Testing the Relationship Between Parents’ and Their ... · evidenced a higher level of subjective...
Transcript of Testing the Relationship Between Parents’ and Their ... · evidenced a higher level of subjective...
RESEARCH PAPER
Testing the Relationship Between Parents’and Their Children’s Subjective Well-Being
Ferran Casas • Germa Coenders • Monica Gonzalez • Sara Malo •
Irma Bertran • Cristina Figuer
� Springer Science+Business Media B.V. 2011
Abstract Casas et al. (J Happiness Stud 9(2):197–205, 2008) found no significant rela-
tionship between paired answers given by parents and their 12–16-year-old children
(N = 266) for a single-item scale on overall life satisfaction (OLS). However, a signifi-
cant, but low (.19) parent–child relationship did appear for the PWI multi-item scale.
Overall, children reported higher subjective well being than parents. In this article, we
present the results obtained from confirmatory factor analysis (CFA), using more scales
and a bigger sample (N = 1,250) of paired parents and children. The study uses three
multiple-item scales: the PWI, the SWLS and the BMSLSS, and six single-item scales: the
OLS, two items from Russell’s scale on core affects, one on overall happiness, Fordyce’s
happiness item and the optional item of the BMSLSS on overall life satisfaction. Separate
CFA for each of the 3 multi-item scales showed good fit statistics. In order to check
comparability between parents and children, we tested equal loading and intercept con-straints. The models with restricted loadings fit only for the PWI and BMSLSS, but none
of the models with restricted intercepts fit. Therefore, it was only possible to estimate two
factor correlations for parents and their children, both very low (.16 for the PWI, .18 for
BMSLSS), and it was not possible to compare factor means. When correlating scores from
the 6 single-item scales for parents and children, they were all found to be significant but
very low. As regards items from the multiple-item scales for parents and children many
correlations are positive and significant, although very low, but others are non significant.
The means of some items were substantially higher for children than for parents. For some
items, differences were minor, non-significant or even reversed. All of the results suggest
that parents’ well-being is very weakly related to their own children’s well-being, in spite
of socialization, common material welfare and genetic influences. However, one out-
standing result is that in our Catalan sample, parents’ well-being seems to have a greater
influence on their female child’s well-being than on their male child’s.
F. Casas (&) � G. Coenders � M. Gonzalez � S. Malo � I. Bertran � C. FiguerInstitut de Recerca sobre Qualitat de Vida, Universitat de Girona (UdG), Girona, Spaine-mail: [email protected]
123
J Happiness StudDOI 10.1007/s10902-011-9305-3
Keywords Subjective well-being � Parents � Parent–child � Adolescents � PWI �SWLS � BMSLSS � Happiness � Life satisfaction � Positive psychology �Psychological assessment � Subjective indicators � CFA � Gender
1 Introduction
Comparing parents’ and children’s subjective well-being (SWB), including happiness, life
satisfaction, and satisfaction with life domains, is a topic of great interest for researchers.
Until now, although adults generally believe they have an important influence on their own
children, and although it might be expected that the combination of socialization and
common material welfare together with shared genetic influences would cause children to
resemble their parents in terms of attitudes, beliefs, routines and values, research results
show a much lower than expected and unclear relationship between a parent’s and his or
her child’s subjective well-being.
The theory of Subjective Well-being Homeostasis (Cummins 2003) proposes that
subjective well-being (SWB) is actively maintained around a set-point for each person, this
set-point being determined by personality. Perhaps acting as a balance between extra-
version and neuroticism, this system strives to maintain a constant, positive level of well-
being that is both highly personal and abstract (Cummins et al. 2003a). Thus, there are two
issues of relevance (Casas et al. 2008):
(a) It is assumed that there is a high level of genetic determination in each person’s
SWB set-point. While the degree of this determination is uncertain, longitudinal studies on
twins have led to estimates that the stable component of SWB has a heritability of some
40% (Lykken and Tellegen 1996; Roysamb et al. 2003). If we accept this value, then a
genetic model with a heritability of 40% predicts that family members who share 50% of
their genes should have a SWB similarity of .20. Moreover, the combination of such levels
of heritability coupled with a shared living environment should generate even higher levels
of shared variance in SWB between parents and their children. However, whether such
additive effects would, in fact, occur, is moot. While Lykken and Tellegen (1996) suggest
nonadditive effects on well-being, the data from Roysamb et al. (2003) suggest otherwise.
In either case the end result of the correlation between the SWB of parents and their
children should not be less than .20.
(b) It is proposed that the aspect of SWB being protected by homeostasis is a highly
abstract sense of self (Cummins et al. 2003a). Homeostasis is concerned with maintaining
the positive non-specific sense of self-satisfaction. As such, the genetically-determined set-
point around which it operates can be reasonably measured by data deriving from the
question ‘‘How satisfied are you with your life as a whole?’’ The data from domain-level
scales, on the other hand, are more specific in their content, in that they ask about aspects
of life (e.g. ‘‘How satisfied are you with your health?’’). Hence, since more variance is
contributed by cognitive processing, the influence of the genetic set-point is predicted to be
less than it is for ‘life as a whole’.
These considerations gave rise to two hypotheses tested by Casas et al. 2008:
Hypothesis (a): Satisfaction with life as a whole will show shared variance between
parents and children, reflecting both genetic and environmental influences.
Hypothesis (b): The shared variance for specific life domains will be less predictable
than it is for life as a whole, since the domains exhibit less genetic influence from the
SWB set-point and more variable influence from the environment.
F. Casas et al.
123
Based on a Spanish sample of N = 266 paired parents and their own children, Casas
et al. (2008) did not find any clear evidence supporting the hypothesis of a relationship
between parents’ and their 12–16-year-old children’s well-being (.083, with a lower
confidence limit of -.052 and an upper of ?.218), when measured using a single item on
overall life satisfaction (OLS). In terms of life domains, no relationship appeared in any of
the following: satisfaction with standard of living, life achievements, personal security,
groups of people I belong to, and relations with other people. Only satisfaction with health
and with security for future could be considered to be significantly related. Last but not
least, a low but significant relationship (.186) was found between parents’ and their
children using the Personal Wellbeing Index (PWI) (Cummins et al. 2003a). Children
evidenced a higher level of subjective well-being than their parents. According to t tests for
paired data, the children’s means were significantly higher for the PWI and the domains
standard of living, life achievements, groups I belong to and security for the future.
The .186 correlation found using the PWI seems to be consistent with the predicted
value of .20 on genetic considerations alone. On the one hand, the PWI has a statistical
advantage over the OLS because, being a summated scale, it is less subject to correlation
attenuation bias. However, its higher correlation may also be due to environmental factors.
