Demographic characteristics of childless middle-aged and elderly couples
Table 1 presents the demographic characteristics of childless middle-aged and elderly couples. The average ages were 69.43 (66.19) years for husbands (wives). High school graduates were highest among husbands at 42.3%; followed by elementary school graduates and lower at 29.0%; and middle school graduates, 16.9%. Among wives, 42.0% were elementary school graduates and lower; 34.3%, high school graduates; and 17.9%, middle school graduates. Among the couples participating in the study, 48.3% (33.3%) of husbands (wives) engaged in economic activities. Regarding their economic activity, most fell under dual-unemployed at 41.5%; single-income, 35.3%; and dual-income, 23.2%. On the subjective economic level, 52.7% (53.1%) of husbands (wives) answered “low.”
Table 1
Demographic characteristics of couples (n = 207 couples)
|
Couples
|
Wives
|
Husbands
|
n(%)
|
n(%)
|
n(%)
|
Age (years)
|
Range
|
|
57–90
|
57–91
|
|
M ± SD
|
|
66.19 ± 7.50
|
69.43 ± 7.95
|
|
≤ 64
|
|
101 (48.8)
|
68 (32.9)
|
|
65–74
|
|
69 (33.4)
|
85 (41.1)
|
|
75–84
|
|
33 (15.9)
|
46 (22.2)
|
|
≥ 85
|
|
4 (1.9)
|
8 (3.8)
|
Education
|
≤ Elementary
|
|
87 (42.0)
|
60 (29.0)
|
|
Middle
|
|
37 (17.9)
|
35 (16.9)
|
|
High
|
|
71 (34.3)
|
88 (42.5)
|
|
≥ College
|
|
12 (5.8)
|
24 (11.6)
|
Perceived economic status
|
High
|
|
3 (1.4)
|
5 (2.4)
|
|
Medium
|
|
94 (45.4)
|
93 (44.9)
|
|
Low
|
|
110 (53.1)
|
109 (52.7)
|
Economic activity
|
Yes
|
|
69 (33.3)
|
100 (48.3)
|
|
No
|
|
138 (66.7)
|
107 (51.7)
|
|
Dual-income
|
48 (23.2)
|
|
|
|
Single-income
|
73 (35.3)
|
|
|
|
Dual-unemployed
|
86 (41.5)
|
|
|
M ± SD = mean ± standard deviation
|
Life satisfaction and depression of male and female dyads
As Table 2 shows, wives had lower life satisfaction than husbands (5.83 vs. 5.95; p = 0.012). Moreover, wives had higher depression than husbands (7.28 vs. 7.04; p < 0.001). Further, regarding the subscales of depression, wives had higher or similar scores than husbands in all domains of depression.
Table 2
Level of life satisfaction and depression among childless middle-aged and elderly couples (n = 207 couples)
Variables
|
Wives
|
Husbands
|
t value
|
p-value
|
Life satisfaction
|
5.83 ± 1.45
|
5.95 ± 1.46
|
5.94
|
0.012
|
Health state
|
5.38 ± 1.81
|
5.49 ± 1.97
|
-2.07
|
0.039
|
Economic state
|
5.37 ± 1.76
|
5.33 ± 1.80
|
-1.04
|
0.069
|
Spouse satisfaction
|
6.41 ± 1.77
|
6.77 ± 1.75
|
6.650
|
< 0.001
|
Overall quality of life
|
6.14 ± 1.74
|
6.20 ± 1.69
|
-8.75
|
< 0.001
|
Depression
|
7.28 ± 6.05
|
7.04 ± 6.48
|
-1.43
|
< 0.001
|
Depressive emotion
|
1.76 ± 2.13
|
1.70 ± 2.41
|
-8.57
|
< 0.001
|
Positive emotion
|
2.65 ± 2.04
|
2.48 ± 1.98
|
-12.50
|
< 0.001
|
Somatic symptom
|
1.86 ± 2.21
|
1.84 ± 2.47
|
-1.41
|
0.095
|
Interpersonal relationship
|
1.01 ± 1.42
|
1.02 ± 1.45
|
-6.24
|
< 0.001
|
M ± SD = mean ± standard deviation |
Test of the measurement model
Before verifying the structural model, CFA via AMOS 25.0 was conducted to verify the validity of the questionnaire items and factors to be incorporated into the structural model. Accordingly, the convergent and discriminant validities were verified, as described below.
