Respondents
An invitation to participate in an online survey was emailed to all 2,995 employees at a Dutch university. They all had the Dutch nationality and had been employed for at least 1 year. Each respondent was given a personal number which enabled them to fill in the questionnaire online. The 142 employees who did not have a personal e-mail address received a paper version at their home address, but it was also made possible for them to respond online. One reminder was sent (by e-mail or in writing) after 10 days.
A total of 1,297 respondents returned the questionnaire (43%). Age had been filled in by 1,112 respondents, which resulted in 37% usable questionnaires. Comparison with the total population showed that the sample gave a fair reflection with respect to age, unit and ‘job classification’ (faculty versus staff). Differences were present especially among faculty. Slightly more women (37% compared to 33%) and older respondents (≥ 55 years) (23% compared to 18%) returned the questionnaire. Thus, (older) lectures were overrepresented (33% compared to 26%), while (younger) PhD students (20% compared to 25%) and faculty with temporary contracts of employment (34% compared to 43%) were underrepresented.
Questionnaire
The questionnaire was based on preliminary research in which interviews were held with stakeholders and two rounds of focus groups to enquire about barriers and factors to support work ability in (older) workers. Subsequently, relevant scales were selected from the questionnaire that is used extensively by “IVA Policy research and advice” in their employee studies (Thunissen and Van der Hoek 2001). Confirmatory factor analyses showed an almost similar classification as can be expected on theoretical grounds (data available on request), with satisfactory reliability which will be presented in the next paragraph. The questionnaire contained scales and items measuring work characteristics (i.e. job demands and job resources) and other relevant scales and items, which we will call ‘other (work) characteristics’.
The outcome measure job satisfaction was assessed using a 7-item scale (α = 0.87) with questions such as “I am satisfied with my job at the moment”, “I enjoy my work” and “I would choose exactly the same job again”.
Workload was obtained by measuring the extent to which the respondents agreed with “all in all, I have problems with workload”.
Conflicts at work was assessed with four items (α = 0.79); e.g. “conflicts are solved easily” (reverse scoring) and “I have conflicts with my colleagues”. Work-home facilitation was assessed with one single item “I can adjust my working hours well in my private life”. “Able to relax sufficiently at home from job demands” was measured with one single item.
Skill discretion was analysed with 5 items (α = 0.85), e.g. “I have enough opportunities within my current job to take on challenging new tasks” and “I can fully use my knowledge and skills during work”. Autonomy was measured with four items (α = 0.81), e.g. “I can determine how to organize my work” and “I can determine my own work pace”. Relation with colleagues was assessed with two items (α = 0.63): “the contact with my colleagues is good” and “I feel respected by my colleagues”).
The support from supervisor scale contained 16 items (α = 0.96), e.g. “my supervisor inspires and motivates me” and “my supervisor regularly discusses opportunities for my personal development”. Opportunities for further education were assessed with three items (α = 0.63): “I receive sufficient opportunities for retraining”, “it is my own responsibility to update the knowledge and skills necessary for my further development” and “the university attaches importance to retraining employees”.
In addition to the aspects from the JD-R model, several other (work) characteristics were assessed. For further exploring differences and similarities concerning workload, two items were analysed: “it is aggravating to have to work longer hours than intended” and “expecting positive results from decreasing workload”. For further exploring social support, “if there is a problem, I can ask someone for help” was included. Appreciation of older workers by the employer was assessed with three items (α = 0.68): “older workers are valued because they are highly experienced”, “older workers are expected to be less productive than younger employees” (reverse scoring) and “knowledge and experience are being passed on from older to younger employees”. Appreciation and feedback were separately assessed with “I receive enough appreciation for my efforts” and “I receive enough feedback”. Contentment with remuneration was assessed with three items. (α = 0.89): “my salary is suitable for the job”, “my salary is in accordance with my knowledge and skills” and “my salary prospects are good”). “Readiness to join in further education” included two items (α = 0.81): “I am prepared to retrain” and “I am prepared to invest time in retraining”. Furthermore, following items were included: “I am ready to take on new tasks all the time”, “I expect positive results from regular attention to career and development opportunities”, “I expect positive results from clarifying the work objectives”. The ‘positive results’ intended by these questions were job satisfaction, employability and optimal performance.