Therefore, these two results together provide very equivocal evidence for a simple genetic
effect and the conclusion that PWI results seem to indicate a detectable influence from a
shared environment rather than a genetic influence.
Among the limitations of these results we may point out the sample size, which is rather
small, the fact that only one multi-item scale was used, and the fact that analyses were
limited to traditional zero-order Pearson correlations.
In order to expand upon the 2008 analysis by Casas et al., new data was collected in
2009 using a much bigger sample of N = 1,250 families, with paired answers from one
parent and one 12–16-year-old child. This article considers the following extensions to the
former study: Firstly, we use a wider range of subjective well-being scales and items in
order to check the sensitivity of the parent–child relationship to the choice of measurement
instrument. Secondly, the equivalence of the scales across generations is tested by means
of confirmatory factor analysis (CFA) models prior to correlating parents’ and children’s
SWB and testing mean differences in SWB between parents and children.
2 Methodology
2.1 Instruments
Three different multiple-item scales and six single-item scales were used: the former were
the PWI (Cummins et al. 2003a; International Wellbeing Group 2006), the Satisfaction
with Life Scale (SWLS) (Diener et al. 1985) and the Brief Multidimensional Students’ Life
Satisfaction Scale (BMSLSS) (Seligson et al. 2003). Of the latter, three well-known single-
item scales were used: one on overall life satisfaction (OLS), another on overall happiness
(HOL), and Fordyce’s single-item scale (Fordyce 1988). Additionally, two items from
Russell’s scale on core affects (CAS) were used, one on happiness and one on satisfaction
(Russell 2003), and the optional item 6 of the BMSLSS was also tested separately.
Scores for all items were collected using a 0–10 scale following Cummins and
Gullone’s (2000) recommendations, and scores for all multi-item scales were transformed
into a 0–100 scale to facilitate comparison.
Relationship Between Parents’ and Their Children’s
123
2.1.1 Personal Well-Being Index (PWI)
This scale was designed as part of the Australian Unity Wellbeing Index. Originally, it
included 7 items on satisfaction with different life domains, which is the version used in
this research. The scale has an end-labelled format, from completely dissatisfied (0) to
completely satisfied (10).
Although the PWI was created to be administered to adults, it has been tested on 12-year-
olds and older adolescents in some countries (Brazil, Chile, Romania, Spain; see Casas et al.
2011) and shown good psychometric properties. A school-children’s version (PWI-SC) has
been developed by Cummins and Lau (2005) and tested on Australian and Chinese popula-
tions, substituting ‘‘satisfaction’’ with life domains for ‘‘happiness’’ with each life domain.
Testing on the Catalan population has raised doubts regarding whether this is an appropriate
change for speakers of other languages, as meanings differ. We therefore decided to retain the
wording of the adult version for our adolescent samples. In fact, Tomyn and Cummins (2011)
also kept ‘‘satisfaction’’ when collecting data from a sample of Australian adolescents using
the PWI-SC. The item referring to ‘‘satisfaction with feeling part of my community’’ has been
substituted by ‘‘satisfaction with the groups of people I belong to’’, like in previous Spanish
samples, as explained in Casas et al. (2011).
The psychometric properties of the PWI have been published in several articles (see, for
example, Lau et al. 2005; International Wellbeing Group 2006; Cummins et al. 2003b).
Psychometric properties found for adolescents and details of the translations have been
published in Casas et al. (2011). In this research, exactly the same version was used as in
Casas et al. (2008).
2.1.1.1 Satisfaction with Life Scale (SWLS) This scale includes 5 items and responses
were originally coded on a scale of 1–7 according to the level of agreement. Several
Spanish adaptations exist. Due to difficulties in understanding because of its negative
wording, most Spanish adaptations have changed the original item 5 (If I could live my life
over, I would change almost nothing) in a way that scores must be reversed before adding
together the 5 items to calculate the index for the overall scale (If I was born again, I would
change quite a lot of things in my life).
The version which we use here (the latter one in the last paragraph) is the same that was
used in Casas et al. (2011), where the 1–7 scale was changed to a 0–10 scale in order to
make it more sensitive, and also for easier comparison of results with other instruments
using a 0–10 scale, with labels for all values from strongly agree to strongly disagree.
The psychometric properties of this scale have been published in various articles. See,
for example, Pavot and Diener (1993) and Diener et al. (1985). Psychometric properties
with adolescents have been published in Casas et al. (2011).
2.1.2 Brief Multidimensional Students’ Life Satisfaction Scale (BMSLSS)
This scale was developed to be used with students aged 8–18. It includes 5 items referring
to satisfaction with different life domains. Responses were originally coded on a scale of
1–7, from terrible to delighted. The psychometric properties of this scale have been
published in different articles (Seligson et al. 2003, 2005; Huebner et al. 2006).
We have changed the 1–7 scale to a 0–10 scale in order to make it more sensitive.
Labels have been given to each value, describing satisfaction with each life domain from
terrible to delighted.
F. Casas et al.
123
One of the original items, satisfaction with school experience, has been substituted in
the parents’ questionnaire by one item on satisfaction with professional experience.
2.1.3 Single Item on Overall Life Satisfaction (OLS)
The importance of including a single-item scale on overall life satisfaction when studying
subjective well-being has been highlighted by Campbell et al. (1976). In our research, we
included a question on Satisfaction with your overall life, using an end-labelled 0–10 scale,
from completely dissatisfied to completely satisfied. The translation of the item back into
English is: At present, to what extent are you satisfied with your life as a whole?
2.1.4 Items Regarding Happiness and Satisfaction on Russell’s Scale (CAS)
Russell’s scale on core affects (2003) is based on the following question: Please indicatehow each of the following items describes your feelings when you think about your life ingeneral. Then a list of affects is presented, including an item on satisfaction and one item
on happiness.
In our questionnaire we have included the original question with these two last items, to
be evaluated from 0 to 10, with the extreme values labelled from not at all to absolutely.
2.1.5 Single Item on Happiness Taking into Account Overall Life (HOL)
Campbell et al. (1976) also pointed out the importance of including a single-item scale on
happiness when studying personal well-being. In our research, we have included the fol-
lowing question: ‘‘Taking into account your overall life, would you say you are…?’’ and
then options are offered using a 0–10 scale, from extremely unhappy to extremely happy.
Only the extreme values are labelled.
2.1.5.1 Fordyce’s One-Item Scale on Happiness Fordyce’s scale is a single-item scale
asking In general, to what extent do you usually feel happy or unhappy? (Fordyce 1988).
Answers are given on a 0–10 scale from completely unhappy to completely happy. Each
value has a label.