Standardized Factor Loadings (FL) > 0.70, Construct Reliability (CR) > 0.70, and Average Variance Extracted (AVE) > 0.50 between the questionnaire and factors verified the convergent validity [24]. Among the items to measure depression, the positive emotions of wives and husbands did not meet the criteria and were, thus, removed. That is, three sub-factors of depression for wives and husbands and four sub-factors of life satisfaction were used in the final analysis to verify conceptual reliability and average variance extraction. Thus, convergent validity was verified above the standard values of 0.70 and 0.50.
Next, this study verified whether the correlation coefficient between the factors was less than the square root of the AVE of each factor to verify the discriminant validity. Accordingly, the correlation coefficient of all factors was lower than the square root of AVE, verifying the discriminant validity.
Finally, to understand how the measurement variables used in this study account for latent variables, we tested the measurement model regarding the life satisfaction and depression of husbands and wives. The results were as follows: χ2 = 66.942 (p < 0.001); degree of freedom (df) = 37; χ2 /df = 1.762; NFI = 0.963; GFI = 0.950; AGFI = 0.900; CFI = 0.980; IFI = 0.982; TLI = 0.966; RMSEA = 0.061; and SRMR = 0.069. The value of the χ2/df index is ≤ 3. SRMR is acceptable if the value is ≤ 0.08 [25]. RMSEA, which considers both model error and simplicity simultaneously, is appropriate if it is ≤ 0.08. NFI, the standard GFI, is appropriate if it is ≥ 0.80. GFI and AGFI values are appropriate if higher than 0.90 [26]. Incremental fit indices, CFI, IFI, and TLI estimates are good if they are ≥ 0.9, being close to 1 [25].
Test of the structural model
Further, to find the actor and partner effects of couples’ life satisfaction on depression, the study evaluated the normality of the measured variables before modeling a structural equation. Moreover, to test the univariate normality of the measured variables, we calculated the skewness and kurtosis. The assumption of normal distribution was satisfied because the skewness was ≤ 3, and kurtosis was ≤ 10 for both husbands and wives. Additionally, as Table 3 shows, correlations ranged from − 0.546 to 0.773.
Table 3
Correlation between life satisfaction and depression among childless middle-aged and elder couples (n = 207 couples) p-value: < 0.05 *, < 0.01**
|
1
|
2
|
3
|
4
|
5
|
6
|
7
|
8
|
9
|
10
|
11
|
12
|
13
|
14
|
15
|
16
|
Wives’ life satisfaction
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 Health state
|
1
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 Economic state
|
.59**
|
1
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 Spouse satisfaction
|
.51**
|
.52**
|
1
|
|
|
|
|
|
|
|
|
|
|
|
|
|
4 Overall quality of life
|
.51**
|
.57**
|
.62**
|
1
|
|
|
|
|
|
|
|
|
|
|
|
|
Wives’ depression
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 Depressive emotion
|
− .46**
|
− .28**
|
− .52**
|
− .37**
|
1
|
|
|
|
|
|
|
|
|
|
|
|
6 Positive emotion
|
− .24**
|
− .18**
|
− .25**
|
− .17*
|
.20**
|
1
|
|
|
|
|
|
|
|
|
|
|
7 Somatic symptom
|
− .51**
|
− .34**
|
− .47**
|
− .35**
|
.76**
|
.19**
|
1
|
|
|
|
|
|
|
|
|
|
8 Interpersonal relationship
|
− .36**
|
− .26**
|
− .43**
|
− .25**
|
.71**
|
.13
|
.68**
|
1
|
|
|
|
|
|
|
|
|
Husbands’ life satisfaction
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
