Sie sind auf Seite 1von 29
British Journal of Industrial Relations 5 34:2 June 1996 0007-1080 pp. 189-217 Job Satisfaction in Britain Andrew E. Clark Abstract Little recent empirical work in economics or industrial relations has examined job satisfaction, despite its demonstrated correlation with labour market behaviour such as quits, absenteeism and productivity. This paper uses information from a study of 5000 British employees to investigate the relationship between three measures of job satisfaction and a wide range of individual and job characteristics. Notably, men, workers in their thirties, the well-educated, those working longer hours and workers in larger establish- ments have lower levels of job satisfaction. The estimated job satisfaction equations are used to calculate a measure of the shadow wage and to provide some evidence that is consistent with the existence of non-compensating differentials in the industry and occupational wage structure. 1. Introduction Job satisfaction remains a variable that has been relatively little studied in economics and industrial relations, despite a wealth of papers in other social science disciplines such as psychology, sociology and management science." The analysis of job satisfaction is of interest for two reasons. First, it is a measure of individual well-being, and many social scientists would consider the distribution of welfare to be one of their principal concerns. Second, the analysis of job satisfaction may give us a number of insights into certain aspects of the labour market. Workers’ decisions about their labour force participation, whether to stay on at a job or to quit, and how much effort to devote to their job are all likely to depend in part upon the workers’ subjective evaluation of their work, in other words on their job satisfaction. The other side of the labour market consists of firms, who prefer that their workers be satisfied. This paper uses both bivariate and regression techniques to examine the distribution of three different measures of job satisfaction in a recent British data set. The empirical results relate workers’ job satisfaction to individual characteristics, such as gender, age and education, and to job character- istics, such as establishment size, promotion, hours and pay. Males, workers Andrew Clark is with the OECD, DEELSA, in Paris. © Blackwell Publishers Ltd/London School of Economics 1996 Published by Blackwell Publishers Ltd, 108 Cowley Road, Oxford, OX4 1JF, and 238 Main Street, Cambridge, MA 02141, USA. 190 British Journal of Industrial Relations in their thirties, those with higher levels of education, workers in larger establishments, workers without opportunities for promotion and those working longer hours are found to have relatively low levels of job satisfaction. Income is strongly positively associated with one measure of satisfaction — pay — but much less strongly correlated with the overall level of job satisfaction. This finding can be explained by the presence of unobserved omitted variables, which are correlated with both overall satisfaction and income, or by the presence of comparison effects in the job satisfaction function, whereby some characteristics of the worker’s job are compared with the job characteristics of some reference group, or with the workers’ own expectations about their jobs. The observed pattern of job satisfaction responses by gender, age and education is argued to be consistent with the presence of such relativities. Section 2 presents a brief discussion of the concept of job satisfaction. The data and the distribution of job satisfaction are presented in Section 3. The fourth section analyses statistically the relationship between job satisfaction and a number of individual and job characteristics. Section 5 uses the results of the regression analysis to estimate job satisfaction responses for a number of ‘benchmark’ individuals. Section 6 proposes a new methodology, based on these regressions, for the calculation of the extra income necessary to compensate a worker for an extra hour of work (the ‘shadow wage’) and investigates the claim that inter-industry and inter-occupational wage differentials represent rents to workers, and do not compensate for un- observed ability or job characteristics. Section 7 concludes. 2. Job satisfaction and utility The Concept of Job Satisfaction A classic reference for the meaning of job satisfaction is Locke (1976), who traces interest in workers’ subjective well-being back to the ideas of scientific management and fatigue reduction at the beginning of the century. Locke defines job satisfaction as ‘a pleasurable or positive emotional state resulting from the appraisal of one’s job or job experiences’ (p. 1300). Satisfaction, according to different schools of thought, depends variously on the individual’s expectations, needs (physical and psychological) and values. More specifically, work attributes that involve mental challenge (autonomy, complexity, the opportunity to use abilities and to learn new ones, responsibility, etc.) lead to satisfaction, although the relationship is concave: too much mental challenge can bring about dissatisfaction as well as too little. Hence both challenge and some measure of achievement are important. Locke also highlights the link between personal interest and recognition and job satisfaction. The role of pay is discussed at length, from the basis of ‘equity’, ‘discrepancy’ or ‘relative deprivation’ theories, which imply that it is not only the absolute level of pay that matters to workers but © Blackwell Publishers Ltd/London School of Economics 1996. Job Satisfaction in Britain 191 also the level of pay relative either to what they expected or to what others receive. This idea of relative utility is discussed further below. Locke’s preferred formulation of satisfaction is as the weighted sum of the discrepancies between how much of a certain valued aspect of working a job delivers and how much of this aspect the individual desires or expects; the weights in this summation are given by the individual’s valuation of the different aspects of the job. Higher job satisfaction may then come about from improvements in the objective aspects of the job, from reduced expectations or desires regarding the job, or equally from a realignment of values so that dissatisfying aspects of the job are downplayed, while those that please are given greater weight. My reading of most psychological research is that, in line with Locke, job satisfaction is readily identified with individual well-being. In sociology, however, this conclusion is less clear. Tausky (1992) draws a distinction between theories of work as a necessary evil in order to survive and work as an activity pleasurable in itself, although he notably talks of satisfaction as being a motivation towards action in both cases. Hodson (1991) provides a scathing critique of the entire concept of job satisfaction, arguing that it is largely irrelevant for workers’ day-to-day activities and that it is theoretic- ally naive, paying too little attention to the multitudinous emotional states underlying individual well-being; he proposes instead a behaviour-based classification of workers.’ On the other hand, both Hakim (1991) and Veenhoven (1991) accept job satisfaction as an index of individual well- being without any qualms. Economists were relative latecomers to the job satisfaction field. One of the first economic papers on this subject, Hamermesh (1977), uses job satisfaction data, which is considered as a direct index of utility from working, to test a model of occupational choice based on workers’ maximization of their expected lifetime utility. However, Freeman (1978) concludes ‘that subjective variables like job satisfaction ... contain useful information for predicting and understanding behaviour, but that they also lead to complexities due to their dependency on psychological states’ (p. 140). It is probably fair to say that economists’ attitudes towards subjective measures of well-being remain sceptical, although a recent resurgence of interest in the analysis of subjective variables may change attitudes in time. Relative Utility This paper considers individuals’ responses to questions about job satisfac- tion as a proxy measure of their utility from working. Following Clark and Oswald (1996), we define a ‘life satisfaction’ function, v, as v =(u, 11), where u is utility from work. and 4: is utility from the non-work spheres of life. The subject of this paper, job satisfaction, is argued to reflect u, which can be considered as a sub-utility function reflecting the level of well-being that the individual receives from all aspects of his or her job. © Blackwell Publishers Ltd/London School of Economics 1996. 