Yes, Western country
Yes, non-Western country
Max. 10 years
15 or more years
School or study
Table 3.2 presents the basic findings for mental illness and mental health across the four points in time. We carried out the MANOVA with repeated measures to analyze this pattern of findings and to find out whether the background variables had any effect on it. When looking at the multivariate tests, we found that the factor scales was significant (F 3,1330 = 47.60; p < .001). That is to say, the mean scores on the different scales differ from each other, independent of the time of measurement. The Helmert contrasts were all significant (p < .001). The absence of psychopathology differs significantly from the presence of mental health, but this is a result of the different answering scales used (a four-point format for psychopathology and a six-point format for mental health). It can be seen in Table 3.2 that the sample scored rather low on psychopathology: approximately 0.35 on a scale from 0 to 3. More interesting is the finding that the means of emotional, psychological, and social well-being differed from each other. On the mental health measure, participants scored highest on emotional well-being, followed by psychological well-being, and lowest on social well-being (Table 3.2). The factor time is not significant (F 3,1330 = 0.78; p = .506), nor is the interaction between scales and time (F 9,1324 = 1.59; p = .113). As can be seen in Table 3.2, the means are remarkably stable across time and so are the differences between the scales.
Frequency distribution of measures of mental illness and health across time (N = 1,340)
How is this pattern qualified by the background variables? First, all background variables (age, gender, migratory status, educational attainment, being married, having paid work, and being in good physical health; all p < .05) showed a significant interaction with the factor scales. Hence, the predictive value of each of these variables was different for the four scales used. We will analyze these differences in more detail below. Furthermore, none of the interactions between the factor time and the background variables were significant (all p > .05). The stability in time observed in Table 3.2 is therefore similar for all characteristics of the participants considered here. However, age showed a significant three-way interaction with scales and time (F 9,1324 = 2.54; p = .007). A similar finding pertains to education (F 9,1324 = 2.53; p = .007). These findings indicate that participants with a different age and those with a different educational background show a somewhat different pattern across time on some scales than other scales. We will also analyze these differences in more detail below.
To further analyze the relations of the different background variables with mental illness and mental health, we carried out a regression analysis for each of the four dimensions of mental illness and mental health. We used the mean score of each dimension across the four time points as the dependent variable and age, gender, migration, education, marital status, employment status, and physical health as the independent variables. The results are presented in Table 3.3.
Ordinary least-squares regression of mental illness and mental health on background variables (N = 1,340)
Adjusted R 2
* p<.05; *** p<.001
It can be seen that age, migration, marital status, employment status, and physical health were related to mental illness: Being young, being a migrant, not being married or employed, and having poor physical health were all related to more mental illness complaints. Together, these variables account for 19% of the explained variance in mental illness. Emotional well-being was related to age, gender, marital status, and physical health: Being older, being female, being married, and having a better physical health were related to better emotional well-being. These variables account for 8% of the variance in emotional well-being. Six percent of the variance of psychological well-being can be explained by age, gender, education, and physical health: A younger age, being female, higher educational achievement, and good physical health were related to better psychological well-being. Finally, the variables explain 3% of the variance in social well-being: Higher educational achievement and good physical health were related to higher social well-being. To conclude, the mean score across time for each dimension of mental illness and mental health was differently related to different background variables.
The within-subjects contrasts in the MANOVA with repeated measures give some more information with regard to the different relations of the background variables with mental illness and mental health. It can be seen in Table 3.3 that an older age was related to less mental illness problems and higher emotional and lower psychological well-being, but it was unrelated to social well-being. All three Helmert contrasts were significant for age. That is to say that the relation of age to mental illness differed from that of mental health, the relation of age to emotional well-being differed from the relation to eudaimonic well-being, and the relation of age to psychological well-being differed from that of social well-being. Being female was related to more emotional and psychological well-being, but not to social well-being and mental illness. The Helmert contrasts were only significant for the comparison between psychological and social well-being. For the other variables, only the Helmert contrast between emotional and eudaimonic well-being was significant. It can be seen that being a migrant and being employed were negatively related to emotional well-being, but positively to psychological and social well-being. Furthermore, higher education was related to more psychological and social well-being, but not to emotional well-being. Being married was related to more emotional well-being, but not to both measures of eudaimonic well-being. Lastly, good physical health was more strongly related to emotional than to psychological and social well-being.
Two three-way interactions were significant: age by scales and time as well as education by scales and time. We carried out post hoc t-tests in order to compare the changes between any two adjacent points in time per scale and age/education group. Figure 3.1 presents the findings for three age groups: a younger group between 18 and 39 years, a middle-aged group between 40 and 59 years, and an older group aged 60 years and over. The scores for mental health range from 1 to 6, those for mental illness from 0 to 3. Straight lines refer to scores on mental illness, dotted lines to mental health. The thicker lines refer to two adjacent means, which were significantly different in the post-hoc paired t-test. It can be seen that the two younger groups (18–39 and 40–50 years) had somewhat less mental illness complaints at the second than the first time of measurement (T0 and T1). Between T1 and T2, we found a significant drop in mental illness complaints and an increase in emotional well-being. Furthermore, the youngest group decreased in emotional but increased in social well-being, whereas the middle age group showed a decrease in psychological well-being. Finally, the oldest age group showed an increase in social well-being and the two youngest groups in emotional well-being. It can be concluded that mental illness and mental health have a somewhat different longitudinal pattern for different age groups. Most important, however, is the finding that a change in one dimension does not always go together with changes in other dimensions.
Mean of mental illness and mental health across time for three different age groups. EWB emotional well-being, PWB psychological well-being, SWB social well-being, MI mental illness, MHC-SF Mental Health Continuum – Short Form, BSI Brief Symptom Inventory
Figure 3.2 presents the findings for three educational groups: lower educated (10 years or less), middle (11–14 years), and higher (15 years and more). There was a decrease in mental illness between the first and second measurement for the middle- and higher educated groups. The lowest educated group showed a significant decrease in mental illness problems 3 months later. Furthermore, all groups declined in psychological well-being, but only the two highest educated groups also declined in emotional well-being. Between T2 and T3, there was a significant increase in well-being: The lowest educated group rose in social and psychological well-being, and the middle group in social and emotional well-being. It can be concluded that mental illness and mental health have a somewhat different longitudinal pattern for different educational groups. Most important, again, is the finding that a change in one dimension does not always go together with changes in other dimensions. These findings therefore provide further evidence for the distinction between the different measures of mental illness and mental health.