2.1.6 Single Item on Satisfaction with My Overall Life (Optional for BMSLSS)
The authors of the BMSLSS add an item on overall life satisfaction to offer the possibility
of using a 6-item version (Seligson et al. 2003). The item is worded ‘‘I would describe mysatisfaction with my overall life as …’’. However, according to some authors (Campbell
et al. 1976; Cummins and Cahill 2000), an overall life satisfaction item should be con-
sidered at a higher level of abstraction than satisfaction with life domain items and should
not be included on the same scale. We therefore report it as a single item.
2.2 Procedure and Participants
A two-stage cluster sampling design was used to select the sample of adolescents. In the
first stage, we randomly selected a number of secondary schools in Catalonia (Spain). At
each school we proceeded in accordance with regular ethical guidelines for administering
Relationship Between Parents’ and Their Children’s
123
questionnaires to children in Catalonia. The aims of the research were reported to the
school director and to the parents’ association in order to obtain permission.
When a school agreed to participate, we randomly selected a number of classes until we
filled a quota for each of the 4 years of compulsory secondary education from each school.
We then requested the class teacher’s co-operation. Following approval, and as soon as the
ethical and formal procedures were concluded, the children were asked for their
co-operation and were informed that their data would be treated confidentially and that
they were free to refuse. The questionnaires were administered to the whole group in their
regular classroom. One of their usual teachers and one or two researchers were present
during administration, and clarified any questions that arose.
The age and gender distribution of the sample is shown in Table 1. The smaller group of
16 year-olds mostly includes adolescents who have repeated one school year. There were
slightly more girls (55.8%) than boys in the sample.
After filling out the questionnaire, all participating students received a sealed envelope
to be delivered to their parents. It included the parents’ questionnaire and a letter pre-
senting the research asking for co-operation and the return of the questionnaire back to the
school in a week’s time. The parents’ questionnaire had a code in order to pair data with
their child’s. Only biological parents’ answers were included. As usual, fathers’ answers
(28.4%) were much fewer than mothers’ (Table 1).
2.3 Statistical Analyses
With the dual aims of correlating parents’ and children’s SWB and of finding differences in
mean SWB between parents and children, the statistical analyses are done in 3 steps:
1. In order to first assess the validity of the factorial structure of multi-item scales, we
tested different confirmatory factor analysis (CFA) models separately for parents and
Table 1 Children’s gender crosstabulation by age and by genderof the parent responding thequestionnaire
Boys Girls Total
Child
12 163 170 333
13.0% 13.6% 26.6%
13 131 181 312
10.5% 14.5% 25.0%
Age
14 122 161 283
9.8% 12.9% 22.6%
15 100 129 229
8.0% 10.3% 18.3%
16 36 57 93
2.9% 4.6% 7.4%
Father responds 169 186 355
13.5% 14.9% 28.4%
Mother responds 383 512 895
30.6% 41.0% 71.6%
Total 552 698 1250
44.2% 55.8% 100.0%
F. Casas et al.
123
children for each of the 3 scales. We used the AMOS 19 software with maximum likeli-
hood estimation and the bootstrap method to compute standard errors. The fit indices
considered were TLI (Tucker & Lewis Index), CFI (Comparative Fix Index), RMSEA
(Root Mean Square Error of Approximation) and SRMR (Standardized Root Mean Square
Residual). Depending on the sources, acceptable TLI and CFI values are above .950 and
acceptable RMSEA and SRMR below .05 or .08 (Batista-Foguet and Coenders 2000;
Arbuckle 2010; Byrne 2010; Browne and Cudeck 1993).
2. In order to compare and relate CFA results across generations, in the second step we
specified models relating parents’ and children’s latent variables for each scale. Results are
comparable only if factor invariance is first found to be tenable. Factor invariance refers to the
extent to which items used in survey-type instruments mean the same to members of different
groups and is a requisite before factors can be compared in a meaningful way; otherwise,
group differences could be attributable to true differences in distributions or to a different
meaning given to variables. There are several types of factor invariance. Metric factorinvariance is a requisite for estimating and interpreting factor correlations across generations.
Strong factor invariance is required for comparing factor means, as in testing the equality of
means in SWB between parents and children, for instance (Meredith 1993). In our case, the
parent and child groups constitute a paired data set and the basic factor invariance ideas are
therefore applied in a slightly different manner than usual. Instead of having a one-factor
model in two populations, we have a two-factor model in one population. The simplest test of
factor invariance consists in a) fitting a CFA to data from both generations by allowing all
parameters to be different across generations. If this model fits the data well, the test continues
by b) constraining the unstandardized factor loadings to make them equal across generations.
Metric factor invariance is tenable if the fit of the constrained model is not considerably worse
than the fit of the unconstrained model (Brannick 1995; Kelloway 1995). If this is the case, the
test continues by c) constraining measurement intercepts (values of the item corresponding to
the zero value of the factor) across generations. Strong factor invariance is again tenable if the
fit of the constrained model is not considerably worse than the fit of the unconstrained model.
To clarify the meaning of ‘‘considerably worse’’, Cheung and Rensvold (2002) recommend
maintaining the invariance hypotheses if CFI decreases by less than .01 when introducing the
constraints.
3. If factor invariance holds, correlations of parents’ and children’s SWB and mean
differences in SWB between parents and children for multi-item scales can be directly
obtained from the estimates of the final constrained CFA model in 2). Otherwise, they are
obtained separately for each item by standard paired data procedures (Pearson correlations
and t tests and confidence intervals).
3 Results
3.1 Data Exploration
The mean responses given by parents and their children to each of the 9 scales are shown in
Table 2 together with their standard deviations, where gender differences are also pre-
sented. The first outstanding result is that children’s scores are always higher than parents’,
regardless of the instrument used. Secondly, although for both parents and children the
lowest average scores are given by the same multiple-item and single-item instruments (the
SWLS and Fordyce’s item), the highest scores clearly differ between generations,
depending on the instrument used.
Relationship Between Parents’ and Their Children’s
123
If we go into more detail and compare children’s and parents’ answers to each of the
items on the 3 multi-item scales (Table 3), the two generations only coincide in the fact
that the lowest responses are for item 5 of the SWLS, as the second lowest item also
differs.
Item 5 of the SWLS has been pointed out as problematic by some authors, even
suggesting that it might reflect traits other than well-being (Veenhoven 1994, 2009). High
scores may reflect more of a conformist personality than well-being. People evaluating new
experiences may be happy with their lives but not want to repeat the same experiences—
considering new experiences to be more desirable. This may be true for many young
people during adolescence.
3.2 Confirmatory Factor Analysis Separately for Parents and Children
The results of the initial CFA models for each of the 3 multi-item scales, relating the items
to a single latent variable for each scale with no constraints or correlated errors, are mixed.