192 British Journal of Industrial Relations There is no agreement in the social sciences about the form that the function u takes. The standard economic model is that individual utility from working depends positively on income, y, and negatively on hours of work, h, and that it depends also on a set of other individual and job character- istics, which are captured in a vector, z: u=u(y, h,2). (1) An alternative specification considers that utility contains a relative component, i.e. that the individual may make comparisons between his/her own level of some of the elements of « and the corresponding level in some reference group. Veblen (1899) was one of the first proponents of such a view in economics, the formal analysis of which is contained in Duesenberry (1949). The idea of individual well-being as depending on some kind of comparison process has heavily influenced social psychology (Adams 1963; Homans 1961) and sociology (Runciman 1966). Suppose that the individual’s own income, y, is compared with that of some reference group, y*. Then an alternative specification of (1) is u=uly, y*, h, 2). Q) ‘The higher is the level of y*, the lower is the individual's own income compared with y*, i.e. the lower is the individual’s relative income, and thus the lower is the individual’s level of utility. Of course, comparisons may take place over any number of job characteristics, and equation (2) serves only to illustrate the general principle of relative utility. However, income is often considered to be one of the most important aspects of the job for the worker, and the level of others’ earnings is more widely known than is, for example, the stress of others’ jobs, the size of others’ offices or their promotion opportunities, about which comparisons could also take place. The empirical testing of equations such as (2) is in its infancy, for two reasons. First, as mentioned above, it is not clear over which elements of the utility function comparisons take place. Second, and more importantly, the researcher almost never has information on how the individual’s y* is calculated; i.e., of whom does this reference group consist? Without this information it is necessary to make some assumptions about the formulation. of y*. Hence any empirical test of relative utility always involves a joint hypothesis: that y* is important, and that the specification used to calculate it is the correct one. Nevertheless, a small body of empirical literature has found evidence consistent with such comparison effects in survey responses to questions on well-being (Clark 1995a,c, Clark and Oswald 1996; Hamermesh 1977; Lévy-Garboua and Montmarquette 1994; and Sloane and Williams 1994), in the income levels that individuals assign to different verbal labels such as ‘excellent’, ‘good’, ‘sufficient’ and ‘bad’ (Hagenaars 1986; Melenberg 1992; and van de Stadt et al. 1985) and in individual savings behaviour (Kosicki 1987). © Blackwell Publishers Ltd/London School of Economics 1996. Job Satisfaction in Britain 193 3. Data The data used in this paper come from wave | of a random sample of approximately 10,000 individuals in 5500 British households. This data set, the British Household Panel Survey (BHPS), includes a wide range of information about individual and household demographics, health, employ- ment, values and finances; for more details see Rose et al. (1991). The data were collected in late 1991. The job satisfaction data come from the following questions asked of all employees in the BHPS. Initially, individuals were asked to rate their satisfaction levels with seven specific facets of their job: promotion prospects, total pay. relations with supervisors, job security, ability to work on their own initiative, the actual work itself, and hours of work. Each of these criteria was to be given a number from | to 7, where a value of 1 corresponded to ‘not satisfied at all’, a value of 7 corresponded to ‘completely satisfied’, and the integers from 2 to 6 represented intermediate levels of satisfaction. After they had rated their levels of contentment with this list of topics, individuals were asked a final question: ‘All things considered, how satisfied or dissatisfied are you with your present job overall using the same 1-7 scale?” One objection that may be raised to the use of job satisfaction responses as measures of individual well-being is that satisfaction is subjective and hence cannot be compared across individuals. How then do we know that the cross-section answers to subjective questions like these contain informa- tion and are not just random draws? A small body of empirical research in economics and psychology has considered this question by relating satisfac- tion scores to subsequent observable labour market behaviour. Freeman (1978) uses American panel data to show that job satisfaction is a significant predictor of quits, with an effect that is, in two of the three data sets examined, at least as powerful as that of wages; Akerlof er al. (1988) and McEvoy and Cascio (1985) reach the same conclusions.‘ Other research has found that job satisfaction is negatively correlated with absenteeism (Clegg 1983) and with non-productive and counter-productive work (Mangione and Quinn 1975). If job satisfaction data were purely idiosyncratic, it is hard to see how the above results could have been obtained. Although it is unlikely that these data are noise-free, the above research results suggest that there is information in its cross-section analysis.* This paper will analyse three of the job satisfaction measures described above: the first, satisfaction with pay, measures the worker’s subjective evaluation of the extrinsic observable monetary reward from working; the second, satisfaction with the work itself, reflects the intrinsic nature of the job; the third, overall job satisfaction, is a useful summary measure.’ Table 1 presents the simple distribution of these three measures. The modal response for all measures of satisfaction is 7, the highest value;’ the median value for overall job satisfaction and for satisfaction with the work itself is 6, while that for satisfaction with pay is 5. There are © Blackwell Publishers Ltd/London School of Economics 1996. 194 British Journal of Industrial Relations TABLE 1 Distribution of Reported Job Satisfaction Levels Pay satisfaction Satisfaction with Overall satisfaction work itself N % N % N % 7 1059-04 248 41.3 1645 31.7 6 5 142 14i 2.0 139% (269 5 888171 ™ 148 995192 4 1045-202 68 134 653126 3 5352. 103 186 3.6 237 46 2 244 47 66 13 90. 17 1 680131 185 36 178 34 5186 100.0 5195 1000 5194 100.0 Notes: These numbers are based on weighted data. The results using unweighted data are almost identical. substantial tails of dissatisfied workers in the distribution, especially for pay satisfaction: 9.7% of workers report overall job satisfaction of 3 or less, with the respective figures for work itself and pay satisfaction being 8.4% and 28.1%. British workers seem well satisfied with their jobs overall, and with the nature of the work itself, but less so with their pay. Table 2 summarizes the bivariate relationships between job satisfaction and a number of worker and job characteristics. Details of all variables are contained in Appendix A. The mean level of the satisfaction score is reported for each characteristic, as is the percentage who are ‘highly satisfied” (reporting satisfaction of 6 or 7 on the 1-7 scale). The majority of these correlations carry over to the multivariate regression analysis of satisfaction, the results of which are reported in Appendix B. As the dependent variable is ordinal (ie., job satisfaction of 6 is not twice as high as job satisfaction of 3), Ordered Probit techniques have been used for these estimations (see Zavoina and McKelvey 1975).* Variables and sets of dummies have been dropped at the 5% significance level in all regressions. TABLE 2 Job Satisfaction Means* Pay Work itself Overall Mean %very Mean %very Mean % very satisfied satisfied satisfied Overall 448 348 S67 63.6 550 587 Individual characteristics Male 433" 312" 5.56% 60.6" = 5.31"* 52.9% Female 465 387 578 66.7 Sm 65.0 16-19 424m 287" 5.40 54.9 3.55% 59.1%* 20-29 427 290 540 556 532 534 30-39 446 336 558 61.0 540 56.1 40-49 452 354 575 66.4 550 586 50-59 472 415 603 74.5 572 663 © Blackwell Publishers Ltd/London School of Economies 1996. Job Satisfaction in Britain 195 60+ S21 S82) 643 85.1 615 155 Health—excellent 460** 373" S81" 6B.2"* 5.68% 65.0" Health—good 448 341 565 62.2 547 $73 Health—other 427 324 546 587 525 509 Education—higher 455 -35.0%* 5.68" 63.79" 53ae* 55.9" Education—A/O/nursing 4446 319 554 60.2 5440 571 Education—other 448 38.2 582 675 5.67 62.9 White/Chinese/Other 449%* 35.0 567 635 ssi* 590 Black O71, 252 557 618 526 509 Indian/Pakistani/Bangladeshi 4.50 32.0 589 673 5.08 507 Married 4624 38.7" S.76** 66.3%* 5.56%" 60.6" Separated 427301 57 68.6 551 648 Divorced 440 33.2 $82 685 555 607 Widowed 5.07 498 6.48 90.1 6.25 79.7 Single 420 268 542 559 533 $3.2 Renter 435** 34.0 5.72" 65.1 561 61.8* Non-renter 452 35.0 565 63.2 347 $79 Inner London 40s 277" 5.48 583* 5.068" 489 Outer London 441 313 552 582 5.49 $79 Rest of South East 456 35.2 568 633 533 597 South West 460 373 568 63.9 556 608 East Anglia 444 370 572 604 5359 597 East Midlands 434 333 578 643 556 595 West Midlands conurbation 4.37.—=35.7 5B 639 561 593 Rest of West Midlands 428 312 575 66.7 545 55.8 Greater Manchester 453 33.7 565 640 350 577 Merseyside 435 286 $56 598 554 575 Rest of North West 468 422 567 65.1 556 615 South Yorkshire 435 300 $52 602 535 56.0 West Yorkshire 443 344 586 70.4 545 614 Rest Yorkshire & Humberside 4.53 39.3 579 66.7 560 616 Tyne and Wear 459 385 337 619 Sai 530 Rest of North 468 398 566 640 562 613 Wales 446 30.1 S73 647 558 611 Scotland 467 383 5.66 63.0 546 577 Work values Promotion 3.82" 178+ 5.30"* 51.88% 5.23%" 46.588 Pay 415 276 517 489 501 433 Relations at work 4.65 411 $.95 68.7 5.89 69.6 Job security 449 35.7 569 633 559 60.6 ‘Actual work itself 454 35.0 582 693 5353 614 Hours 5.00 49.9 590 702 594 681 Initiative/somethingelse 4.69 37.2 577 681 556 616 Job characteristics Monthly pay <£500 4.62" 41.0% —5.81** 67.18% 5.84%" 68.3" £500-£999 413 269 5438 586 $35 $52 £1000-£1499 443 303 563 613 S31 526 £1500+ 499 484 5.79 66.7 548 582 <16 hours pw S10" 49.4% 5.83% 69.0% 5.93" 70.0" 16-30 hours pw 585 67.7 584 683 30-40 hours pw 535 598 536 545 >40 hours pw 573 66.1 S41 56.3 Agriculture S92e* TB7* 5.886% 62.0%" Energy 537 596 534 568 Extraction S41 56.7 535 541 Engineering 559 621 S15 487 Other manufacturing 5.40 58.7 524 493 Construction 388 673 548 57.0 © Blackwell Publishers Ltd/London School of Economics 1996. 