-
Journal of Family Psychology2000, Vol. 14, No. 3, 420-435
Copyright 2000 by the American Psychological Association,
Inc.0893-3200/O0/$5.0D DOI: 10.1O37//O893-32OO.14.3.420
Parenting Practices and Adolescent DepressiveSymptoms in Chinese
American Families
Su Yeong Kim and Xiaojia GeUniversity of California, Davis
This study examined parenting practices and adolescent
depressive symptomsamong Chinese Americans. First, confirmatory
factor analyses revealed that self-reported parenting practices by
mothers and fathers and adolescent perception ofparenting practices
loaded adequately on three subscales: Inductive
Reasoning,Monitoring, and Harsh Discipline. Second, parents'
depressive symptoms wererelated to disrupted parenting practices,
which, in turn, were significantly related tothe negative
evaluation of these behaviors by the adolescents. Adolescents'
per-ceptions of such parenting practices were significantly
associated with their de-pressive symptoms. Third, the
relationships were robust even after parental income,education, and
generation status were statistically controlled. Overall, the
relation-ships between parenting practices and adolescent
depressive symptoms amongChinese Americans seemed to echo those
found among European Americans.
The rising prevalence of depression duringadolescence has drawn
increased research atten-tion (Compas, Ey, & Grant, 1993; Ge,
Lorenz,Conger, Elder, & Simons, 1994; Kandel & Da-
Su Yeong Kim and Xiaojia Ge, Department ofHuman and Community
Development, University ofCalifornia, Davis.
Support was provided through a Scott Mesh Hon-orary Scholarship
for Research in Psychology fromthe American Psychological
Association of GraduateStudents; a Jastro-Shields Research Award
from theUniversity of California, Davis, College of Agricul-tural
and Environmental Sciences; and a GraduateResearch Mentorship Award
from the University ofCalifornia, Davis, Office of Graduate
Studies.
We would like to thank the administrators at theschool district
and the families that participated in thestudy. Research assistance
is highly appreciated fromMarilyn Luong, Selina Li, Wilson Luk,
Sun-MeeKang, Cindy Kwan-Yee Lau, Chor-Yi Wong,Charles Inada, Jenny
Jiun-Ling Wang, Kunise Stroh,Laura Kawate, Marco Cordova, Elaine
Chinn, andElizabeth Wong. Valuable comments on an earlierversion of
this article were received from M. BrentDonnellan.
Correspondence concerning this article should beaddressed to Su
Yeong Kim, Department of Humanand Community Development, University
of Califor-nia, 1 Shields Avenue, Davis, California
95616-8523.Electronic mail may be sent to [email protected].
vies, 1982; Radloff, 1991). Various factors thatmay explain this
phenomenon have been exam-ined. One line of research involves
investigationof the linkage between parenting and
adolescentdepressive symptoms. Research conductedamong European
American families suggeststhat the way parents interact with their
childrenis related to adolescent depressive symptoms.For example,
several studies (Ge, Best, Conger,& Simons, 1996; Ge, Conger,
Lorenz, & Si-mons, 1994) have shown that parenting withwarmth
and involvement is negatively related todepressive symptoms and
that parenting char-acterized by harshness and disciplinary
incon-sistency is positively related to adolescent de-pressive
symptoms in European Americanfamilies. However, it is unclear
whether similarrelationships exist within ethnic minority
fami-lies. To fill this gap in knowledge, the presentstudy explored
the relationship between parent-ing processes and adolescent
depressive symp-toms using a sample of Chinese American fam-ilies,
the largest Asian American ethnic group inthe United States (Uba,
1994).
Despite the popular "model minority" image,studies suggest that
some Asian Americanadults do experience an elevated level of
de-pressive symptoms (Aldwin & Greenberger,1987; Kuo, 1984;
Okazaki, 1997; Ying, 1995).Although studies on Asian American
adoles-
420
-
SPECIAL ISSUE: PARENTING AND DEPRESSIVE SYMPTOMS 421
cents are fewer than those involving adults,there is some
evidence that adolescents alsoexhibit elevated distress levels (Kim
& Chun,1993). Important progress has been made inidentifying
some of the possible predictors ofdepressive symptoms among Asian
Americans,including different cultural orientations (Ying,1995),
recency of arrival (Kuo, 1984), and typeof self-construal (Okazaki,
1997). Relativelyfew studies, however, have explored the
signif-icance of family processes in adolescent depres-sive
symptoms.