9 Health state
|
.51**
|
.47**
|
.32**
|
.41**
|
− .44**
|
− .21**
|
− .51**
|
− .44**
|
1
|
|
|
|
|
|
|
|
10 Economic state
|
.44**
|
.60**
|
.36**
|
.51**
|
− .25**
|
− .12
|
− .32**
|
− .21**
|
.63**
|
1
|
|
|
|
|
|
|
11 Spouse satisfaction
|
.35**
|
.34**
|
.56**
|
.40**
|
− .41**
|
− .23**
|
− .40**
|
− .38**
|
.42**
|
.40**
|
1
|
|
|
|
|
|
12 Overall quality of life
|
.44**
|
.49**
|
.47**
|
.58**
|
− .39**
|
− .15**
|
− .46**
|
− .34**
|
.57**
|
.62**
|
.58**
|
1
|
|
|
|
|
Husbands’ depression
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
13 Depressive emotion
|
− .41**
|
− .31**
|
− .44**
|
− .30**
|
.71**
|
.19**
|
.70**
|
.63**
|
− .54**
|
− .29**
|
− .39**
|
− .42**
|
1
|
|
|
|
14 Positive emotion
|
− .23**
|
− .123
|
− .24**
|
− .14
|
.19**
|
.77**
|
.17*
|
.14*
|
− .19**
|
− .09
|
− .27**
|
− .12
|
.19**
|
1
|
|
|
15 Somatic symptom
|
− .40**
|
− .38**
|
− .40**
|
− .32**
|
.65**
|
.15*
|
.69**
|
.57**
|
− .55**
|
− .34**
|
− .32**
|
− .40**
|
.77**
|
.14*
|
1
|
|
16 Interpersonal relationship
|
− .38**
|
− .24**
|
− .36**
|
− .28**
|
.57**
|
.17*
|
.62**
|
.69**
|
− .42**
|
− .19**
|
− .33**
|
− .27**
|
.69**
|
.16*
|
.64**
|
1
|
The model has the following characteristics: χ2 = 66.942 (p < 0.001), df = 38, χ2/df = 1.762, NFI = 0.963, GFI = 0.950, AGFI = 0.900, CFI = 0.984, IFI = 0.984, TLI = 0.966, RMSEA = 0.061, and SRMR = 0.069. Thus, the model was found to have a good fit, as shown in Table 4.
Impact of life satisfaction on depression at the dyadic level
Table 4 and Fig. 1 shows the results of the actor and partner effects of couples’ life satisfaction on depression. The life satisfaction of wives (β = -0.285, p = 0.004) and husbands (β = -0.403, p < 0.001) had a significant effect on individual depression. Husbands’ life satisfaction (β = -0.255, p = 0.011) had a significant effect on wives’ depression, and wives’ life satisfaction (β = -0.375, p < 0.001) had a significant effect on husbands’ depression.
Table 4
Effect coefficients for the hypothetical model (n = 207 couples)
|
Wives
|
Husbands
|
Β (SE)
|
t value
|
p-value
|
Β (SE)
|
t value
|
p-value
|
Total depression
|
Actor’s life satisfaction
|
-0.285(0.046)
|
2.889
|
0.004
|
-0.403(0.025)
|
3.880
|
< 0.001
|
Partner’s life satisfaction
|
-0.375(0.038)
|
3.766
|
< 0.001
|
-0.255(0.029)
|
2.530
|
0.011
|
Depressive emotion
|
Actor’s life satisfaction
|
-0.197(0.062)
|
1.449
|
0.007
|
-0.401(0.026)
|
3.420
|
< 0.001
|
Partner’s life satisfaction
|
-0.395(0.021)
|
3.890
|
< 0.001
|
-0.376(0.021)
|
3.877
|
< 0.001
|
Somatic symptom
|
Actor’s life satisfaction
|
-0.554(0.029)
|
5.074
|
< 0.001
|
-0.371(0.037)
|
2.267
|
< 0.001
|
Partner’s life satisfaction
|
-0.122(0.051)
|
1.377
|
0.030
|
-0.108(0.064)
|
1.382
|
0.142
|
Interpersonal relationship
|
Actor’s life satisfaction
|
-0.566(0.027)
|
5.102
|
< 0.001
|
-0.339(0.040)
|
1.178
|
< 0.001
|
Partner’s life satisfaction
|
-0.382(0.029)
|
3.857
|
< 0.001
|
-0.096(0.071)
|
1.215
|
0.168
|
Regarding the effect on the depressive emotion domain, the life satisfaction of wives (β = -0.197, p = 0.007) and husbands (β = -0.401, p < 0.001) as actors significantly affected their depressive emotion domain. Regarding partner’s life satisfaction, husbands’ life satisfaction (β = -0.395, p < 0.001) significantly affected wives’ depressive emotion domain, and wives’ life satisfaction (β = -0.376, p < 0.001) significantly affected husbands’ depressive emotion domain.