196 British Journal of Industrial Relations Distribution 4a5 35.2 562 633 552 60.2 Transport 438327 555 S74 $19 513 Finance 468 35.7 348 55.1 549 357.4 Other services 4.55 36.6 5.91 70.8 5.75 663 Managers 484°" 41.24 6,02" 75.8%" 5.60%" 64.18 Professionals 43 581 659 544 55.8 Associate professionals 456 5.89 70.0 557 605 Clerical 4.66 548 S71 551 59.2 Craft 4.09 5.60 60.9 5.29 516 Personal 447 601 75.4 585 694 Sales 4.40 558 61.0 562 60.7 Plant operative 4.03 Isai: 1952 509 474 Other 477 547 572 560 613 Establishment size: 1-24 452 374" -S.86M* 68.6" 5.698 63,984 Establishment size:25-199 4.43 32.8 562 624 345575 Establishment size: 200+ 451 342 5.48 58.7 5.32 53.6 Union 445 342 5.60" 62.3 5388" 54,5¢* Non-union 450 49 572 646 336 60.7 Incentive payments 4.62"* 362 537 60.9% $44. 57.0 No incentive payments 443 2 510 645 552 593 Promotion opportunities 4.60"* 35.1 568 63.2 5.58" 60.2" No promotion opportunities 4.36 34.4 565 639 542 572 Manager 456 352 586"* 69.0" 555 603 Non-manager 444 34S 556 60.6 3a? 579 Temporary/contract job 446 36.4 357 62.6 342 578 Permanent job 449 346 568 63.9 551 587 Second job 429% 31.0 357 603 542 570 No second job 450 352 568 63.9 551 589 *‘Very satisfied’ denotes job satisfaction of 6 or 7 on the 1~7 scale, For each set of characteristics, the asterisks denote the significance level of the F-test that the means are not identical: * significant at the 5% level and ** significant at the 1% level. All figures refer to weighted data, 4, Individual and job characteristics Gender Table 2 shows that nearly two-thirds of women report overall job satisfac- tion of 6 or 7, compared with just over one-half of men; a similar difference exists for satisfaction with pay and with the work itself. Moreover, when other characteristics are controlled for in the regressions, the variable ‘male’ attracts a significant negative coefficient for each of the measures of satisfaction examined.’ This is surprising for two reasons. First, it is well established that there is sex discrimination in the British labour market (see, among others, Brown Johnson et al. 1992; Equal Opportunities Commission 1992; and Wright and Ermisch 1991). Second, when life satisfaction/ subjective well-being scores are analysed, women report higher levels of stress than do men (see Clark and Oswald 1994; Lane 1991). There are a number of plausible explanations for this finding. First, men and women workers differ both in terms of the types of jobs that they do and in terms of their personal characteristics (for example, working men have © Blackwell Publishers Ltd/London School of Economics 1996. Job Satisfaction in Britain 197 rather different qualifications from working women and work longer hours). Second, from the responses to the BHPS question on the aspects of a job that the respondent finds important, it is apparent that men and women value work for different reasons.” Third, there is a participation effect: for cultural reasons, women who are dissatisfied at work may find it easier to leave the labour force than their male equivalents; thus, satisfied women workers may be a statistical construct, as more of the women who would be dissatisfied at work are not working. Last, men and women may answer job satisfaction questions in different ways because, although their objective job characteristics may be the same, their expectations of what their job should be like, or the reference groups against which they compare, are different. The first two explanations, controlling for individual and job character- istics and work values, are tested in the Ordered Probit regressions in Appendix B. As can be seen, the estimated coefficient on males remains negative and significant for all three measures of job satisfaction. The remaining hypotheses are discussed in Clark (1995b). The formal testing of a selection effect is complicated in an Ordered Probit model, but the preliminary results do suggest that the selection effect helps to explain women workers’ higher job satisfaction. The last explanation proposes that men’s and women’s responses to job satisfaction questions may differ because they use different expectations or reference groups to evaluate their jobs. For example, if it is true that women are socialized into expecting less from their employment, then they will be more satisfied than a man with any given job. It is not easy to test for such an effect, as most surveys contain no information about individual expectations or reference groups. (The data analysed by Melenberg (1992) are an exception.) The inclusion of measures of comparison income into job satisfaction regressions using the BHPS data set (Clark 1995a, c; and Clark and Oswald 1996) does not explain the lower levels of job satisfaction reported by male workers. However, Clark (1995b) finds that younger women, professional women and higher-educated women report the same, or Jower, levels of job satisfaction as their male counterparts, which is consistent with cohort, experience and education effects on women’s expectations regarding their jobs." Age A significant relationship between job satisfaction and age is apparent in Table 2’s cross-tabulations. Satisfaction with pay and satisfaction with the work itself rise nonlinearly with age, with larger rises in satisfaction for the older age groups. There is some evidence of a U-shaped relationship between overall job satisfaction and age, with those in their twenties or thirties being the least satisfied. For all three measures, workers aged sixty or over are the most satisfied, followed by workers in their fifties. When other variables are controlled for in the regressions, a slightly different picture emerges. Here, both age and its square are entered as explanatory variables to reflect the nonlinear relationship suggested by the © Blackwell Publishers L1d/London School of Economics 1996. 198 British Journal of Industrial Relations cross-tabulations. The significant negative coefficient on age and the positive coefficient on age-squared in all three regressions imply a U-shaped relationship between age and job satisfaction. The minima of these U- shapes are at ages 36, 27 and 34 respectively for satisfaction with pay, satisfaction with the work itself and overall job satisfaction. The relationship between job satisfaction and age in the BHPS data set has been explored by Clark et al. (1996), who find that explanations based on matching, whereby older people have had more time in the labour force to find a good job, and on a cohort effect, whereby workers born in the 1950s have always had low satisfaction, are not supported by the data. Older workers’ higher levels of job satisfaction could result from a participation effect, as for the gender difference in job satisfaction discussed above. The idea that dissatisfied older workers find it easier to leave the labour market is a natural one. However, early retirement probably starts to become numerically important only in the fifties and sixties, whereas satisfaction starts to rise in the thirties. The participation effect probably helps to explain why workers over the age of fifty have such high levels of job satisfaction, but does not provide an underpinning for the whole of the U- shaped relationship. The conclusion proffered by Clark et al. (1996) is in terms of workers’ perceptions of their job in relation to their job expectations. The U-shape may be explained not by individual or job characteristics, but by changing expectations over time. Young workers may feel satisfied because of the novelty of their situation and because they have little information about the world of work with which to evaluate their job. As they become older they become able to make this comparison, and it may be that this explains the drop in satisfaction towards the mid-thirties. The subsequent rise in satisfaction up until the age of retirement probably partly results from the participation effect discussed above; it could also come from older workers’ reduced aspirations as they realize that they have fewer alternative jobs open to them, or even from the reduced importance that older workers might attach to such aspirations.” Health There is a strongly significant positive relationship between self-reported physical health and job satisfaction in Table 2. This could reflect that workers in poor health have a tendency to report low levels of satisfaction with all aspects of their life, or that they can only obtain relatively unsatisfying jobs. As the introduction of the job description (hours, income, promotion opportunities, managerial responsib so on) in the regressions does not remove the health effect, it seems likely that it is the propensity of those in poor health to be less satisfied that is driving this correlation. ‘© Blackwell Publishers Ltd/London School of Economics 1996. Job Satisfaction in Britain 199 Education The effect of education on labour market behaviour has been heavily researched. The results show that those with higher levels of education earn more (see e.g. Blanchflower and Oswald 1994), are promoted more quickly and, in general, end up with better jobs. In light of these findings, the predicted correlation between education and job satisfaction should be unambiguously positive.'