Previous studies of Asian American parent-ing practices tended
to focus on the averagedifferences between Asian and European
Amer-ican parents. In comparison with their EuropeanAmerican
counterparts, Asian American parentsare described as more
controlling, stricter indiscipline, and more restrictive (Chao,
1994;Kelley & Tseng, 1992; Steinberg, Dombusch,& Brown,
1992). When individual differenceswithin Asian Americans are
examined, studieshave primarily focused on the relationship
be-tween parenting practices and students1 aca-demic achievement
(Dombusch, Ritter, Leider-man, Roberts, & Fraieigh, 1987;
Steinberg etal., 1992). Relatively little attention has beenpaid to
individual differences in Chinese Amer-ican parenting practices and
their associationwith adolescent socioemotional
developmentaloutcomes.
There is some evidence to suggest that, de-spite the significant
mean differences in parent-ing practices between Asian and
EuropeanAmericans, the association between parentingbehaviors and
some psychological outcomesappears to be similar. For example,
Steinberg,Mounts, Lamborn, and Dombusch (1991) re-ported that
authoritative parenting, conceptual-ized as a composite of
acceptance-involvement,firm control, and democratic discipline, is
asso-ciated with less depressive symptoms amongAsian American
adolescents. Other studies havealso shown that warm parenting is
negativelyrelated to depressive symptoms among adoles-cents of both
European and Asian descent(Chiu, Feldman, & Rosenthal, 1992;
Green-berger & Chen, 1996). Collectively, the re-search
suggests that, within Asian Americanfamilies, parenting practices
characterized byhigher levels of warmth, involvement, and
dem-ocratic disciplinary practices are associatedwith decreased
levels of adolescent adjustment
problems. On the basis of these findings, weexpected that Asian
American adolescentswhose parents were more involved with
theirlives through monitoring their activities wouldmanifest fewer
depressive symptoms. We alsoexpected that those adolescents whose
parentsused more reasoning in their disciplinary prac-tices would
display fewer depressive symptoms.
Although parental warmth has been found tobe associated with
less depressive symptomsamong Chinese or Chinese American
adoles-cents (Chiu et al., 1992; Greenberger & Chen,1996), the
relationship between parenting di-mensions other than warmth and
adolescent de-pressive symptoms is less well explored. Thismay be
due in part to limited understanding ofparenting practices among
Asian Americans.When examining parenting practices in
ChineseAmericans, researchers remain uncertain as towhich aspects
of parenting other than warmthmay be related to adolescent
depressive symp-toms. For example, whereas Chiu et al.
(1992)examined parental control and involvement,Greenberger and
Chen (1996) focused on fam-ily conflict and cohesion. The
inconsistencyacross previous studies in the assessment ofparenting
practices suggests a need for a carefulexamination of measurement
strategies. Thus,the first phase of this study was devoted to
anexamination of several parenting dimensionsamong Chinese American
families.
Moreover, the existing studies of AsianAmerican parenting
practices typically adoptedBaumrind's (1991) parenting typology
(Dom-busch et al., 1987; Leung, Lau, & Lam, 1998).Recently,
however, Chao (1994) and Gonzales,Cauce, and Mason (1996) cautioned
that such aframework may not be entirely appropriate forassessing
parents from other cultures. In light ofthese warnings, the first
phase of the studysought to ensure the adequacy of the measures.In
doing so, we examined whether several par-enting dimensions
developed for EuropeanAmericans, such as inductive reasoning,
moni-toring, and harsh discipline, would be appropri-ate to the
study of parenting practices amongChinese American families.
Studies of Asian American parenting typi-cally adopt measures
used with EuropeanAmericans (Chiu et al., 1992; Steinberg et
al.,1991). Although ethnic differences have beenobserved (Chao,
1994; Gonzales et al., 1996),researchers have not adequately
explored
-
422 KIM AND GE
whether parenting items used to assess induc-tive reasoning,
monitoring, and harsh disciplinein European Americans would be
appropriatefor the assessment of Chinese American parent-ing. It is
plausible to expect Chinese Americanparents who are high in
inductive reasoning toprovide more explanations, give reasons
fortheir decisions, ask their adolescent for opinionswhen making
decisions, and explain rules totheir adolescent. Similarly, Chinese
Americanparents who are high in monitoring are expectedto know the
whereabouts of their adolescent,whom their adolescent is with, and
whethertheir adolescent complies with the set bedtime.Conversely,
Chinese American parents who arehigh in harsh discipline may be
more likely tospank and hit their adolescent or lock the
ado-lescent out of the house. In other words,whereas the level at
which parents use certainparenting practices may differ across
ethnicity,items making up the dimensions should showa certain
degree of similarity regardless ofethnicity.