Table 2 Mean responses bychildren and parents to 6 single-item subjective well-being scalesand three multiple-item sub-jective well-being scales (PWI,SWLS and BMSLSS)
Overall parents’ and children’smeans are bold in order tofacilitate their comparison
Boys Girls Totalchildren
Fathers Mothers Totalparents
PWI
Mean 81.47 81.47 82.12 71.52 74.45 73.62
SD 13.66 13.66 12.20 12.46 13.13 13.00
SWLS
Mean 74.40 74.40 74.44 68.53 71.26 70.49
SD 14.84 14.84 15.68 12.94 14.96 14.46
BMSLSS
Mean 80.29 80.29 79.72 77.84 79.82 79.26
SD 13.96 13.96 12.45 12.36 11.41 11.72
Item 6 BMSLSS
Mean 8.23 8.23 8.23 7.82 8.08 8.01
SD 1.67 1.67 1.60 1.37 1.41 1.41
CAS satisfied
Mean 7.98 7.98 7.86 7.45 7.67 7.61
SD 1.83 1.83 1.91 1.49 1.66 1.61
CAS happy
Mean 8.41 8.41 8.31 7.61 7.88 7.80
SD 1.74 1.74 1.87 1.53 1.63 1.60
Fordyce
Mean 7.93 7.93 7.84 7.38 7.44 7.42
SD 1.61 1.61 1.75 1.37 1.55 1.50
HOL
Mean 7.93 7.93 7.89 7.37 7.56 7.51
SD 1.66 1.66 1.70 1.23 1.52 1.45
OLS
Mean 8.41 8.41 8.26 7.53 7.86 7.77
SD 1.71 1.71 1.92 1.43 1.62 1.58
F. Casas et al.
123
Table 3 Mean responses by children and parents for each item of the three multiple-item scales
Boys Girls Total children Fathers Mothers Total parents
PWI
Health
Mean 8.42 8.57 8.51 7.25 7.38 7.34
SD 1.97 1.71 1.83 1.89 2.20 2.12
Standard living
Mean 8.50 8.62 8.57 7.09 7.31 7.25
SD 1.68 1.47 1.57 1.66 1.85 1.80
Achievements
Mean 7.95 8.07 8.02 7.28 7.63 7.53
SD 1.82 1.80 1.81 1.62 1.71 1.69
Safety
Mean 7.88 7.54 7.69 7.21 7.42 7.36
SD 1.99 2.11 2.06 1.57 1.77 1.72
Groups I belong to
Mean 8.54 8.69 8.62 7.32 7.84 7.69
SD 1.85 1.88 1.87 1.54 1.59 1.59
Future security
Mean 7.58 7.62 7.60 6.56 6.70 6.66
SD 2.20 1.94 2.06 1.86 2.04 1.99
Relations people
Mean 8.15 8.37 8.28 7.35 7.83 7.70
SD 2.08 1.77 1.91 1.46 1.52 1.52
SWLS
1. Life close to ideal
Mean 6.82 6.57 6.68 6.76 6.91 6.87
SD 2.18 2.20 2.19 1.67 1.89 1.83
2. Excellent life conditions
Mean 9.01 9.01 9.01 7.71 7.98 7.90
SD 1.35 1.42 1.39 1.33 1.59 1.52
3. Satisfied with life
Mean 8.41 8.29 8.34 7.63 7.85 7.79
SD 1.94 2.01 1.98 1.47 1.72 1.65
4. Got important things
Mean 7.74 7.92 7.84 7.69 7.94 7.87
SD 2.30 2.00 2.14 1.51 1.64 1.61
5. I’d change a lot in my life
Mean 4.79 4.54 4.65 5.48 5.05 5.17
SD 3.58 3.45 3.51 2.77 2.99 2.93
BMSLSS
Family
Mean 8.45 8.25 8.34 8.20 8.44 8.37
SD 1.81 1.78 1.80 1.60 1.50 1.53
Friends
Relationship Between Parents’ and Their Children’s
123
The models showed proper goodness of fit statistics for the SWLS with both adolescents
and parents (see Model P1 for parents and Model A1 for adolescents in Table 4).
In both the parents’ and the adolescents’ models, item 5 of the SWLS showed much
lower estimated standardized loadings than any other item. For that reason we tested the
same models without this item (Models P2 and A2 in Table 4) and both models did well
enough.
The initial BMSLSS CFA model did not fit when used with parents, and the PWI did not
fit with either parents or adolescents. By allowing some errors to covariate, we achieved
modified models with good fit statistics. For the BMSLSS, the errors we allowed to
covariate in the parents’ model (Model P3) were satisfaction with professional experience
and satisfaction with self. Allowing the equivalent errors to covariate in the children’s
model, the fit indices also clearly improved, therefore we used this modified model also for
Table 4 CFA. Fit statistics for factorial structure in different models separately for parents’ (models P1–P4) and adolescents’ well-being (models A1–A4)
Model Parents v2 df p Value TLI CFI RMSEA(confidenceinterval)
SRMR
P1 SWLS 19.931 5 .001 .989 .995 .049 (.028–.072) .0174
P2 SWLS (4 items) 5.176 2 .075 .996 .999 .036 (.000–.075) .0086
P3 Modified BMSLSS (1 pair oferrors allowed to covariate)
10.313 4 .035 .989 .996 .036 (.008–.063) .0136
P4 Modified PWI (4 pairs oferrors allowed to covariate)
12.847 10 .232 .998 .999 .015 (.000–.036) .0096
A1 SWLS 17.942 5 .003 .977 .988 .046 (.024–.069) .0211
A2 SWLS (4 items) 10.966 2 .004 .971 .990 .060 (.029–.097) .0186
A3 Modified BMSLSS (1 pair oferrors allowed to covariate)
7.252 4 .123 .993 .997 .026 (.000–.055) .0119
A4 Modified PWI (3 pairs oferrors allowed to covariate)
32.379 11 .001 .983 .991 .039 (.024–.056) .0184
P Parents, A Adolescent Children
Table 3 continued
Boys Girls Total children Fathers Mothers Total parents
Mean 8.29 8.60 8.46 7.59 7.99 7.88
SD 1.73 1.49 1.61 1.52 1.51 1.52
School experience (parents: prof. exp)
Mean 7.16 7.32 7.25 7.63 7.52 7.55
SD 2.14 1.90 2.01 1.68 1.82 1.78
Myself
Mean 7.89 7.59 7.72 7.59 7.74 7.70
SD 1.99 2.01 2.00 1.51 1.59 1.57
Place I live in
Mean 8.36 8.09 8.21 7.91 8.22 8.13
SD 1.84 2.05 1.96 1.76 1.70 1.72
Overall parents’ and children’s means are bold in order to facilitate their comparison
F. Casas et al.
123
the adolescents (Model A3). For the PWI (Model P4 for parents and Model A4 for
adolescents), the errors we allowed to covariate for adolescents were: (1) satisfaction with
interpersonal relationships and satisfaction with groups I belong to; (2) satisfaction with
standard of living and satisfaction with health; and (3) satisfaction with standard of living
and satisfaction with life achievements. For parents, we needed to allow one additional
error to covariate: (4) satisfaction with standard of living and satisfaction with future
security.