? Table 2’s results do indeed indicate that there is a strong relationship between job satisfaction and education, but of the opposite sign. For all three measures of job satisfaction, the percentage of those claiming to be very satisfied is greatest for the group with the lowest level of education. There is weak evidence that mean satisfaction with pay increases with education, but those with the highest level of education report mean overall job satisfaction of 5.38, against a figure of 5.44 for those with intermediate education and 5.67 for those in the lowest educational group. When other explanatory variables are included, higher levels of education are unambiguously associated with lower levels of all three measures of job satisfaction. One interpretation of these results is presented in Clark and Oswald (1996), where it is suggested that, even though better educated workers have better jobs, education is positively correlated with workers’ expectations of what kind of job they should have. The causal mechanism of this relationship is ambiguous: the process of education could itself raise workers’ expectations, or those who already have high expectations (influenced by their parents or their early schooling, for example) could be more likely to continue their education. The expected correlation between education and job satisfaction becomes ambiguous once such comparisons are taken into consideration. The statistical results discussed above show that the ‘comparison’-level effect associated with education seems to outweigh the positive effect through the type of jobs that better educated workers have; thus, workers with higher levels of education report themselves as relatively dissatisfied. Race The cross-tabulation results show that whites report the highest level of overall job satisfaction and that blacks appear to be relatively dissatisfied with their pay.’ In the regressions, the only significant racial effect is that workers from the Indian sub-continent are more likely to be satisfied with their work itself. The scarcity of significant results here is partly due to the relatively small number of ethnic minority workers interviewed (68 black and 71 Indian sub-continent workers). Marital Status Marital status is strongly correlated with overall job satisfaction and ‘© Blackwell Publishers Ltd/London School of Economics 1996. 200 British Journal of Industrial Relations satisfaction with the work itself, both in the cross-tabulations and in the regression equation. Married workers report the highest level of overall job satisfaction, but the most satisfied group with respect to the work itself are the widowed. This latter correlation is of note because widow(er)hood is probably the only marital status involving no personal choice, and thus can be considered as completely exogenous to the individual.’ Those who are widowed may value more highly the social contacts that work brings, which is consistent with the large and significant coefficient on widow(er)hood appearing in the satisfaction with the work itself equation. Housing Tenure Renters are more satisfied at work than home owners or home buyers. This could reflect ease of mobility between jobs. Individuals cannot always, for a variety of reasons, change from one job to another, even if they would find such a change satisfying; house ownership or purchase could represent an obstacle to geographic job mobility, whereas renters are presumably more mobile, and thus more able to leave unsatisfying jobs. Alternatively, if renting is seen as a proxy for social class, and thus for the individual’s reference group, renters could make comparisons against a reference group with worse jobs, and thus report higher levels of job satisfaction. Region Workers in London, especially inner London, are relatively dissatisfied with their jobs. This effect is most pronounced for satisfaction with pay. The estimates from the overall job satisfaction equation imply that, ceteris paribus, the most satisfied workers are found in the ‘rest of the North’ and the least satisfied are found in inner London and Scotland. Overall there is little evidence of a North-South divide in job satisfaction. One potential explanation is that satisfaction is lower when the price level in the region is higher. To test this hypothesis, the regional dummies were replaced by a regional price variable. Regional prices were insignificant in the overall job satisfaction equation but were negative bordering on significance (at the 7% level) in the satisfaction with the work itself equation, and negative and significant at the 0.1% level in the pay satisfaction equation. However, examination of the log-likelihoods from the two specifications reveals that the regional dummies fit the data better than does the regional price variable. The results thus offer some support for the hypothesis that job satisfaction is lower in high price regions, but there appear to be other, unmeasured, regional characteristics which contribute to workers’ reported levels of job satisfaction. Work Values All respondents who were active in the labour force were asked to choose their first most important aspect of a job from a list of seven specific job © Blackwell Publishers Ltd/London School of Economics 1996, Job Satisfaction in Britain 201 attributes, the same as those highlighted for the job satisfaction questions above, and an eighth catch-all category, ‘something else’. For the regression analysis this last category and ‘initiative’ are excluded, leaving six dummy variables. These turn out to be very strong predictors of job satisfaction. The regression results show that workers who say that promotion opportunities or pay is the most important aspect of a job report substantially lower levels of job satisfaction. The estimated effects are large: choosing pay as the most important aspect of a job has a comparable negative effect on job satisfaction to being male in two out of the three regressions. On the other hand, there is some evidence that those who choose hours of work or relations at work report higher levels of satisfaction. Income and Hours The cross-tabulation results do not provide strong evidence for the hypothesis that high pay is associated with higher levels of job satisfaction. Although the highest paid group report the highest mean level of pay satisfaction, for the other two measures it is the lowest paid who have the highest satisfaction. Once other variables are controlled for in the regres- sions, income is strongly positively correlated with satisfaction with pay, exhibits a weaker positive correlation with overall job satisfaction, and is uncorrelated with satisfaction with the work itself."* It may be surprising that pay has a relatively weak effect on overall job satisfaction. It is possible that this result stems from the unobservability of some variables which are correlated with both income and satisfaction. If one of the functions of pay is to compensate workers for the difficulty of their job, then higher-paid workers may be doing harder jobs, and therefore will not necessarily be more satisfied."” An alternative explanation relies on the idea that income is evaluated relative to some comparison level and not in an absolute sense, as discussed in Section 2. Absolute income may then act as a poor measure of relative income, which would expiain its relative weakness in predicting overall job satisfaction. Previous work using the BHPS data set has found strong correlations between job satisfaction and comparison income, this latter being measured by an econometrically predicted ‘going rate’ for the job (Clark and Oswald 1996), the pay of other workers in the same household (Clark 1995a) or the pay that the respondent has received in the past (Clark 199Sc). These comparison income measures are shown to be negatively correlated with overall job satisfaction, suggesting that the income compar- isons are important determinants of workers’ reported well-being." Weekly hours of work are very strongly negatively correlated with satisfaction with pay and, less strongly, with overall job satisfaction. Industry and Occupation In general, workers are relatively satisfied in agriculture and dissatisfied in engineering. Workers in ‘other services’ report low levels of satisfaction with © Blackwell Publishers Ltd/London School of Economics 1996. 202 British Journal of Industrial Relations pay but are one of the most satisfied groups overall. A compensating differentials interpretation is that workers in ‘other services’ are being paid badly because there are other (unobserved, in this data set) aspects of their jobs that make them satisfied. Those at the higher end of the occupational scale, professionals and managers, report higher satisfaction with their work itself but are relatively dissatisfied with their pay. The same pattern holds for craft and sales workers. Personal sector workers are the most satisfied overall according to the regression estimates, and plant operatives are the least satisfied. Establishment Size The most satisfied workers are to be found at smaller establishments (see Idson 1990). This effect is weaker for satisfaction with pay because we know that larger firms (to which larger establishments must belong) pay higher wages (see Brown and Medoff 1989). The effect of establishment size is strongest for satisfaction with the work itself. A stratification approach was adopted to test the hypothesis that this establishment size effect is weaker for those who value the pay and promotion opportunities that larger, more hierarchical, establishments provide. Separate regression equations were estimated for those who considered pay, promotion or hours (extrinsic characteristics of the job) as the most important aspect of working, and for those who chose relations at work or the work itself (intrinsic characteristics of the job). The establish- ment size dummies were insignificant for the former group, but (except for satisfaction with pay) very significant for the latter. This suggests that part of the attraction of small establishments comes from their provision of intrinsic rewards, which raise significantly the job satisfaction of those workers who value such job characteristics.” Union Membership A negative correlation is found between union membership and all three measures of job satisfaction; this relationship has been one of the main points of interest of the small extant empirical