Previous research typically examined the in-fluence of parents
on Asian American adoles-cent developmental outcomes using child
re-ports of parenting behaviors. In the presentstudy, data from
multiple informants, includingmothers, fathers, and adolescents,
were mod-eled with a confirmatory factor-analyticalframework to
provide a more comprehensiveunderstanding of Chinese American
family pro-cesses. Two general hypotheses guided this firstphase of
analyses. First, we hypothesized thatthree dimensions of parenting
practices (i.e.,harsh discipline, inductive reasoning, and
mon-itoring) would emerge, as evidenced by (a) anadequate fit of a
measurement model to our dataand (b) satisfactory factor loadings
of the as-sessment items on the latent dimensions. Sec-ond, this
three-dimensional factor pattern wouldbe similar across multiple
informants, as evi-denced by (a) a consistent pattern across
multi-ple informants and (b) a general absence ofsignificant
differences in estimated factorloadings.
The second phase of this study was devotedto testing a process
model linking parents' de-pressive symptoms, parenting practices,
and ad-olescent depressive symptoms. Three generalhypotheses were
tested. First, we hypothesizedthat parents' depressive symptoms
would beassociated with their parenting practices. This
hypothesis was derived from Downey andCoyne (1990), who
suggested that depressionacts to reduce a parent's ability to carry
outeffective parenting tasks. According to theseauthors, depressed
mood may increase irritableresponses by parents toward their
children andhinder their level of involvement with them.Indeed,
Conger, Conger, Elder, and Lorenz(1992) have shown that depressed
mood signif-icantly increases harsh and hostile
disciplinarypractices among European American parents. Inthe
current study, we expected depressivesymptoms of parents to be
negatively associatedwith parenting practices characterized by
mon-itoring and inductive reasoning techniques andpositively
related to parenting practices charac-terized by harsh
disciplining.
Second, we hypothesized that the way par-ents interact with
their adolescent child wouldbe significantly related to the
adolescent's ap-praisal of these behaviors. In other words,
howparents behave toward their children signifi-cantly influences
the cognitive evaluation ofthese behaviors by the adolescent. An
increas-ing number of studies suggest that adolescents'cognitive
evaluation of parenting plays an im-portant role in their emotional
development. Forexample, Harold, Osborne, and Conger (1997)have
recently shown that adolescents* percep-tions of parenting
behaviors are directly associ-ated with their developmental
outcomes. Draw-ing on this line of research, we expected
thatself-reported parenting practices by mothersand fathers would
be significantly related toadolescents' perceptions of parenting
practices.
Finally, we hypothesized that how adoles-cents perceived
parenting practices would besignificantly associated with their own
depres-sive symptoms. This hypothesis is consistentwith Conger et
al. (1992) and Patterson, De-Baryshe, and Ramsey (1989), who
suggestedthat disrupted parenting practices are signifi-cantly
related to adverse adolescent outcomes.It is also consistent with
Harold et al.'s (1997)finding that the child's perception of
parentalhostility has a direct effect on child maladjust-ment. On
the basis of these findings, we ex-pected that adolescents'
evaluations of parent-ing practices would be significantly
associatedwith their own depressive symptoms. Specifi-cally, we
expected that parenting practices char-acterized by higher degrees
of harshness andlower degrees of monitoring and inductive rea-
-
SPECIAL ISSUE: PARENTING AND DEPRESSIVE SYMPTOMS 423
sorting would be associated with reports of moreadolescent
depressive symptoms. We reasonedthat a higher level of perceived
harshness fromparents would heighten adolescents' feelings
ofhostility from their surroundings, which in turnwould increase
their sense of hopelessness anddecrease their sense of self-worth.
