Introduction

Intergenerational cultural dissonance (ICD)—a clash between parents and children over cultural values—occurs so commonly among immigrant families that it is regarded as a normative experience (Lee et al. 2005; Sluzki 1979; Ying et al. 1999). A typical scenario finds immigrant parents adhering to their traditional cultural beliefs while their children endorse dominant Western values, resulting in a clash. This clash is likely to be more serious among families from non-Western cultures, such as Vietnamese and Cambodian families, who share few commonalities with mainstream U.S. culture (Ying and Chao 1996). Asian American adolescents, specifically Vietnamese, Cambodian, and Filipino youth, report higher levels of culture conflict with parents than other groups of adolescents (Phinney et al. 2000; Rumbaut 1996). However, we have limited knowledge about how such cultural conflicts influence youth and their families.

Several scholars have suggested connections between ICD (also labeled acculturation gap) and a variety of negative consequences for families and children, including troubled parent–child relationships, psychological difficulties among family members, and school problems among youth (e.g., Chung 2001; Kibria 1993; Lee et al. 2005; Min 1998; Nguyen and Williams 1989; Szapocznik et al. 1978). However, only a few studies have specifically examined the relationship between such dissonance and problem behaviors, or the mechanisms through which ICD contributes to youth problem behaviors (see e.g., Go and Le 2005). It is possible that this dissonance leads to problem behaviors either directly or indirectly by, for example, disrupting family processes, that is, increasing family conflict and weakening bonding between parent and child (McQueen et al. 2003). A more specific understanding of ICD and its connection to problem behaviors can lead to more effective interventions to prevent problem behaviors.

Vietnamese and Cambodian Immigrant Families

Studies of Asian Americans as a group are scarce (Choi and Lahey 2006; Harachi et al. 2001; Levesque 2007). Subgroup-specific studies are even scarcer (Choi, in press; Harachi et al. 2001). Among the few existing studies, Southeast Asians, including Vietnamese, Cambodians, Laotians, and Hmong, are often analyzed together, despite critical differences, such as premigration background, language, culture, religion, and the level of industrialization in their country of origin (e.g., Chung 2001). For example, Vietnamese culture is heavily influenced by China, while other Southeast Asian cultures such as Cambodian are more influenced by India (Bankston and Hidalgo 2007). Even though Buddhism is a traditional religion for both countries, Cambodians generally practice Hinayana Buddhism while the Vietnamese practice Mahayana Buddhism (Bankston and Hidalgo 2007; Kim 2002). The latter emphasizes compassion to and cohesion with others as a pathway to individual enlightenment, while Hinayana focuses more on achieving individual enlightenment and personal development as a path to helping others (Kim 2002).

Vietnamese and Cambodian youth have adapted differently to life in the U.S. Vietnamese youth, in general, are faring well academically and behaviorally, while Cambodian youth are doing less well (Choi, in press; Kim 2002; Portes and Rumbaut 2001). Cambodian youth demonstrate greater academic difficulties on average (e.g., high school dropout and truancy) and show signs of other maladjustment (e.g., involvement in ethnic gangs) (Goldberg 1999; Ima 1995; Kim 2002). Chung (2001) reports that cultural conflicts are greatest among the foreign-born (first generation) parents and U.S.-born children, a category into which a large proportion of Vietnamese and Cambodian families now falls. These two refugee/immigrant groups received some attention during their early years of settlement in the 1980s, particularly on the effects of war and refugee status and subsequent high rates of psychopathology and other adjustment problems among adults (e.g., Kinzie et al. 1984; Mollica et al. 1987). Yet little research has examined the adjustment of children whose parents may remain affected by their war-conflict experiences. Thus, despite the potential high vulnerability and significant group differences, our knowledge and understanding of family dynamics among Vietnamese and Cambodian immigrant and refugee families remain particularly limited (Dinh et al. 1994). Comparative studies can provide valuable insights into ethnic-specific and common factors that predict youth behaviors, including ICD and family processes (Chung 2001). An enhanced understanding of family dynamics and conflict and any mediating effects across Vietnamese and Cambodian families can shed light on the apparent differences and disparities between these two groups and suggest appropriate interventions.