job satisfaction literature (Borjas 1979; Freeman 1978; Meng 1990; and Miller 1990). There is an obvious issue of endogeneity here, since, if unions address issues of worker dissatisfac- tion, the more dissatisfied workers will be the most attracted by union membership. Another potential explanation relies on the fact that unions, by providing workers with a voice (see Freeman 1980), encourage them tostay in jobs they dislike and to try to change their working conditions.” Incentive Payments, Promotion Opportunities and Managerial Responsibilities First, the BHPS data set includes information on whether the employee’s pay ever includes incentive bonuses or profit-related pay. In Table 2, such ‘© Blackwell Publishers Ltd/London Schoo! of Economics 1996, Job Satisfaction in Britain 203 incentive payments are associated with higher pay satisfaction but lower levels for the other two satisfaction measures. However, when the other variables are controlled for in the regressions, there is no significant effect of incentive payment systems on any of the measures of satisfaction considered and so no estimates are reported. This finding is consistent with that of Marsden and Richardson (1994), who argue that the role of these compensation systems in bringing about employee satisfaction does not appear promising. Variables measuring the worker's position, or potential position, in the firm’s hierarchy are strong predictors of job satisfaction. The availability of opportunities for promotion has a positive effect on all three measures of job satisfaction: the increase in overall job satisfaction for having promotion opportunities outweighs the lower satisfaction associated with working in the largest (200 workers or more) establishment or being a union member. Managerial responsibilities are strongly positively correlated with satisfac- tion with the work itself, but show no relationship with the other two satisfaction measures. Temporary or Contract Work The flexibility of the work-force has been a much-discussed topic in industrial relations since the mid-1980s (see e.g. Hunter et al. 1993). One aspect of this flexibility has been the growth in the number of workers on temporary or fixed-term contracts, a change that has sometimes been Suggested not to be in the workers’ interests. However, no effect of such status on workers’ job satisfaction is found: the cross-tabulations in Table 2 show that temporary or contract workers report levels of job satisfaction similar to those of workers on permanent contracts. Further, the variable for temporary or contract work was insignificant in all three regression equations.” Second Job The image of the worker with a second job is not that of someone who is necessarily satisfied with the world of work. Table 2 shows that workers with a second job do indeed report lower levels of job satisfaction, but only significantly so for pay satisfaction. The second job variable is insignificant for all three measures of job satisfaction once the other explanatory variables are controlled for in Appendix B. 5, Predicted job satisfaction The results of the regressions reported in Appendix B may be used to predict the probabilities with which workers with certain types of character- © Blackwell Publishers Ltd/London School of Economics 1996. 204 British Journal of Industrial Relations istics will report different levels of job satisfaction. As an illustration, Table 3 presents the estimated probabilities that four ‘benchmark’ individ- uals will report job satisfaction of 7, the highest level, or of 6 or 7 (and are thus ‘highly satisfied’ according to the definition used in Table 2). A brief description of each individual is given in the table; the full details are provided at the foot of Appendix A. TABLE3 Estimated Percentage Probabilities of Reporting High Job Satisfaction Satisfaction Satisfaction Overall job with pay withwork satisfaction itself 7 60r7 7 60r7 7 60r7 1. Female, 25, single, £1300 pem, 273 441 35.7 598 = 287 58.1 40 hours pw, education high 2. Male, 35, married, £2000 pem, 27 3S 52.7 153 26.9 55.9 45 hours pw, education high 3. Female, 45, married, £1600 pem, 306 479 446 «684 = 32.5 62.2 35 hours pw, education medium 4. Male, 55, divorced, £1300 pem, 181 325-409 65.0 = 143 38.1 38 hours pw, education low. Table 3 shows that all four individuals are estimated to be more likely to report high satisfaction with work itself than high satisfaction with pay or high overall job satisfaction, as Table 1 suggests. The third individual typically has the highest predicted probability of the four of reporting high levels of job satisfaction, as she combines a number of favourable characteristics. Marginal effects on these estimated probabilities can be calculated by changing the characteristics of one of the individuals, say individual 3. All of the following figures apply to overall job satisfaction. A change of gender from woman to man reduces the estimated probability of responding 7 from 33% to 24%; a high level of education reduces this probability to 28%, and an increase in weekly hours from 35 to 45 has only a small effect, cutting the probability to 30%. On the other hand, a widow with the same character- istics as individual 3 is estimated to have a 37% probability of reporting overall job satisfaction of 7, while increasing individual 3’s age from 45 to 60 increases this probability from 33% to 46%, and that of being highly satisfied from 62% to 74%. 6. Compensating differentials and wage rents A lively literature has developed recently concerning the existence of compensating differentials in the labour market: are workers paid more to © Blackwell Publishers Ltd/London School of Economics 1996. Job Satisfaction in Britain 205 undertake jobs with certain unpleasant characteristics, and if so how much? One specific facet of this research has examined the extra pay that would compensate workers for supplying one more hour of labour, in other words the shadow wage rate. The shadow wage, which is essential for an understanding of labour supply and the structure of pay, has typically been estimated with wages and hours variables from cross-section or panel data. This paper proposes an alternative methodology whereby we consider changes in wages and hours that leave a worker's job satisfaction un- altered. A related area of research has looked at the inter-industry and, to a lesser extent, inter-occupational structure of wages. It is often observed that the standard set of explanatory variables cannot adequately explain wages, in that industry or occupational dummies remain significant in the estimated wage equation. One interpretation of this finding is that these dummies represent compensating differentials, ic. unmeasured aspects of the job or worker which differ across industries and occupations and for which workers are compensated. This paper performs a simple test of this supposition using the estimated job satisfaction equations. The results suggest that inter-industry and inter-occupational wage differentials are Positively associated with job satisfaction differentials, and thus may well represent rents to workers, rather than compensating workers for the type of job that they do. Shadow Wages The regression results can be used to estimate the shadow wage in an obvious way: if an individual works one more hour, how much would his/ her pay need to rise to leave his/her job satisfaction unaltered? Ideally this calculation would be performed on overall job satisfaction, but the weakness of the correlations between pay and overall job satisfaction, especially in the sub-sample who do not wish to change their hours of work (analysed below), precludes its use in estimating the shadow wage. However, both hours and pay are strongly associated with pay satisfac- tion, and it is the results from this regression that provide the estimates below.” One problem that has dogged many previous estimates of the shadow wage can be resolved thanks to the richness of the BHPS data set. The calculation of the shadow wage requires that workers be on their labour supply curve. However, the discreteness of the wage and hours distribution that firms offer implies that this condition often does not hold. All BHPS working respondents are asked if, given their current hourly wage, they would prefer to change their weekly hours of work: 30% of workers said that they would prefer to work fewer hours, and 9% would prefer to work longer hours. The shadow wage calculations reported below use data from the remaining 61% of the sample. For this sub-sample, an increase of one hour of work per week (from the © Blackwell Publishers Lid/London School of Economics 1996. 206 British Journal of Industrial Relations mean level of 29 hours per week) can be compensated by a rise in gross monthly income of £37, giving a shadow wage rate of £8.60 per hour (ph) compared with this sub-sample’s actual average hourly earnings of £5.40 per hour.” The same regression with income and hours interactions by gender reveal that men’s estimated shadow wage (£10 ph; actual average earnings £6.40 ph) is higher than that for women (£6.75 ph; actual average earnings £4.60 ph). Interactions by level of education produce estimates of £11 ph for the higher-educated group, £7.50 ph for workers with a medium level of education and £6.50 ph for workers with the lowest level of education (compared with their actual average earnings of £7.60, £5 and £4.20 ph respectively).