Similarly,adolescents who perceive a lack of monitoringby their
parents may experience increased feel-ings of parental
irresponsibility, which may re-sult in an increased sense of
helplessness. Fi-nally, parents1 lack of inductive reasoning
andexplanation may exacerbate an adolescent'ssense of uncertainty
and sense of frustration,which are important components of
depressivesymptoms.
Despite the theoretical appeal of the preced-ing hypotheses,
limited research has been di-rectly tailored toward testing these
hypothe-sized relationships among Asian Americanadolescents. We are
not aware of any studiesthat have examined the relationship
betweenharsh disciplinary practices and depressivesymptoms in this
population. One study (Chiuet al., 1992) did include items related
to parentalmonitoring. However, the monitoring itemswere combined
with items related to parentalschool engagement to form a composite
vari-able. In another study (Steinberg et al., 1991),questions
similar to the parental monitoring di-mension were subsumed under
the authoritativeparenting construct. Although Chiu et al.
(1992)reported no significant association betweenschool
engagement-monitoring and adolescentdepressive symptoms, Steinberg
et al. (1991)reported a significant association between
au-thoritative parenting and adolescent depressivesymptoms. It is
apparent that a clear associationbetween the monitoring dimension
and adoles-cent depressive symptoms remains to be inves-tigated.
Although studies have examined ethnicdifferences in parental use of
inductive reason-ing (Kelley & Tseng, 1992;
Kobayashi-Winata& Power, 1989), it is unclear whether, within
agroup, those adolescents who perceive their par-ents to be higher
in levels of inductive reason-ing would show lower levels of
depressivesymptoms.
In the third phase of the study, we sought totest the robustness
of the hypothesized relation-ships by statistically controlling for
adolescentgender and parent generation status, income,and
education. Gender was selected because
several studies have reported significant genderdifferences in
depressive symptoms during ad-olescence (Ge, Lorenz, et al., 1994;
Nolen-Hoeksema & Girgus, 1994). The remainingcontrol variables
were selected because of thesalient roles they play in ethnic
minority familyprocesses (Garcia Coll, Meyer, & Brillon,
1995;Uba, 1994). The research focus during thisphase was on the
possible alteration of esti-mated relationships among the variables
in themodel after the inclusion of control variables.We anticipated
that first-generation status ofparents would be related to more
adolescentdepressive symptoms as a result of intergenera-tional
conflicts in immigrant families (e.g.,Dinh, Sarason, & Sarason,
1994; Rumbaut,1994). We also predicted that higher incomeand
education levels of parents would relate tolower levels of
depressive symptoms amongadolescents, mothers, and fathers,
consonantwith the large body of literature showing anegative
relationship between socioeconomicstatus and depressive symptoms
(e.g., Dohren-wend, 1990). We hypothesized that the modeltested in
the second phase of the study wouldnot be significantly altered
after the inclusion ofthese control variables. That is, we
expectedthat the models should remain fit to the data andthat the
structural coefficients would not besignificantly reduced.
In all three phases of analyses, we modeledour data separately
for mothers and fathers.Previous research on Asian American
parentingpractices relied almost exclusively on data con-cerning
mothers (Chao, 1996; Chen, Green-berger, Lester, Dong, & Guo,
1998). This canbe attributed to the view that Asian Americanmothers
assume the primary caregiving role asopposed to Asian American
fathers (Uba,1994). Given the importance of fathering sug-gested by
Parke (1996) and Lamb (1997), thisstudy explored the role of both
mothers1 andfathers1 parenting practices in adolescent de-pressive
symptoms.
Method
Procedure
Participants were recruited from four middleschools in a major
metropolitan area of northernCalifornia. With the aid of school
administrators, theresearchers identified all Chinese American
students.All eligible families were then sent a letter
describing
-
424 KIM AND GE
the research project and were informed that a re-search
assistant would be contacting them by tele-phone to request their
participation in the studywithin a few weeks. Bilingual research
assistantswere available to contact families whose home lan-guage
was not English. After phone consent had beenobtained,
questionnaires for the mother, father, andtarget adolescent in the
family were mailed to thehome. Participants were instructed to
complete thequestionnaires alone and to not discuss answers
withfriends or family members. They were also instructedto seal
their questionnaires in the envelopes providedimmediately after
completing their responses. Withinapproximately 3-5 weeks of
sending the question-naire packet, researchers visited each middle
schooltwice to collect the completed questionnaires on des-ignated
days during the students' lunch period. Tar-get adolescents
returning the family questionnaireswere compensated a nominal
amount of money fortheir participation.