Intergenerational Cultural Dissonance and Family Processes

Many scholars have speculated an indirect effect of ICD, in which ICD disrupts family processes and complicates the typical generation gap during adolescent years by increasing misunderstandings and miscommunications. The latter may eventually result in arguments and disagreements between parents and adolescent children (e.g., Lee et al. 2000; McQueen et al. 2003). However, studies are limited that explicitly investigate this mechanism. Using findings from Mexican American samples, McQueen et al. (2003) suggested that a high level of acculturation among youth increased deviant behaviors, mediated by high levels of conflict and lack of secure bonding between parent and child. However, their study did not test cultural conflict per se. Their underlying assumption was that the high level of acculturation among youth would likely produce cultural conflict between parents and children. The difficulty with this assumption is that, without specifically testing the discrepancy in cultural values, it is impossible to identify parents’ level of acculturation. Parents may in fact be as acculturated as their children and hence the high level of acculturation of youth will not necessarily lead to dissonance. Using the level of ICD reported by youth, we test the indirect and predictive role of ICD on problem behaviors by first hypothesizing that ICD leads to parent–child conflict (see Fig. 1).

Fig. 1
figure 1

Conceptual model

Full size image

Parent–Child Conflict and Bonding

A large body of literature documents parent–child conflict during childhood (characterized by frequent disagreements, fights, arguments, and anger) as a significant risk factor for a range of both externalizing youth problems (e.g., antisocial behaviors, aggression, and conduct problems) and internalizing problems (e.g., depression, anxiety, and low self-esteem) (Barber and Delfabbro 2000; Formoso et al. 2000; Hawkins et al. 1992; Hill et al. 2005; Ingoldsby et al. 2006). The detrimental impact of parent–child conflict during childhood may go beyond the adolescent years, for example, influencing social and psychological adjustment during young adulthood (Hannum and Dvorak 2004). The impact has also been found to be significant across genders (Harachi et al. 2006b) and racially and ethnically diverse samples (Constantine 2006).

On the other hand, a strong and positive bonding between parent and child can serve as a protective factor against adolescent problems (Arbona and Power 2003; Dryfoos 1998). The effect of secure parent–child bonding has been confirmed for a variety of youth outcomes and among diverse samples, such as problem behaviors among Latino adolescents (Eamon and Mulder 2005) and among a group of African American, European American, and Mexican American adolescents (Arbona and Power 2003), and on a psychological and social adjustment indicator for college students (Hannum and Dvorak 2004). Throughout adolescence, parents continue to influence their children (Arbona and Power 2003; Catalano and Hawkins 1996). Although changes in parent–child relationships, including the nature of parental attachment, are likely inevitable during adolescence, the healthy process of maturation has generally been characterized by continuous and close relationships with parents while also developing autonomy (McQueen et al. 2003). In this study, it is hypothesized that parent–child conflict positively predicts problem behaviors while positive and strong parent–child bonding negatively predicts problem behaviors (see Fig. 1).

Findings are mixed on the specific mechanisms by which parent–child conflict leads to youth problems. Although several studies have found a direct association between parent–child conflict and problems among young and school-aged children (e.g., Ingoldsby et al. 2006), others have noted that the effect of parent–child conflict is consistent but not large (Formoso et al. 2000) and, when considered simultaneously with other family processes such as bonding with parents, the effect may no longer be significant (e.g., Eamon and Mulder 2005; Fleming et al. 2002; Hill et al. 2005). The effect of parent–child conflict may be indirect, that is, mediated by positive parent–child bonding. For example, Hannum and Dvorak (2004) have found that parent–child bonding fully mediates the link between parent–child conflict and problem behaviors. In other words, parent–child conflict reduces the level of secure parent–child bonding, which in turn is related to social and psychological adjustments.

At the same time, studies have shown that parent–child conflict and bonding significantly and independently influence youth behaviors. Specifically, they both influence substance use and deviant behaviors among Mexican American youth (McQueen et al. 2003) and depression and conduct problems among an ethnically diverse sample of youth (Formoso et al. 2000). Constantine (2006) further showed that parent–child conflict had both a direct and indirect effect on youth behaviors, with positive parent–child bonding partially and significantly mediating the effect of perceived conflicts. A clearer understanding of these relationships can help develop specific and effective interventions.