These ‘other (work) characteristics’ were not included in the multivariate analyses as they were not included in the research model by Van Ruysseveldt (2006). However, literature shows that they are associated with job satisfaction (Chen et al. 2006; Bilimoria et al. 2006; Winefield et al. 2008) and therefore of interest to get better insight into differences and similarities between the age groups.
Control variables included into the multivariate models are “presence of chronic disease”, “normal job performance is impeded by poor health”, sex and job classification [“faculty” (professors, lectures and researchers) versus “staff” (all other employees)]. The first two variables are included since the prevalence of chronic disease and poor health increases with age. Personal characteristics included age, the number of working hours per week, contract of employment (temporary or permanent), term of appointment, number of years in the same position and having children at home. They were all assessed with one single item.
Most items were scored on a 5-point scale either to indicate the level of agreement with a statement (1 completely disagree, 5 completely agree) or to measure the extent to which a statement applied to the respondent (1 not at all, 5 to a large extent). An exception was “normal job performance is impeded by poor health”, which was assessed with a 4-point answering scale (1 not/hardly, 4 greatly). Furthermore, a few items simply required a yes or no.
For all scales, a scale score was calculated by averaging the item scores. In all scales and items, higher scores mean more agreement with the proposition. Thus, higher scores for skill discretion means that the respondents experience more skill discretion (desirable), whereas higher scores for conflicts at work means that the respondents are confronted with more conflicts at work (which is undesirable).
In the statements with a positive formulation, mean scores higher than 3.5 were considered to be satisfactory, because this level suggested that the relevant work conditions were solid. In the negative formulations, this applied to mean scores of 2.5 and lower. In addition, the portion of the respondents with satisfactory scores was calculated. This means the percentages of workers with satisfactory mean scale scores (i.e. >3.5 or ≤2.5) or the percentages workers with satisfactory answers for items [i.e. either agree to moderate or to large extent or (completely) agree].
Analyses
Analyses were conducted on four age groups: younger than 35, 35–44, 45–54 and 55 years and older. This choice of classification was based on the probable major differences in home situation (e.g. younger versus older children at home) and work experience (e.g. duration of professional tenure) between the age groups that were likely to interfere with work characteristics and job satisfaction (Lynn et al. 1996). Data were analysed using SPSS version 17.0 (SPSS Inc., Chicago, IL, USA).
Differences in personal characteristics were analysed with χ2-tests (Table 1). “Normal job performance is impeded by poor health” was dichotomized. Impediment was assumed when the respondents indicated to agree ‘slightly’, ‘moderately’ or ‘greatly’ with the proposition.
Table 1 Personal characteristics per age group
In order to answer the first research question, factorial ANOVA was used to test the correlation between age and several work characteristics while adjusting for sex and job classification (Table 2).
Table 2 Differences and similarities in work characteristics between the four different age groups (sex and job classification adjusted mean [SE] and percentage respondents with satisfactory mean scores)
In order to answer the second research question, blockwise linear regression analyses were used in each age group separately to investigate variables associated with job satisfaction (Table 3). First, before including into the regression analyses, the answers of four items were dichotomized; normal job performance is impeded by poor health, problems with workload, work-home facilitation, “able to relax sufficiently at home from job demands”. Agreement with the statement (completely agree, agree and neither agree nor disagree) was appointed a one, while disagreement (disagree and completely disagree) was appointed a zero. In normal job performance is impeded due to poor health, an one was assigned to agreement (slightly, moderately and greatly) and a zero to disagreement (not/hardly). Secondly, we checked multicollinearity by computing tolerances and variance inflation factors (VIFs). Following the guidelines (Bowerman and O’Connell 1990; Menard 1995), we concluded that there was no reason for concern (adapted from Field 2002) (but available on request). The regression model with the independent variable ‘job satisfaction’ comprised three blocks. First, the control variables (presence of chronic disease, normal job performance is impeded by poor health, sex and job classification) were entered. Next, into the second block, job demands (problems with workload, conflicts at work, work-home facilitation and “able to relax sufficiently at home from job demands”) were entered. Finally, into the third block, job resources (skill discretion, autonomy, support from supervisor, relation with colleagues and opportunities for further education) were entered. Statistical significance was set at α ≤ 0.05.
Table 3 Summary of linear regression analyses on variables to explain variance in job satisfaction in the four different age groups