Participants
Intact families were recruited to examine the in-fluence of both
mothers and fathers on adolescentoutcome. From a pool of 254
families, 59 familiesdeclined to participate. Of the 195 families
thatagreed to participate in the study, 56% of adoles-cents, 52% of
mothers, and 50% of fathers returnedcompleted questionnaires.
However, responses from14 adolescents, 15 mothers, and 14 fathers
weredropped as a result of mixed race-ethnicity mar-riages or
parental divorce. The final sample consistedof 95 adolescents (52%
female) along with 86 moth-ers and 84 fathers. One study (Chen et
aL, 1998) witha similar study design reported a comparable
partic-ipation rate of 58% for Chinese American mothers.Comparable
participation rates for adolescents andfathers were not available.
The participation rate inthe present study was well above the 30%
averageparticipation rate for mailed surveys (Cauce, Ryan,
&Grove, 1998).
The families were of working-class to middle-classbackgrounds.
The mean family income level wasslightly under $40,000 a year, with
the average parentcompleting high school but not college.
Eighty-eightpercent of mothers and 86% of fathers were em-ployed.
The largest self-reported category for occu-pation was professional
(mother, 26%; father, 45%).Average ages were 43.58 years for
mothers and 47.24years for fathers. Average length of marriage
wasabout 18 years. Most parents were immigrants (77%of mothers and
74% of fathers). Among the latergenerations, parents were more
likely to be secondgeneration (mother, 14%; father, 18%) than
thirdgeneration or beyond (mother, 9%; father, 8%). Themajority
(70%) of couples were of the same genera-tion status. Couples of
different generations weremore likely to be non-immigrants. The
mean age of
adolescent respondents was 13.05 years (SD 0.75),representative
of seventh (52%) and eighth grades(48%). The majority of
adolescents were born in theUnited States, with 58% being second
generation and27% being third generation and beyond. Only 15% ofthe
adolescents were immigrants. The high propor-tions of
first-generation parents and second-generation adolescents in this
study parallel nationaldemographic patterns for Chinese Americans
(Portes& Zhou, 1993; Zhou & Bankston, 1998).
As a means of including first-generation parentswhose language
of preference for completing ques-tionnaires was not English,
parent questionnaireswere prepared in English and Chinese. The
question-naires were first translated to Chinese and then
back-translated to English. Any inconsistencies with theoriginal
English version scale were then resolved bytwo bilingual speakers
with careful consideration ofculturally appropriate meanings of
items. Fifty-twopercent of mothers and 55% of fathers completed
theEnglish version questionnaires, and the other mothersand fathers
completed Chinese version questionnaires.
Measures
Parenting practices. Parenting practices wereassessed through a
scale adapted from the IowaYouth and Families Project (Conger,
Patterson, &Ge, 1995; Ge et al., 1996). Using a scale
rangingfrom 1 (always) to 7 (never), respondents rated 10items
intended to measure three dimensions of par-enting: inductive
reasoning (e.g., give reasons fordecisions and explain rules),
monitoring (e.g., knowwhereabouts of adolescent, whom adolescent is
with,and whether the adolescent complies with the setbedtime), and
harsh discipline (e.g., spank, hit, orlock adolescent out of the
house). These measureswere internally consistent, with alpha
coefficientsranging from .64 to .83 (Table 1). There were nomajor
differences in internal consistencies across lan-guage versions.
Because the statistical distribution ofthe harsh discipline
dimension was skewed for allinformants, these items were
log-transformed.
Depressive symptoms. Depressive symptomswere assessed with the
Center for EpidemiologicStudies Depression Scale (CES-D; Radloff,
1977).