> Rents in Industry and Occupational Wages Most studies of the wage structure have found that workers in some industries or occupations systematically earn more than their individual and job characteristics would predict (see Gibbons and Katz 1992; and Hwang et al. 1992). The consequent discussion has focused on whether these wage differentials are compensating, reflecting unmeasured aspects of the worker and the job, or non-compensating, in which case they are referred to as rents. This section uses the results from job satisfaction and pay regressions to present some preliminary evidence that these wage differentials are non- compensating. To aid intuition, consider a wage equation that regresses log income (y) on a series of individual and job characteristics, the latter including a set of industry dummies, I, (the , subscript indicating that these dummies are in the wage equation). The estimation of the wage equation yields estimates B.,, for these industry dummies. The explanatory variables also include a set of occupation dummies, O,, with analogous estimated coefficients B,,. The estimated coefficients on these industry and occupation dummies represent the part of wages that cannot be explained by other explanatory variables, but which varies systematically by industry and occupation. Similarly, the estimation of a job satisfaction equation yields estimates of B,,, and By,, for the industry and occupation dummies respectively. If wage differentials are non-compensating, then higher wages should be reflected in higher worker job satisfaction; that is, there should be a positive correlation between B,, and 8,,,, and between 8, and 8,,,. On the other hand, if wage differ- entials are compensating, there should be no such relationship. As we are interested in whether different levels of wages by industry bring about job satisfaction, wages are not controlled for in these satisfaction regressions. (To do so would evaluate job satisfaction at the same wage.) This test was carried out using the BHPS data. Both the wage equation and the job satisfaction equation contained a rich array of explanatory variables.” The relationship between the industry coefficients in the two regressions was analysed by the regression of one set of industry estimates on the other: 8,,, = a + yB,,. The same technique was used for the © Blackwell Publishers Ltd/London Schoo! of Economics 1996. Job Satisfaction in Britain 207 analysis of the relation between the occupation dummy estimates. Indus- try and occupation were defined at the two-digit SIC and SOC level respectively, giving 58 observations by industry and 75 by occupation. These observations were then weighted by the number of respondents in each two-digit group. The resulting estimates of y, with t-statistics in parentheses, are presented in Table 4. The estimates of y are positive and significant in every case (bar the industry estimate for satisfaction with work itself): industries and occupations that ‘inexplicably’ pay more also have workers who are ‘inexplicably’ more satisfied with their jobs. This result is consistent with the hypothesis that inter-industry and inter-occupational wage differentials are not compensating differentials. However, the causality of this correlation is not clear, and it could be the case that more satisfied individuals make better workers and are rewarded accordingly, which need not imply the existence of wage rents. The definitive test between these two hypotheses would require either much better information on worker productivity than is typically available, or the analysis of panel data. In the latter case, a good worker who is satisfied should be well rewarded even if she changes industry or occupation: panel data thus distinguishes between explana- tions that are worker-based and those that are industry- or occupation- based.” TABLE4 Wage Rents and Job Satisfaction Rents: Estimates of Satisfaction Satisfaction Overall with pay with Work Satisfaction Iself By industry 1.56(33.0) 0.03 (0.9) 0.35 (8.3) By occupation _ 0.33 (16.0) 0.61 (29.2) 0.15 (7.1) 7, Conclusion This paper has provided an up-to-date empirical analysis of job satisfaction in Britain. The satisfaction levels of some 5000 workers have been related to a wide range of their personal and job characteristics. The results have shown that job satisfaction is higher for women, older workers, and those with lower levels of education. It has been argued that this finding may be explained both by the different levels of labour force participation and job expectations and by comparisons by sex, age and education. The types of job that workers have are also strong predictors of job satisfaction. Workers with long hours, those in large establishments, union members and those without promotion opportunities are more likely to be dissatisfied at work. Income is strongly positively correlated with pay satisfaction and much less strongly with overall job satisfaction. © Blackwell Publishers Ltd/London School of Economics 1996. 208 British Journal of Industrial Relations In addition, this paper has suggested that the use of data on workers’ subjective evaluations of their job may give new insights into a number of active areas of labour market research. Compensating differentials can be calculated as the extra income that would compensate a worker, i.e. leave him just as satisfied, for dissatisfying aspects of his job. The average worker would need to receive an estimated extra £8.60 to compensate for an extra hour of work. This shadow wage is higher for men and for those with higher levels of education. A growing literature has considered whether inter- industry or inter-occupational wage differentials are compensating or whether they represent rents to workers. A positive relationship was found between the estimates on industry/occupation dummies in a wage equation and their counterparts in a job satisfaction equation: industries/occupations that pay more than expected have workers who are more satisfied than expected. This finding is consistent with the interpretation of these inter- industry and inter-occupational wage differentials as rents. Appendix A Variables and definitions Overall job satisfaction: Scaled 1-7 where 7 is the highest category. Job satisfaction with pay: Scaled 1-7 where 7 is the highest category. Job satisfaction with the work itself. Scaled 1-7 where 7 is the highest category. Male dummy: Respondent is male. Age: Age of respondent at date of interview. Health dummies (3): Respondents classify their own health, compared with people of their own age. Categories: Excellent; Good; Fair to very poor. Omitted category: fair to very poor. Education dummies (3): High — degree, teaching qualification or other higher qualification; Medium — nursing qualification, A-levels, O-levels or equivalent; Low — neither of the above. Omitted category: Low. Indian dummy: Respondent considers that they below to the Indian, Pakistani or Bangladeshi racial group. Black dummy: Respondent considers that they belong to the black Caribbean, black African or other black racial group. Marital status dummies (5): Married; Separated; Divorced: Widowed: Never Married. Omitted category: Never Married. Renter dummy: Respondent lives in rented or rent-free accommodation. Region dummies (18): Standard regions plus seven metropolitan areas. Work values dummies (7): Aspect of a job which respondent finds most important: Promotion prospects; Pay; Relations at work; Job security; The actual work itself; Hours; and Using your initiative or ‘something else’. Omitted category: Using your initiative or something else. Log income: Natural log of usual monthly gross pay from respondent’s main job. © Blackwell Publishers Lid/London School of Economics 1996. Job Satisfaction in Britain 209 Log hours: Natural log of usual weekly hours (excluding overtime). Industry dummies (10): Agriculture, forestry and fishing; Energy and water supplies; Extraction of minerals and ores other than fuels, manufacture of metals, mineral products and chemicals; Metal goods, engineering and vehicles industries; Other manufacturing industries; Construction; Distribution, hotels and catering; Transport and com- munication; Banking, finance, insurance, business services and leasing; Other services. Omitted category: Other services. Occupation dummies (9): Managers and administrators; Professional; Associate professional and technical; Clerical and secretarial; Craft and related; Personal and protective service; Sales; Plant and machine operative; Other. Omitted category: Other. Establishment size dummies (3): Number of workers at establishment is <25; 25-199; 200+. Omitted category: 200+. Union member dummy: Respondent is a member of a recognized union at his/her workplace. Incentive payments dummy: Respondent's pay has included incentive bonuses or a profit-related element. Promotion opportunities dummy: Respondent has opportunities for pro- motion in his/her current job. Manager dummy: Respondent has some managerial or supervisory duties. Temporary or contract work dummy: Respondent’s current job is sea- sonal, temporary, casual or a job done under contract or for a fixed period of time. Second job dummy: Respondent earns money from a second job. Characteristics used for the predictions in Section 5 In addition to the characteristics described in the text, the following at- tributes were used to calculate the predicted probabilities. All individuals are home owners. Individual I; health excellent; white; West Yorkshire; pay most impor- tant; Other Services; Clerical; establishment size 25-199; non-union; promotion opportunities; manager. Individual 2: Health good; black; West Midlands conurbation; work itself most important, Construction; Manager; establishment size 200+; non-union; promotion opportunities; manager. Individual 3: health good; white; Inner London; job security most important; Other manufacturing; Personal; establishment size 25-199; non-union; promotion opportunities; manager. Individual 4: health other; white; Scotland; ‘something else’ most important; Transport; Plant operative; establishment size 1-25; union; no promotion opportunities; not a manager. © Blackwell Publishers Lid/London School of Economics 1996. 