3.3 Test of Factor Invariance for Scale Comparability
Models PA2, PA3 and PA4 in Table 5 relate each of the SWB multi-item scales between
parents and children without constraints and exhibit acceptable goodness of fit measures.
When we test metric factor invariance by restricting factor loadings (Models PA2L to
PA4L in Table 5), only the BMSLSS and the PWI reduce CFI by less than .01. Correla-
tions between parents’ and children’s factors are only interpretable for these two dimen-
sions and were .181 for the BMSLSS (95% bootstrap confidence interval .091–.277) and
.156 for the PWI (95% bootstrap confidence interval .077–.237). See model PA3L in Fig. 1
and Model PA4L in Fig. 2.
When we test strong factor invariance by restricting both factor loadings and intercepts,
the fit of all models (PA2LI, PA3LI and PA4LI in Table 5) deteriorated considerably, thus
making it impossible to test the equality of means on any of the three subjective well-being
scales across generations. These results mean that items on the selected scales relate
differently to subjective well-being for parents and children.
From the above analyses it follows that we can generally estimate intergenerational
correlations and mean differences at the item level. The only two interpretable multiple-
item scale correlations have already been presented in this section.
Table 5 Fit statistics for different models relating parents’ and adolescents’ well-being
Models Parents–children v2 df p Value TLI CFI RMSEA(confidenceinterval)
SRMR
PA2 SWLS (4 items)unconstrained
43.566 19 .001 .990 .993 .032 (.020–.045) .0237
PA3 Modified BMSLSSunconstrained
97.015 32 .000 .966 .976 .040 (.031–.050) .0303
PA4 Modified PWIunconstrained
144.134 69 .000 .983 .987 .030 (.023–.036) .0266
PA2L SWLS (4 items) loadingsrestricted
112.883 22 .000 .967 .974 .058 (.047–.068) .0323
PA3L Modified BMSLSS loadingsrestricted
117.504 36 .000 .962 .969 .043 (.034–.051) .0332
PA4L Modified PWI loadingsrestricted
180.961 75 .000 .979 .982 .034 (.027–.040) .0298
PA2LI SWLS (4 items) loadingsand intercepts restricted
550.591 25 .000 .834 .852 .130 (.120–.139) .0383
PA3LI Modified BMSLSS loadingsand intercepts restricted
257.322 40 .000 .908 .918 .066 (.058–.074) .0332
PA4LI Modified PWI loadings andintercepts restricted
507.534 81 .000 .920 .929 .065 (.060–.070) .0326
Relationship Between Parents’ and Their Children’s
123
3.4 Intergenerational Correlations
When we correlate the 6 single-item scales, paired parents’ and children’s scores show
positive and significant, although very low, correlations. The widely-used OLS yields a
95% confidence interval for the correlation between parents and children from .032 to .142.
The highest upper confidence limits are close to .2 (Fordyce, CAS happy and CAS sat-
isfied), see Table 6.
Most correlations between parents’ and children’s scores on the PWI domains are
positive and significant, although very low—the exceptions being satisfaction with future
security and satisfaction with groups of people I belong to, which do not reach significance.
Correlations between parents and children for items on the SWLS are all significant, but
low. Most correlations for the BMSLSS domains between parents and children are positive
and significant, although very low—the exceptions being satisfaction with friends and
satisfaction with school experience (satisfaction with professional experience for parents),
which do not reach significance (Table 7). Although PWI and BMSLSS items correspond
to specific life domains and SWLS items are more general and context-free, when con-
sidered globally their correlations are of about the same size and similar to those for the
one-item scales in Table 6—the mentioned non-significant cases excepted. Overall, we can
say that Hypothesis (a) is confirmed, although the shared intergenerational variance arising
from genetics and environment is very low. Hypothesis (b) is disconfirmed, because the
Fig. 1 Standardized estimatesfor Model including parents’ andchildren’s BMSLSS scores withconstrained unstandardizedloadings (Model PA3L)
F. Casas et al.
123
environment-driven domain correlations are approximately as high as correlations for life
as a whole. The results thus suggest that environment has a weak effect, and genetics an
even weaker one.
Fig. 2 Standardized estimates for Model including parents’ and children’s PWI scores with constrainedunstandardized loadings (Model PA4L)
Table 6 Zero order Pearsoncorrelations between parents andchildren for the 6 single-itemscales (n = 1,250 pairs) withconfidence intervals(1 - a = 95%)
Pointestimate
Lowerconfidencelimit
Upperconfidencelimit
OLS .087 .032 .142
HOL .133 .078 .188
Fordyce .145 .090 .200
CAS happy .145 .090 .200
CAS satisfied .147 .092 .202
Item 6 BMSLSS .127 .072 .182
Relationship Between Parents’ and Their Children’s
123
3.5 Intergenerational Mean Differences
Positive mean differences in Table 8 are interpreted as a higher subjective well-being for
children than for parents. This is significantly the case for all 6 single-item scales. These
differences between parents and children are also significantly dependent on scale choice.
The largest differences (OLS, CAS happy) have confidence intervals which do not overlap
with the lowest differences (CAS satisfied and Item 6 of the BMSLSS). Cohen’s d
Table 7 Zero order Pearsoncorrelations for all items on the 3multiple-item scales (PWI,SWLS and BMSLSS) and confi-dence intervals (1 - a = 95%)
* Non significant correlation
Pointestimate
Lowerconfidencelimit
Upperconfidencelimit
PWI
Health .089 .033 .144
Standard living .183 .128 .237
Achievements .100 .045 .155
Safety .097 .042 .152
Groups I belong to .041* -.015 .096
Future security .051* -.005 .106
Relations people .063 .008 .118
SWLS
1. Life close to ideal .107 .051 .162
2. Excellent life conditions .192 .138 .246
3. Satisfied with life .152 .097 .207
4. Got important things .081 .026 .137
5. I’d change a lot in mylife
.094 .039 .149
BMSLSS
Family .091 .137 .246
Friends .029* -.026 .084
School experience(parents: sat. profess exp)
.032* -.023 .087
Myself .103 .048 .158
Place I live in .139 .084 .194
Table 8 Mean differences and confidence intervals for parent and child (1 - a = 95%), and Cohen’s d(n = 1,250 pairs) for the 6 single-item scales
Mean differencepoint estimate
SE meandifference
Mean differencelower confidencelimit
Mean differenceupper confidencelimit
Effect size formean differences(Cohen’s da)
OLS .56 .065 .44 .69 .24
HOL .40 .059 .29 .51 .19
Fordyce .46 .059 .34 .57 .22
CAS happy .55 .063 .43 .68 .25
CAS satisfied .31 .065 .18 .43 .13
Item 6 BMSLSS .22 .057 .11 .33 .11
a Absolute mean difference divided by the standard deviation of differences (Cohen 1988; Rosenthal 1991)
F. Casas et al.
123
(a measure of effect size) shows mean differences to be relatively low. As a rule of thumb,
Cohen referred to effect sizes as ‘‘small, d = .2,’’ ‘‘medium, d = .5,’’ and ‘‘large, d = .8’’.