Table 1Internal Consistency Alpha Coefficients forParenting
Practices Measure
Variable
Inductive reasoningMonitoringHarsh discipline
Adolescentperception
Mother Father
.71 .82
.64 .75
.81 .78
Self-report
Mother
.81
.73
.79
Father
.71
.83
.74
-
SPECIAL ISSUE: PARENTING AND DEPRESSIVE SYMPTOMS 425
This 20-item scale has been found to be reliable foruse with
adults as well as adolescents (Radloff.1991). It has been used in
many studies involvingadolescents from both European (e.g., Ge et
al., 1996;Lamborn, Mounts, Steinberg, & Dornbusch, 1991)and
Asian (e.g., Chiu et al., 1992; Greenberger &Chen, 1996)
ancestries. Using a scale ranging from 0{rarely or none of the
time) to 3 (most or all of thetime), respondents rated their
depressed affect (e.g.,lonely or sad), positive affect (e.g., happy
or hope-ful), somatic symptoms (e.g., poor appetite or
restlesssleep), and interpersonal problems (e.g., people
areunfriendly). The measure indicated good internalconsistencies
across adolescents (a = .88), mothers(a = .91), and fathers (a =
.89). No major differ-ences were found in the internal
consistencies acrosslanguage versions. We also log-transformed
theCES-D measure because of its skewness. Descriptivestatistics and
zero-order correlations for all studyvariables are presented in the
Appendix.
Demographic variables. Information was col-lected on demographic
characteristics such as age,years of marriage, family income,
ethnicity, highestlevel of education completed, and generation
status.Both mothers and fathers were queried regardingannual gross
family income. Generation status wasdetermined by considering the
U.S.-born or foreign-born status of the respondent and his or her
parents.The first-generation or immigrant designation wasassigned
to individuals born outside the United Sta-tus. The
second-generation designation was assignedto individuals born in
the United States with a parentborn outside the United States. The
third-generationdesignation was assigned to individuals born in
theUnited States with a parent who was also bom in theUnited
States. Fourth-generation status and fifth-generation status were
determined accordingly.
Results
Measurement Model of ParentingDimensions
The first phase of the analyses examined theadequacy of the
hypothesized parenting dimen-sions for use with Asian Americans
because theparenting scales were originally developed forEuropean
Americans. The confirmatory factor-analytical model for the three
hypothesized par-enting dimensions is presented in Figure 1. Inthis
model, the A coefficients represent the fac-tor loadings of the 10
items on the three hypoth-esized parenting latent constructs:
inductivereasoning, monitoring, and harsh discipline.These latent
constructs were also hypothesizedto be intercorrelated, as
indicated by the i/r co-efficients. This model was tested for (a)
adoles-cents' perceptions of mothers' parenting, (b)adolescents'
perceptions of fathers' parenting,(c) self-reported parenting by
mothers, and (d)self-reported parenting by fathers.
Table 2 presents the descriptions of the par-enting items, model
fit indexes, and standard-ized and unstandardized coefficients
based onmaximum-likelihood estimates for each of thefour models.
Items making up the parentingdimensions were confirmed as
hypothesized.The model tested showed a nonsignificant chi-square
value with multiple fit indexes above .90.Meeting these criteria
indicates a reasonablygood fit, as suggested by Tanaka (1987)
andMarsh, Balla, and McDonald (1988). The only
1 2 3 4 5 6 7
Figure 1. Measurement model of Chinese American parenting
dimensions. The psi (t/r) coefficients refer tointercorrelations
among the latent constructs.
-
426 KIM AND GE
Table 2Maximum-Likelihood Estimates for Measurement Model of
ParentingDimensions for Chinese Americans
Dimension and item
Inductive reasoning (t/rl)A U . Talk to childA2,l. Give reasons
for decisionsA3,l. Ask opinion about decisionsA4.1. Explain
rules
Monitoring (*//2)A5,2. Whereabouts of childA6,2. Whom child is
withA.7,2. In bed by set time
Harsh discipline (i/>3)A8,3. SpankA9,3. Lock out of
houseA10,3. Hit
Interrelationships among parentingdimensions
$2. 1. Monitoring, inductivereasoning
ij/3. 1. Harsh discipline, inductivereasoning
t//3. 2. Harsh discipline.monitoring
X*PdfGFICFIIFIn
Adolescent perception
Mother
STD
.65
.61
.70
.56
.59
.61
.48
.72
.65
.93
.61
- . 3 1
- . 6 6 .