210 British Journal of Industrial Relations Appendix B Table BI lists the Ordered Probit job satisfaction regressions TABLE B1 Ordered Probit Job Satisfaction Regressions Satisfaction Satisfaction Overall job with pay with work satisfaction itself Individual characteristics Male 0.29%" 0.184 ~0.26** (0.04) (0.04) (0.04) Age 0.050" 0.024 -0,043"* (0.009) (0.010) (0.010) Age-squared/1000 0.69"* 0.47** 0.62" (0.11) (0.12) (0.12) Health Excellent 022+ 0.27 0.40%* (0.05) (0.05) (0.05) Good 0.15** ‘0.10* 0.18** (0.04) (0.04) (0.04) Education Higher -0.12* 0.34" 0.328 (0.05) (0.05) (0.06) A/O/Nursing ~0.10* -0.19** 0.21" (0.04) (0.04) (0.04) Race Black 0.08 (0.15) Indian 035+ (0.14) Marital status Married 0.06 o.1ae (0.05) (0.05) Separated -0.12 0.03 (12) (0.13) Divorced 0.05 0.04 (0.08) (0.08) Widowed ae 0.27 (0.16) (0.15) Renter 0.09* ose (0.04) (0.04) Region dummies (18) Yes Yest Yes* Work values Promotion -0.30"* ~0.22" 022+ (11) (11) Pay 17 0.29% (0.07) (0.07) Relations at work 0.06 0.20 (0.08) (0.08) Job security 0.002 0.10 (0.06) (0.06) Actual work itself 0.04 0.01 (0.06) (0.06) Hours 0.20 0.11 (0.10) (@.1) © Blackwell Publishers Ltd/London School of Economics 1996. Job Satisfaction in Britain 211 Job characteristics Log income 0.60** 0.09" (0.04) (0.04) Log hours -0.89+« 025+ (0.06) (0.06) Industry dummies (10) Yes* Yest* Occupation dummies (9) Yest* Yest Establishment size: 1-24 o21+* (0.05) Establishment size: 25-199 0.07 (0.04) Union 0.10" 0.134" (0.04) (0.04) Promotion opportunities O.17* o24e* (0.04) (0.04) Manager Mu(1) -1.29 -2.73 Mu (2) ~1.07 Mu (3) 0.70 Mu (4) 0.12 Mu (5) 0.36 Mu (6) O81 N 4299 Log-likelihood 7783.2 7098.7 Log-likelihood at zero 8050.5 ~7406.3 Notes: * denotes significance at the 5% level and ** significance at the 1% level, Standard errors in parentheses, Final version accepted 8 March 1995. Acknowledgements This paper refers in part to joint work with Andrew Oswald and Peter Warr. I am grateful to them both for their enthusiasm and insights. I also thank Peter Dolton, David Gray, Dan Hamermesh, John Treble, Ruut Veen- hoven, Frances Woolley and two anonymous referees for helpful comments. The data used in this paper were made available through ESRC Data Archive. The data were originally collected by the ESRC Research Centre on Micro-social Change at the University of Essex. Neither the original collectors of the data nor the Archive bear any responsibility for the analyses or interpretations presented here. Part of this research was carried out at CEPREMAP, Paris, and DELTA, Paris, whose hospitality is gratefully acknowledged. This work was supported by the European Union. The views expressed in this article are those of the author and do not represent those of the OECD or of its member countries. Notes 1, Locke (1976) notes that a 1957 survey included 1795 references to papers on job satisfaction, and estimates that the corresponding 1976 figure was over 3350. © Blackwell Publishers Ltd/London School of Economics 1996. 212 British Journal of Industrial Relations 2. Locke examines in detail Maslow’s need hierarchy theory and Herzberg’s Motivator-Hygiene theory. 3. This distinction between behaviour and job satisfaction diminishes in light of the esearch showing links between job satisfaction and labour market behaviour. Hodson’s point that workplace setting and the worker's reaction to it are important is a good one, but it is not clear how his resulting classification of worker types could be implemented empirically. The 1973 GHS report showed that those who were very dissatisfied with their job were thirteen times more likely to report an intention to quit than those who were very satisfied. 5. Recent methodological defences of the analysis of subjective variables have been made by Harsanyi (1986), Tinbergen (1991) and Van Praag (1991). 6. This claim can be confirmed by factor analysis. 7. Data from the 1989 British Social Attitudes survey, the General Household Surveys, the 1989 ISSP, which holds comparable individual-level labour market data for 11 countries, and the US General Social Surveys show similar bunching. . Positive coefficients in the Ordered Probit regressions are associated with a higher estimated probability that the individual reports job satisfaction of greater than level i (see Greene 1993). The Ordered Probit procedure chooses estimates b to maximize EIn(p,), where p, is the estimated probability of the observed response and the summation is over all of the observations in the data set. The probability of observing level i is p; = Pr(j1.. < x'b+u S ,), where u is assumed to be normally distributed. The b coefficients are estimated by the procedure, as are the thresholds, 4, }», ..., Hy... Where 7 is the number of categories of the ordered dependent variable. jj) is taken to be -« and pit; the probabilities thus sum to 1, Say that an individual j has characteristics x, such that x;'b = -0.7: then the estimated probability that j reports overall job satisfaction of 7, using the estimates of the y’s from Appendix 2, is Pr(-0.22 < -0.7+u < + «) = Pr(0.48 < u) = 1-F(0.48) = 31.6% (where F is the cumula- tive normal distribution). Similarly, the estimated probability that j responds “6 is Pr(-0.98 < -0.7+u < 0.22) = 1-F(-0.28) - (1-F(0.48)) = F(0.48)-F (-0.28) = 29.4%. The estimated probability that j responds ‘1’ is F(-2.03) = 21%. 9. Analogous US results are found in Blanchflower et al. (1993). 10. For example, 34% of female employees say that the most important aspect of working is the actual work itself, as opposed to 24% of male employees; for pay these figures are 13% and 19% respectively. 11. The results of separate job satisfaction regressions by gender are presented in Clark (1995b). 12. The psychological theory of cognitive dissonance implies that workers who have made a career choice will ignore information which implies that they have chosen badly (see Akerlof and Dickens 1982). Higher satisfaction could also result from the lowering of the desired or expected level of job rewards. The rising part of the age-job satisfaction profile could thus be explained either by the progressive discarding of dissatisfying comparisons with age, as the choice of career becomes increasingly fixed, or by the progressive diminution of desires with age (resignation to the job that one has). 13. This proposition rules out the scenario of perfect information and no financial constraints in which schooling is chosen optimally, which implies no correlation between education and job satisfaction. = ge © Blackwell Publishers Ltd/London School of Economics 1996. te 18. ve 20. 21. 22. 23. 24. Job Satisfaction in Britain 213 }. Blanchflower et al. (1993) and Freeman (1978) find that blacks have lower levels of job satisfaction than whites, although Bartel (1981) finds the opposite correlation. . Marriage is, however, endogenous. Veenhoven (1989) emphasizes that ‘hap- pier’ individuals have a better chance in the marriage market. . Since the log transformation forces concavity, it might be an inappropriate functional form in light of the cross-tabulation results. Experiments with the level of monthly gross pay and its square, as well as with the log of pay and the square of the log, proved unsuccessful. However, these additional es- timates confirmed the strong positive relationship between monthly pay and satisfaction with pay, and the weaker relationship with overall job satisfac- tion. None of the results were changed by the use of net (as opposed to gross) monthly pay. Mortimer et al. (1988) show that job satisfaction rises with the autonomy of the job, and Miller (1980) shows that job satisfaction is negatively correlated with a number of job pressures (such as time pressure or dirtiness). These types of variables, unmeasured in the data set, could be compensated by higher pay. If reference groups are occupation-based, then separate regressions by occupa- tion will hold y* constant and allow less-biased estimates of y. The results yielded two occupation groups, Craft and Related and sales, for which the estimated income coefficient, despite the much smaller sample size, was better-determined than that from the equation in Appendix B. A similar exercise by industry revealed strong positive income coefficients in Other manufacturing (SIC4) and Construction (SICS). There is thus some evidence consistent with the existence of occupation- or industry-based reference groups, but only for a minority of workers in the sample. This relationship is consistent with the positive correlation between firm size and sickness absenteeism reported by the CBI (1994): the lowest rate of absenteeism (2.3%) was found in firms with fewer than 100 employees, while the highest rate (4.3%) was in firms with more than 5000 employees. It might be countered that part of the attraction of smaller establishments is their likely proximity to the worker’s home. However, a variable measuring travel-to-work time was insignificant in all of these regressions and did not affect the conclusions. Rebitzer and Tsuru (1993) note that 22% of maie union members would not tell anyone about their dissatisfaction with wages, as opposed to 33% of non-union workers. Gender interactions showed that temporary or contract work has a greater negative effect on women’s job satisfaction than on men’s; however, none of the estimates were statistically significant. The true measurement of shadow wages requires information about changes in the worker’s welfare level. However, an estimated job satisfaction relationship is a reduced form which reflects both workers’ and firms’ behaviour. The Presence of a number of firm controls in these equations is designed to keep firm characteristics as constant as possible, and hence to enable us to identify the worker’s evaluation of income and hours changes. The estimated coefficient on log hours for this sub-sample is -0.8767, and that on log income is 0.5257. A rise of one hour of work per week from the mean of 29.3 thus has a negative effect on the (latent) dependent variable of ~0,8767(In(30,3)-In(29.3)) = -0.0294. A rise in monthly income of £37.29 © Blackwell Publishers Ltd/London School of Economics 1996. 