There are very diverse differences between parents’ and children’s mean responses to
single items of multi item scales (Table 9). One group of items shows substantially higher
child SWB (in the region of one point on an 11-point scale): health, standard of living,
groups and future security (PWI) and the second item on the SWLS. These items also show
the highest Cohen’s d statistics, well above what Cohen referred to as a ‘‘small’’ rela-
tionship. Another group of items shows higher child SWB, but to a lesser degree: the
remaining PWI items, the third SWLS item and friends (BMSLSS). A third group of items
shows non-significant differences between parents and children: the fourth SWLS item,
and family, self and place (BMSLSS). A fourth group of items shows lower child SWB: the
first and fifth SWLS items, and school/profession (BMSLSS).
Table 9 Mean differences and confidence intervals (1 - a = 95%) for parent and child for all items on the3 multiple-item scales (PWI, SWLS and BMSLSS) and Cohen’s d
Meandifferencepointestimate
SE meandifference
mean differencelowerconfidence limit
mean differenceupperconfidence limit
Effect size formean differences(Cohen’s da)
PWI
Health 1.17 .076 1.02 1.31 .44
Standard living 1.32 .061 1.20 1.44 .61
Achievements .48 .066 .35 .61 .21
Safety .33 .072 .19 .47 .13
Groups I belong to .93 .068 .80 1.06 .39
Future security .94 .079 .78 1.09 .34
Relations people .58 .067 .45 .71 .24
SWLS
1. Life close toideal
-.19 .076 -.34 -.04 .07
2. Excellent lifeconditions
1.11 .052 1.00 1.21 .60
3. Satisfied withlife
.56 .067 .43 .69 .23
4. Got importantthings
-.03* .073 -.17 .12 .01*
5. I’d change a lotin my life
-.52 .123 -.76 -.28 .12
BMSLSS
Family -.03* .060 -.15 .09 .01*
Friends .58 .062 .46 .70 .27
School experience(parents: sat.profess exp)
-.30 .075 -.45 -.16 .11
Myself .03* .068 -.11 .16 .01*
Place I live in .08* .069 -.06 .21 .03*
a Absolute mean difference divided by the standard deviation of differences (Cohen 1988; Rosenthal 1991)
* Non significant
Relationship Between Parents’ and Their Children’s
123
3.6 Differences in Parent/Child Correlations According to the Gender of the Child
When we analyze the correlations for the 6 single-item scales and their confidence limits
by child gender (Table 10), we observe that they are significant for all scales between
parents and a female child, but not when the child is a boy. The OLS, Fordyce’s scale and
item 6 of the BMSLSS do not correlate significantly between parents and a male child.
Excepting for the CAS-happy, correlations for all scales are clearly lower for male children
than for female children.
When we analyze the correlations for each of the items on the 3 multiple-item scales
and their confidence limits (Table 11), we observe that they are significant for 13 out of 17
items between parents a female child, but only for 8 items between parents and a male
child. Among the significant correlations in the male child group, only satisfaction with life
achievements and item 4 on the SWLS have a slightly higher upper confidence interval for
boys than for girls—all other significant correlations being lower for boys.
We also performed similar analyses for parent gender and found no relevant differences.
We thus chose to report only the differences based on child gender. In any case, parent
gender and child gender were statistically unrelated (Pearson’s v2 = 2.38 with 1 df;p value = .122; Cramer’s V = .044), showing that mothers or parents were equally
inclined to respond regardless of the child’s gender, and that the parent gender variable
does not appear capable of affecting the results of the child gender variable.
4 Discussion
In the studied sample, the subjective well-being of 12–16-year-old adolescents appears to
be significantly higher than that of their parents, whichever scale is used. While differences
are not dramatic, this result seems to be consistent with findings showing that 12-year-olds
in different countries score significantly higher in SWB than the mean population
(Argentina, Australia, Brazil, Chile, Romania and Spain), although these scores decrease
from that age onwards. Different authors disagree on the interpretation of these findings.
Tomyn and Cummins (2011) attribute such results to increased depressive symptoms at
Table 10 Zero order Pearson Correlations between parents and children for the 6 single-item scales andtheir confidence limits, depending on child gender
Boys Girls
Pointestimate
Lowerconfidencelimit
Upperconfidencelimit
Pointestimate
Lowerconfidencelimit
Upperconfidencelimit
OLS .049* -.035 .132 .113 .039 .186
HOL .094 .011 .177 .160 .087 .233
Fordyce .073* -.010 .156 .194 .121 .267
CAS happy .150 .068 .233 .141 .068 .215
CASsatisfied
.119 .036 .201 .165 .092 .239
Item 6BMSLSS
.049* -.034 .132 .185 .112 .258
* Non significant correlation
F. Casas et al.
123
these ages of adolescence, while Casas et al. (2011) state that these results may be related
to habitual adolescent development in the context of many societies—and that such results
had not been highlighted before for the simple reason that it is only in the last two decades
that researchers have started to collect data on the subjective well-being of adolescents in
different countries using the same instruments, usually not using the more sensitive scales
(0–10).
Differentiation from parents’ criteria may be part of a natural process of building one’s
own identity. We also suspect that the older they are, the more meticulous and precise most
children and adolescents try to be when facing a scale in a questionnaire—although this is
another hypothesis to be tested in the future, probably related to cognitive skills devel-
opment. However, we should also be open to the hypothesis that the ‘‘sources’’ of sub-
jective well-being are generally different among adults than among 12–16-year-old
adolescents.