28.440.65
320.931.001.02
82
UNSTD
1.901.762.511.32
0.601.201.05
0.220.160.21
Father
STD
.57
.80
.76
.82
.87
.84
.48
.68
.76
.82
-56,,
- . 2 0 "
- . 3 2
35.840.29
320.920.990.99
85
UNSTD
1.642.212.462.50b
2.25b2.28b1.44
0.200.160.24
Self-report
Mother
STD
.74
.83
.67
.69
.76
.61
.73
.66
.79
.79
.73
- . 6 0
- . 8 4
46.180.05
320.890.950.95
74
UNSTD
1.631.071.011.01
.781.041.06
0.180.210.21
Father
> STD
.63
.64
.61
.58
.88
.72
.74
.79
.61
.72
87C
- . 4 0
- . 4 9
36.900.25
320.910.970.98
75
UNSTD
1.251.191.370.89
1.080.811.18
0.200.090.13
Note. Maximum-likelihood estimates were significant at p <
.05 level except where indicated. Subscript aidentifies significant
differences in unstandardized maximum-likelihood estimates between
adolescent reportof mother and mother self-report; subscript b
identifies significant differences between adolescent report
offather and father serf-report; and subscript c identifies
significant differences between mother and fatherself-reports. STD
= standardized estimate; UNSTD = unstandardized estimate; GFI =
goodness-of-fit index;CFI = comparative fit index; IFI =
incremental fit index.a Marginally significant atp < .10.
exception involved the model of mothers' self-reported parenting
(goodness-of-fit index [GFI]= .89). This model, however, had
comparativefit index (CFI) and incremental fit index (IFI)values of
.95 with a nonsignificant chi-squarevalue, suggesting an acceptable
model fit.
As shown in Table 2, the analyses revealedthat the 10 items used
to assess parenting prac-tices all significantly loaded on the
three par-enting dimensions. The standardized lambdacoefficients
ranged from .48 (A72) to .93 (A103).The lowest loading for
adolescents' perceptionsof both mothers' and fathers' parenting
in-volved the item concerning how often parents
know their adolescent is in bed by the set time(A72). No
consistent pattern of low loadingsappeared for self-reported
parenting by mothersand fathers. Taken together, the resulting
sig-nificant factor loadings for these models suggestthat the 10
items developed for European Amer-icans were, in general, adequate
measures of thethree latent parenting constructs in
ChineseAmericans.
The standardized coefficient estimates of theintercorrelations
(i/r's) among the latent con-structs for the four models ranged
from - .20(p < .10) between fathers' harsh discipline
andinductive reasoning perceived by adolescents to
-
SPECIAL ISSUE: PARENTING AND DEPRESSIVE SYMPTOMS 427
.87 (p < .01) between fathers* self-reportedmonitoring and
inductive reasoning. The coef-ficients were all statistically
significant exceptfor the marginally significant relationship
be-tween fathers1 harsh discipline and inductivereasoning as
perceived by adolescents. It ap-peared that for adolescents, as
well as for par-ents, the harsher parents are as
disciplinarians,the less likely they are to monitor their
adoles-cent and use inductive reasoning in their par-enting
practices. The significant correlation co-efficients between
monitoring and inductivereasoning across the four models (ranging
from.56, p < .05, to .87, p < .05) suggest thatparents who
use inductive reasoning in theirparenting practices are more likely
to monitortheir adolescents' behaviors.
To examine how parents and their adoles-cents conceive parenting
practices, we con-ducted three sets of comparisons using
multi-group comparison procedures in structuralequation modeling
(Jbreskog & SOrbom, 1996).These analyses were performed to
determinewhether there were significant differencesamong (a)
adolescents' perceptions of mothers'parenting and self-reports by
mothers, (b) ado-lescents' perceptions of fathers' parenting
andself-reports by fathers, and (c) self-reported par-enting
practices by mothers and fathers. Thesecomparisons were made by
constraining themodels to be equivalent between the two groupsunder
comparison. We then relaxed the equiv-alence constraints to
determine whether the chi-square statistic changed significantly. A
signif-icant change in chi-square would provideevidence that there
was a significant differencein how the parenting items were
conceived be-tween the two groups.