214 British Journal of Industrial Relations from the mean of £647.62 produces an exactly compensating effect: 0.5257(In(684.91)-In(647.62)) = 0.0294, 25. Analogously, the estimated coefficients in Appendix B imply that the average worker would need to be paid an extra £230 per month to compensate for the lack of promotion opportunities (i.e. in order to be just as satisfied with his/her pay). 26. The set of explanatory variables used comprises those already discussed, minus Work values and income, plus Job tenure and its square and dummy variables for Organization type (e.g. public or private sector), the Sex mix of the workplace, Trade union recognition at the workplace, Pension membership, Time of day worked, and a variable for whether the individual had had an accident within the past year. A large number of regressors were used to avoid the possibility of omitted variable bias. 27. Gibbons and Katz (1992) examine such data for exogenously displaced workers and find that workers who change industries also change wages in a way that mirrors the inter-industry wage profile; this finding casts doubt on worker-based explanations of the wage structure. References Adams, J.S. (1963). ‘Toward an understanding of inequity’. Journal of Social Psychology, 67: 422-36. Akerlof, G. A. and Dickens, W. T. (1982). ‘The economic consequences of cogni- tive dissonance’. American Economic Review, 72: 307-19. —— Rose, A. K. and Yellen, J. L. (1988). ‘Job switching and job satisfaction in the US labour market’. Brookings Papers on Economic Activity, 2: 495-582. Argyle, M. (1989). The Social Psychology of Work, 2nd edn. Harmondsworth: Penguin, Bartel, A. P. (1981). ‘Race differences in job satisfaction: a reappraisal’. Journal of Human Resources, 16: 295-303. Blanchflower, D. G. and Oswald, A. J. (1992). ‘Entrepreneurship and supernormal returns: evidence from Britain and the US’. Mimeo, Dartmouth College. ——- —— (1994), “Estimating a Wage Curve for Britain’. Economic Journal, 104: 1025-43. ——~——and Warr, P. B. (1993). ‘Well-being over time in Britain and the USA’. Mimeo, Dartmouth College. Borjas, G. (1979). ‘Job satisfaction, wages and unions’. Journal of Human Resources, 14:21-40. Brown, C. and Medoff, J. (1989). ‘The employer size-wage effect’. Journal of Political Economy, 97: 1027-59. Brown Johnson, N., Bobko, P. and Hartenian, L. $. (1992). ‘Union influence on local union leaders’ perceptions of job insecurity: an empirical test’. British Journal of Industrial Relations, 30: 45-60. Clark, A. E. (1995a). ‘Is utility relative? Evidence from household data’. Mimeo, DELTA. —— (1995b). ‘Job satisfaction and gender: why are women so happy at work?” Discussion Paper no. 95-10, DELTA. —— (1995c). ‘Wages, raises and well-being: evidence for a relative utility function’. Mimeo, DELTA. ‘© Blackwell Publishers Ltd/London School of Economics 1996. Job Satisfaction in Britain 215 ——and Oswald, A.J. (1994). ‘Unhappiness and unemployment’. Economic Journal, 104: 648-59. ———— (1996). ‘Satisfaction and comparison income’. Journal of Public Economics 69: 57-81 ——— and Warr, P. B. (1996). ‘Is job satisfaction U-shaped in age?’ Journal of Occupational and Organizational Psychology, 69: 57-81 Clegg, C. W. (1983). ‘Psychology of employee lateness, absence and turnover: a methodological critique and an empirical study’. Journal of Applied Psychology, 68: 88-101. Confederation of British Industry (CBI) (1994). Managing Absence — In Sickness and in Health, London: CBI. Duesenberry, J. $. (1949). Income, Saving and the Theory of Consumer Behaviour. Cambridge, Mass.: Harvard University Press. Equal Opportunities Commission (1992). Pay and Gender in Britain: 2. London:IRS. Freeman, R. (1978). ‘Job satisfaction as an economic variable’. American Economic Review, 68: 135~41 ——— (1980). ‘The exit-voice tradeoff in the labour market: unionism, job tenure, quits and separations’. Quarterly Journal of Economics. 94: 643-74 General Household Survey (GHS) (1973). GHS Introductory Report. London: OPCs. Gibbons, R. and Katz, L. (1992). ‘Does unmeasured ability explain inter-industry wage differentials?” Review of Economic Suidies, 59: 515-35. Greene, W. (1993). Econometric Analysis, 2nd edn, London: Macmillan. Hagenaars, A. J. (1986). The Perception of Poverty. Amsterdam: North-Holland. Hakim, C. (1991). ‘Grateful slaves and self-made women; fact and fantasy in women’s work orientations’. European Sociological Review, 7: 101-21. Hamermesh, D. S. (1977). ‘Economic aspects of job satisfaction’. In O. C. Ashen- felter and W. E. Oates (eds.), Essays in Labor Market Analysis. New York: John Wiley. Harsanyi, J. C. (1986). ‘Interpersonal utility comparisons’, in J. Eatwell, M. Milgate and P. Newman (eds.), The New Palgrave Dictionary of Economics. London: Macmillan. Hodson, R. (1991). ‘Workplace behaviours: good soldiers, smooth operators and saboteurs’. Work and Occupations, 18: 271-90. Homans, G.C. (1961). Social Behaviour: lis Elementary Forms. New York: Harcourt Brace. L., McGregor, A., MacInnes, J. and Sproull, A. (1993). ‘The “flexible trategy and segmentation’. British Journal of Industrial Relations, 31: 383- Hwang, H., Reed, W.R. and Hubbard, C. (1992). ‘Compensating wage differ- entials and unobserved productivity’, Journal of Political Economy, 100: 835-58. Idson, T. L. (1990). ‘Establishment size. job satisfaction and the structure of work’. Applied Economics, 22: 1007-18. Kosicki, G. (1987). ‘A test of the relative income hypothesis’. Southern Economic Journal, 54: 422-34. Lane, R. (1991). The Labour Market Experience. Cambridge: Cambridge University Press. Lévy-Garboua, L. and Montmarquette, C. (1994). ‘On reported job satisfaction: a test of subjective well-being models and a new interpretation’. Mimeo, Université Paris I. © Blackwell Publishers Ltd/London School of Economics 1996. 216 British Journal of Industrial Relations Locke, E. A. (1976). ‘The nature and causes of job satisfaction’. In M. D. Dunnette (ed.), Handbook of Industrial and Organizational Psychology. Chicago: Rand- McNally. Mangione, T.W. and Quinn, R.P. (1975). ‘Job satisfaction, counterproductive behavior, and drug use at work’. Journal of Applied Psychology, 60: 114-16. Marsden, D. and Richardson, R. (1994). ‘Performing for pay? The effects of “merit pay” on motivation in a public service’, British Journal of Industrial Relations, 32: 243-61. McEvoy, G.M. and Cascio, W.F. (1985). ‘Strategies for reducing employee turnover: a meta-analysis’. Journal of Applied Psychology, 70: 342-53. Melenberg, B. (1992). ‘Micro-econometric models of consumer behaviour and welfare’. Mimeo, Tilburg University. Meng, R. (1990). ‘The relationship between unions and job satisfaction’. Applied Economics, 22: 1635-48. Miller, J. (1980). ‘Individual and occupational determinants of job satisfaction’. Sociology of Work and Occupations, 7: 337-66. Miller, P. (1990). ‘Trade unions and job satisfaction’. Australian Economic Papers, 29: 226-48. Mortimer, J., Finch,M. and Maruyama, G. (1988). ‘Work experience and job satisfaction: variation by age and gender’, in J. Mortimer and K, Borman (eds.), Work Experience and Psychological Development through the Life Span (AAAS Selected Symposium 107). Boulder, Colo.: Westview Press. Rebitzer, J. and Tsuru, T. (1993). ‘The limits of enterprise unionism: an empirical explanation of the causes of union decline in Japan’. Mimeo, MIT. Rose, D. and 14 others (1991). ‘Micro-social change in Britain: an outline of the role and objectives of the British Household Panel Study’. Working Papers of the ESRC Research Centre on Micro-social Change, Paper 1. Colchester: University of Essex. Runciman, W.G. (1966). Relative Deprivation and Social Justice. Henley: Routledge & Kegan Paul. Sloane, P.J. and Williams, H. (1994). ‘Job satisfaction, comparison income, and gender differences in earnings’. Mimeo, University of Aberdeen. Tausky, C. (1992). ‘Work is desirable/loathsome: Marx versus Freud’. Work and Occupations, 19: 3-17. Tinbergen, J. (1991). ‘On the measurement of welfare’. Journal of Econometrics, 50:7-15. Van de Stadt, H., Kapteyn, A. and Van de Geer, S. (1985). ‘The relativity of utility: evidence from panel data’. Review of Economics and Statistics, 67: 179-87. Van Praag, B.M. (1991). ‘Ordinal and cardinal utility’. Journal of Econometrics, 50: 69-89. Veblen, T. (1899). The Theory of the Leisure Class. London: George Allen & Unwin. Veenhoven, R. (1989). ‘Does happiness bind? Marriage chances of the unhappy’. In R. Veenhoven (ed.), How Harmful is Happiness? Rotterdam: Universitaire Pers Rotterdam. —— (1991). ‘Questions on happiness: classical topics, modern answers, blind spots’. In F. Strach, M. Argyle and N. Schwarz (eds.), Subjective Well-Being. Oxford: Pergamon Press. Warr, P.B. (1992). ‘Age and occupational well-being’. Psychology and Aging, 7: 37-45 ‘© Blackwell Publishers Ltd/London School of Economies 1996. Job Satisfaction in Britain 217 Wright, R.E. and Ermisch, J. F. (1991). ‘Gender discrimination in the British labour market: a reassessment’. Economic Journal, 101: 508-22. Zavoina,R. and McKelvey, W. (1975). ‘A statistical model for the analysis of ordinal level dependent variables’. Journal of Mathematical Sociology, Summer: 103-20. ‘© Blackwell Publishers Lid/London School of Economics 1996.

Das könnte Ihnen auch gefallen