Our higher scores among adolescent children cannot be explained by a general posi-
tivity bias. Not all items show a higher satisfaction among the children. Exceptions are
Table 11 Zero order Pearson correlations between parents and children for items on the 3 multiple-itemscales and their confidence limits, depending on child gender
Boys Girls
Pointestimate
Lowerconfidencelimit
Upperconfidencelimit
Pointestimate
Lowerconfidencelimit
Upperconfidencelimit
PWI
Health .062* -.021 .145 .112 .038 .185
Standard living .117 .034 .200 .241 .169 .313
Achievements .098 .015 .181 .103 .029 .177
Safety .050* -.033 .133 .142 .068 .215
Groups I belong to .068* -.016 .151 .019* -.055 .094
Future security .019* -.064 .102 .080 .006 .154
Relations people .071* -.012 .155 .055* -.019 .129
SWLS
1. Life close to ideal .090 .007 .173 .117 .043 .191
2. Excellent lifeconditions
.110 .027 .193 .256 .185 .328
3. Satisfied with life .091 .008 .174 .200 .128 .273
4. Got important things .090 .007 .174 .073* -.001 .147
5. I’d change a lot in mylife
.020* -.064 .103 .154 .081 .227
BMSLSS
Family .169 .087 .252 .206 .134 .279
Friends -.005* -.089 .078 .062* -.012 .137
School experience(parents: sat. professexp)
-.036* -.120 .047 .091 .018 .165
Myself .071* -.012 .155 .125 .051 .198
Place I live in .124 .042 .207 .144 .071 .217
* Non significant correlation
Relationship Between Parents’ and Their Children’s
123
observed when we analyze the items on the multiple-item scales: items 1 and 5 on the
SWLS and satisfaction with school on the BMSLSS—which was compared with satis-
faction with professional experience for parents—showed significantly higher satisfaction
among parents than among their children.
At the item level, item 2 on the SWLS and satisfaction with health, with standard of
living, with groups I belong to and with security for future on the PWI show significant and
high mean differences between parents and children (around 1 point higher for children
over an 11-point scale). These three PWI items were also previously identified as showing
the highest generational difference by Casas et al. (2008)—together with satisfaction with
life achievements—not showing such a large difference in the present study -, but satis-
faction with health showing much higher mean differences than in the quoted one.
The fact that models for the PWI and BMSLSS did fit with restricted loadings suggests
that it is possible to compare correlations between parents and their children. However,
although they are significant, they are also very low.
The fact that none of the well-being-scale models fitted when restricting intercepts
suggests that the behaviour of the items differs between parents and children when related
to the respective overall score for the scale. These results lead us to the conclusion that it is
not possible to compare either factor means in CFA or means for the summated scales
between parents and their children. This at least partially results from the fact that for some
items child means are substantially higher than parent means, for some items both means
are about equal, and for yet other items parent means are higher than child means.
This study serves as a good example of how results obtained using a specific scale for
different groups or cultures—including age groups, or generations—cannot always be
compared. Such results highlight a challenge for future research: Would it be possible to
design a well-being scale which is not only valid for both adults and adolescents when
considered separately but also comparable between the two? Here we have used a scale
that was designed to be used with children and adolescents (the BMSLSS), one that was
designed for adults (the SWLS) and another which, although originally aimed at adults, is
commonly applied to both adult and adolescent populations (the PWI). At the item level,
the PWI has proved to be the scale which detects the largest average generational dif-
ferences in satisfaction.
On the other hand, we have correlated a broad range of single-item well-being scales for
parents and children and results seem to be consistent with the fact that parents’ well-being
is only very weakly related to that of their child.
When we correlate each individual item on the multiple-item well-being scales—i.e.
satisfaction with a concrete life domain—results suggest that some domains are rather
unrelated (friends, groups of people I belong to, future security, school/work), with family
and standard of living showing the highest correlations. In the case of school/work, results
suggest that work experience is not quite comparable with school experience. With future
security, doubts can be raised regarding whether it is understood in the same way by
parents and children, and with groups, it often happens that children and their parents have
unrelated core groups of friends. Family and standard of living constitute a common
intergenerational environment, however.
Evidence from different studies shows ‘‘friends’’ to be the only life domain where
satisfaction does not decrease during the 12-to-16 adolescent period in some countries—
whereas satisfaction with family does decrease (Casas et al. 2011). Friends become sup-
portive and very important in social networks at this age, while during adulthood it is
common for family to again become the most significant source of social support, par-
ticularly in families with children in a Mediterranean cultural context.
F. Casas et al.
123
One outstanding result is that in our Catalan sample parents’ well-being seems to have a
bigger influence on the female child’s well-being than on the male child’s. We have
checked whether the gender of the parent may have any influence on these results, but it
seems not to. A possible explanation is that in Catalan culture boys receive more
encouragement to act independently from the family than girls, and they are therefore less
likely to adopt their families’ lifestyle and criteria for evaluating satisfaction with life
domains and with life as a whole. Girls are thought by many parents to be more vulnerable
to certain dangers than boys, and are therefore more often ‘‘protected’’ or under parental
control. On the other hand, many Catalan girls are much more interested in knowing about
the family history than boys, which could hypothetically be related to a more empathic
approach to the family’s style of evaluating satisfaction with life experiences, and there-
fore, well-being. Finally, yet importantly, along the same lines as those suggested for
explaining intergenerational differences in SWB, we cannot for now reject the hypothesis
that the ‘‘sources’’ of SWB may be somehow different among adolescent boys and girls in
the context of each specific culture.
Satisfaction with friends, relations with other people and groups I belong to show no
significant correlation between parents and children, regardless of the child’s gender. This
probably means that both boys and girls have significant satisfactory and meaningful
relationships outside of those with family members, particularly with their peers, and while
family members form part of a psychosocial environment that is shared with parents, peers
and friends do not.
Our related hypothesis—to be tested in the future—is that the bigger influence of
parents’ well-being on the female child’s well-being is not specific to Catalan culture.
However, it is unclear to us whether this may be observed only in other related cultures,
like the cultures of the Mediterranean region (i.e. more family-oriented, more sexist,
cultures), or the Latin-language speaking countries, or also in more dissimilar cultures.
Our findings may make some adults uncomfortable, as parents generally expect to have
an important influence on their own children. We believe that these results may reinforce
the idea that adolescent cultures may be constructed independently of and even be unre-
lated to adult cultures (Casas 2008), particularly when communication between generations
is not appropriate or intensive enough. Such reflections may also be applicable to sub-
jective well-being, which may have a correlation between parents and children, but may
also be largely or even completely unrelated, even if objective well-being is likely to be
strongly related.
Last but not least, our results seem to support the conclusions by Casas et al. (2008) that
correlations between parents’ and their children’s overall well-being are not high enough to
provide evidence of a simple genetic effect. The fact that correlations are no lower for
domain items than for single overall items seems to indicate a detectable influence in
shared environment rather than genetics, thus disconfirming Hypothesis (b).