The results from these comparisons suggestthat there was no
significant overall differencebetween adolescents and their
mothers, Ax^lO,N = 156) = 7.99, p > .05, or between
spouses,A^CIO, N = 149) = 10.12, p > .05. There was,however, a
significant overall difference be-tween adolescents and their
fathers, Aj^ClO,N = 160) = 35.05, p < .05. To investigatewhich
specific items were differentially con-ceived by adolescents and
their fathers, we fur-ther compared the factor loadings for the
10items. Three of the 10 items (explain rules,know whereabouts of
child, and know whomchild is with) were found to be
significantlydifferent between adolescents and their fathers
(see Table 2). Interestingly, 2 of them con-cerned the
monitoring construct (know where-abouts of the child and know whom
the child iswith). Taken together, the results suggest
that,overall, adolescents and their mothers con-ceived parenting
dimensions in a similar way.There also was no apparent difference
betweenfathers and mothers in their conceptions ofvarious parenting
dimensions. It appeared thatadolescents and fathers tended to
conceive par-enting dimensions somewhat differently, partic-ularly
in the area of monitoring.
We next examined whether relationshipsamong the three parenting
dimensions would bedifferent across multiple informants. Therewere
significant differences between adoles-cents and their parents as
well as betweenspouses: adolescents and mothers, A^2(3, N =156) =
8.49, p < .05; adolescents and fathers,A^(3, N = 160) = 25.35, p
< .05; andspouses, A*2^, N = 149) = 9.41, p < .05.Further
examination indicates that the differ-ence between adolescents and
mothers wasmainly attributable to the difference in the mag-nitude
of relationship between monitoring andharsh discipline, (^3i2)
wherein the magnitudeof the negative correlation was
significantlysmaller for adolescents' perceptions of
mothers'parenting. Another relationship, that betweenmonitoring and
inductive reasoning, ( t ^ ) wasstatistically different between
spouses and be-tween adolescents' perceptions of fathers'
par-enting and self-reports by fathers. These resultssuggest that
there were some differences be-tween informants in the magnitudes
of the rela-tionships among parenting dimensions, particu-larly in
comparison with fathers' self-reportedparenting.
Despite some differences in the parentingconceptualization, we
decided to treat the par-enting dimensions in the same manner for
allinformants for three reasons. First, we wantedour parenting
dimensions to be driven by theoryrather than based on the results
of the factorcoefficients. We therefore considered the facevalidity
of the questions for all informants. Sec-ond, treating the
parenting dimension in a dis-similar way across informants would
create aninconsistency across analyses, leading to diffi-culty in
interpretation of the results. If a differ-ence existed, one would
not be certain whetherit was due to a real difference or due to
adifference in the selection of parenting items
-
428 KIM AND GE
across informants. Finally, when we examinedthe standardized
factor loadings, we did not seedramatic differences in the factor
loadingsacross informants.
Structural Model Linking ParentingPractices to Adolescent
DepressiveSymptoms
We used structural equation modeling to testour hypotheses that
(a) depressive symptoms ofparents would be associated with their
parentingpractices, (b) parenting practices reported byparents
would be significantly associated withadolescents' evaluation of
parenting practices,and (c) adolescents' perception of
parentingpractices would be significantly associated withtheir
depressive symptoms. The resulting stan-dardized coefficients for
the tested model areprovided in Figure 2. Results involving
mothersare presented above those involving fathers. Inthis model,
parenting practices is a latent con-struct consisting of harsh
discipline, inductivereasoning, and monitoring. Because the
threeparenting dimensions were highly correlated(see Appendix),
clustering them together as alatent construct provided a meaningful
way tocircumvent the problems of multicollinearityand measurement
errors. Residuals (e) for thesame parenting dimensions were allowed
to cor-
relate between adolescents and parents becauseof the possible
method variance resulting fromthe same measures used by the
informants. Thecorrelated measurement errors are not enumer-ated
for figure clarity. The Appendix providesmeans, standard
deviations, and zero-order cor-relations for all variables examined
in themodel.
As shown in Figure 2, both mothers' andfathers' models yielded a
good fit to the data:mothers, ;^(16, N = $3) = 17.46, p = .36,GFI =
.95, CFI = .90, IFI = .99, and fathers,^ (16 , # = 82) = 29.14, p =
.02, GFI = .92,CFI = .90, IFI = .91. These fit indexes
areconsidered acceptable according to Tanaka's(1987) suggestion for
a minimum p value of .01.The tests of equivalence between the
models formothers and fathers showed no significant dif-ference
between them. Therefore, for compara-tive purposes, we retained the
fathers' modelunmodified.
In Figure 2, it can be seen that depressivesymptoms of parents
were negatively related toparenting practices that are
characterized by lowharshness and high inductive reasoning
andmonitoring for both mothers (/321 ~ .48, p