Regarding the effect on the somatic symptom domain, the life satisfaction of wives (β = -0.554, p < 0.001) and husbands (β = -0.371, p < 0.001) as actors had a significant effect on their somatic symptom domain. As partners, husbands’ life satisfaction (β = -0.122, p = 0.030) significantly affected wives’ somatic symptom domain, whereas wives’ life satisfaction (β = -0.108, p = 0.142) had no significant effect on husbands’ somatic symptom domain.
Regarding the effect on the interpersonal relationship domain, the life satisfaction of wives (β = -0.566, p < 0.001) and husbands (β = -0.339, p < 0.001) as actors significantly affected their interpersonal relationship domain. As partners, husbands’ life satisfaction (β = -0.382, p < 0.001) significantly affected wives’ interpersonal relationship domain, whereas wives’ life satisfaction (β = -0.096, p = 0.168) had no significant effect on husbands’ interpersonal relationship domain.
This study examines whether the actor and the partner effects differed between men and women; the two coefficients were considered equal and compared using the chi-square test for the constrained and unconstrained (saturated) models (Table 5). No significant difference was noted between the constrained model, where both the husband and wife’s actor effects were constrained equally, and the unconstrained model (χ2 = 0.344, p = 0.557). This result reveals that the life satisfaction effect of husband and wife actors had a similar impact on depression. However, there was a significant difference between the constrained model, where the partner effects of husbands and wives were constrained equally, and the unconstrained model (χ2 = 9.455, p = 0.002). The findings indicate that the life satisfaction partner effect of husbands significantly impacted depression more than that of wives. Further, there was a significant difference in the equivalence constraints between the actor and partner effects for husbands’ life satisfaction (χ2 = 6.360, p = 0.030), and that of wives had no significant difference. (χ2 = 2.988, p = 0.084). This result reveals that the partner effect of husbands’ life satisfaction on depression was more influential than that of wives.
Table 5
χ2 Differences in the test between the basic and equivalent constraint models
Model
|
χ2
|
df
|
TLI
|
CFI
|
RMSEA
|
△χ2
|
p-value
|
basic
|
66.942
|
38
|
.966
|
.984
|
.061
|
|
|
equivalence constraint 1 (a = b)
|
67.286
|
39
|
.973
|
.984
|
.059
|
0.344
|
0.557
|
equivalence constraint 2 (a1 = b1)
|
76.397
|
39
|
.972
|
.984
|
.060
|
9.455
|
0.002
|
equivalence constraint 3 (a = b1)
|
73.302
|
39
|
.970
|
.983
|
.062
|
6.360
|
0.030
|
equivalence constraint 4 (b = a1)
|
69.930
|
39
|
.970
|
.982
|
.062
|
2.988
|
0.084
|
CFI = Comparative Fit Index; df = Degree of freedom; RMSEA = Root Mean Squared Error of Approximation; TLI = Turker–Lewis Index; a = Husband’s actor effects; b = Wives’ actor effects; a1 = Husbands’ partner effects; b1 = Wives’ partner effects |