Acknowledgments The Spanish research presented here has been funded by the Spanish Government’sMinistry of Science and Education, with reference number SEJ2007-62813/PS.
References
Arbuckle, J. L. (2010). IBM SPSS� AmosTM
19 User’s Guide. Crawfordville, FL: Amos DevelopmentCorporation.
Batista-Foguet, J. M., & Coenders, G. (2000). Modelos de ecuaciones estructurales. Madrid: La Muralla.
Relationship Between Parents’ and Their Children’s
123
Brannick, M. T. (1995). Critical comments on applying covariance structure modeling. Journal of Orga-nizational Behaviour, 16, 201–213.
Browne, M. W., & Cudeck, R. (1993). Alternative ways of assessing model fit. In K. A. Bollen & J. S. Long(Eds.), Testing structural equation models (pp. 136–162). Thousand Oaks: Sage.
Byrne, B. M. (2010). Structural equation modeling with AMOS. Basic concepts, applications and pro-gramming (2nd ed.). New York: Routledge.
Campbell, A., Converse, P. E., & Rogers, W. L. (1976). The quality of American life: Perceptions, eval-uations, and satisfactions. New York: Russell Sage.
Casas, F. (2008). Children’s cultures and new technologies: A gap between generations? Some reflectionsfrom the Spanish context. In A. James & A. James (Eds.), European childhoods: Cultures, politics andchildhood in Europe (pp. 61–81). Houndmills, Hampshire: Palmgrave MacMillan.
Casas, F., Coenders, G., Cummins, R. A., Gonzalez, M., Figuer, C., & Malo, S. (2008). Does subjectivewell-being show a relationship between parents and their children? Journal of Happiness Studies, 9(2),197–205. Published on-line in 2007: http://dx.doi.org/10.1007/s10902-007-9044-7.
Casas, F., Sarriera, J. C., Alfaro, J., Gonzalez, M., Malo, S., Bertran, I., et al. (2011). The personal well-being index: Its functioning with 2 new items in 3 samples of 12–16 years-old adolescents in 3countries. Social Indicators Research. doi: 10.1007/s11205-011-9781-1.
Cheung, G. W., & Rensvold, R. B. (2002). Evaluating goodness-of-fit indexes for testing measurementinvariance. Structural Equation Modeling, 9, 233–255.
Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: LawrenceEarlbaum Associates.
Cummins, R. A. (2003). Normative life satisfaction: Measurement issues and a homeostatic model. SocialIndicators Research, 64, 225–256.
Cummins, R. A., & Cahill, J. (2000). Avances en la comprension de la calidad de vida subjetiva. Inter-vencion Psicosocial, 9(2), 185–198.
Cummins, R. A., Eckersley, R., Lo, S. K., Okerstrom, E., Hunter, B., & Davern, M. (2003). Australian unitywellbeing index: Cumulative psychometric record. Report 9.0. Australian Centre on quality of life.http://www.deakin.edu.au/research/acqol/instruments/PWI/Cumulative_Psychometric_Record_Australian_data.doc.
Cummins, R. A., Eckersley, R., van Pallant, J., Vugt, J., & Misajon, R. (2003). Developing a national indexof subjective well-being: The Australian unity well-being index. Social Indicators Research, 64,159–190. Updated in: http://www.deakin.edu.au/research/acqol/instruments/well-being_index.htm.
Cummins, R. A., & Gullone, E. (2000). Why we should not use 5-point Likert scales: The case for subjectivequality of life measurement. In Proceedings 2nd international conference on quality of life in cities(pp. 74–93). Singapore: National University of Singapore.
Cummins, R. A., & Lau, A .L. D. (2005). Personal wellbeing index—School children (PWI-SC) (English)(3rd ed.). Manual. Reviewed May 2006. http://www.deakin.edu.au/research/acqol/instruments/PWI/PWI-school.pdf.
Diener, E., Emmons, R., Larsen, R., & Smith, H. L. (1985). The satisfaction with life scale. Journal ofPersonality Assessment, 49(1), 71–75.
Fordyce, M. W. (1988). A review of research on the happiness measures: A sixty second index of happinessand mental health. Social Indicators Research, 20(4), 355–381.
Huebner, E. S., Seligson, J. L., Valois, R. F., & Suldo, S. M. (2006). A review of the brief multidimensionalstudents’ life satisfaction scale. Social Indicators Research, 79, 477–484.
International Wellbeing Group (2006). Personal wellbeing index—adult—manual (4th version). Melbourne:Australian centre on quality of life, Deakin University. http://www.deakin.edu.au/research/acqol/instruments/well-being_index.htm.
Kelloway, E. K. (1995). Structural equation modeling in perspective. Journal of Organizational Behaviour,16, 215–224.
Lau, A. L. D., Cummins, R. A., & McPherson, W. (2005). An investigation into the cross-cultural equiv-alence of the personal well-being index. Social Indicators Research, 72, 403–430.
Lykken, D., & Tellegen, A. (1996). Happiness is a stochastic phenomenon. Psychological Science, 7(3),186–189.
Meredith, W. (1993). Measurement invariance, factor analysis and factorial invariance. Psychometrika, 58,525–543.
Pavot, W., & Diener, E. (1993). Review of the satisfaction with life scale. Psychological Assessment, 5(2),164–172.
Rosenthal, R. (1991). Meta-analytic procedures for social research. Newbury Park, CA: Sage.Roysamb, E., Tambs, K., Reichborn-Kjennerud, T., Neale, M. C., & Harris, J. R. (2003). Happiness and
health: Environmental and genetic contributions to the relationship between subjective wellbeing,
F. Casas et al.
123
perceived health, and somatic illness. Journal of Personality and Social Psychology, 85(6),1136–1146.
Russell, J. A. (2003). Core affects and the psychological construction of emotion. Psychological Review,110(1), 145–172.
Seligson, J. L., Huebner, E. S., & Valois, R. F. (2003). Preliminary validation of the brief multidimensionalstudent’s life satisfaction scale. Social Indicators Research, 61, 121–145.
Seligson, J. L., Huebner, E. S., & Valois, R. F. (2005). Validation of a brief life satisfaction scale withelementary school students. Social Indicators Research, 73, 355–374.
Tomyn, A. J., & Cummins, R. A. (2011). The subjective wellbeing of high-school students: Validating thepersonal wellbeing index—School children. Social Indicators Research, 101, 405–374.
Veenhoven, R. (1994). El estudio de la satisfaccion con la vida. Intervencion Psicosocial, III(9), 87–116;IV(10), 125–127.
Veenhoven, R. (2009). Medidas de felicidad nacional bruta. Intervencion Psicosocial, 18(3), 279–299.
Relationship Between Parents’ and Their Children’s
123