In this study, parent–child conflict is hypothesized to have both a direct and indirect effect on problem behaviors (Fig. 1). Parent–child conflict is fairly common among families with adolescents (Phinney et al. 2000), with a certain level of conflict regarded as a normal and necessary process of establishing independence during adolescence (Steinberg 1991). Many adolescents who experience conflict with parents do not have serious adjustment problems (Formoso et al. 2000). It may be because a mutual bonding can be maintained while tolerating the conflict that might occur while youth increasingly assert their autonomy (Constantine 2006). Parent–child bonding, in fact, can gain more salience, especially in the presence of familial conflict (Constantine 2006). A strong and positive parent–child bond can assure youth that parents will be a continued source of stability and support, despite conflicts or disrupted family relationships (Formoso et al. 2000). Although several studies (mentioned above) have demonstrated quite strong evidence of a direct effect of parent–child conflict on youth behaviors, findings on the mediating effect of parent–child bonding have been somewhat inconsistent, and relatively little research has been devoted to explicitly testing the mediated relationship (Formoso et al. 2000). This study adds to the literature by testing the direct and mediated relationships among parent–child conflict and bonding. The study also contributes to the literature by testing these relationships with Vietnamese and Cambodian immigrant families.

Present Study and Hypotheses

This study empirically tests the relationships shown in Fig. 1 by ethnic group. The hypotheses are summarized as follows: (1) ICD will be associated with increased parent–child conflict as reported by both youth and parents; (2) parent–child conflict, in turn, will directly and positively predict problem behaviors, and positive parent–child bonding will negatively predict problem behaviors; and (3) the effects of conflict are anticipated to be both direct and mediated by positive parent–child bonding. Because it is important to include both parents’ and youths’ perceptions of family conflict (Lee et al. 2005), the study uses both maternal and youth reports of conflict.

It is impossible to develop specific a priori hypotheses regarding group differences, owing to a lack of current research on these two groups. Studies have documented different demographic characteristics or general cultural differences across these two groups (e.g., Waters et al. 2007), but none has specifically related family dynamics to youth outcomes. Thus, the study simply tests the null hypotheses, which assume the relationships will not be significantly different across groups.

This study uses longitudinal data from the Cross Cultural Families (CCF) Project. Unlike many of the studies on Southeast Asian immigrants and refugees, the CCF Project, although regional, used community-based random samples of the populations. Clinical or nonrandom convenience samples not only limit the generalizability of findings to the general population, but also bias the findings owing to sample-specific characteristics. Findings from CCF Project data can better inform intervention targets for the general population.

As the number of immigrant families continues to rise—by 2010, children of immigrants alone will account for 22% of the school-age population (Portes and Rumbaut 2001)—it is imperative that those who work with youth and their families understand the difficulties that immigrant families face. Moreover, the majority of contemporary immigrants, including Vietnamese and Cambodian families, are from cultures that emphasize less independence among youth and more interdependence, obedience, and family obligations (Zhou and Lee 2004). Thus, for these immigrant parents, the reach for autonomy during adolescence can be unanticipated and puzzling (Ying et al. 1999). It also comes with added challenges such as cultural conflict. Yet, children whose cultures emphasize family loyalty and preservation of strong emotional ties among family members may be more likely to benefit from the buffering effects of positive bonding when faced with high levels of cultural dissonance and parent–child conflict (Formoso et al. 2000). Thus, if cultural conflicts within immigrant families indeed affect problem behaviors, it is critical to illuminate the pathways by which ICD influences problem behaviors to identify the source of risk as well as the buffering factors that mitigate the impact of conflicts.

Methods

Overview of Study and Sample Selection

The CCF Project (P.I. Tracy W. Harachi) is a longitudinal study examining the developmental trajectories of a panel of Cambodian (n = 164) and Vietnamese (n = 163) youth, including the occurrence of problem behaviors and the relations between risk and protective factors and different outcome behaviors. Potential respondents were identified through locator information obtained from an urban Pacific Northwest school district. In 2001, the project contacted a random sample of Vietnamese and Cambodian families with a child enrolled in Grades 3–6 in the school district. The overall consent rate was approximately 85%. Data collected for the longitudinal study include annual maternal interviews, youth interviews, teachers’ report of student behaviors, and school and police records. We used maternal and youth data collected in 2004 (wave 4) and 2005 (wave 5) for our analysis.

Sample Description

The sample includes 327 mothers and children respectively. At the first interview (in 2001), the average age of the Cambodian mothers was 41.8 years (SD = 8.37), and 42.3 years (SD = 6.87) for Vietnamese mothers. On average, the Cambodian mothers had lived in the U.S. for 16 years (SD = 4.09), and Vietnamese mothers had lived in the country for 11 years (SD = 5.15). A total of 72% of mothers in both groups reported receiving public assistance, food stamps, or qualifying for the free/reduced price school lunch program. Slightly more than one-fourth (27%) of Vietnamese mothers and one-half of Cambodian mothers were single parents. Parental education levels were low. Among Cambodian mothers, 80.1% had less than a high school degree, while 65.8% of the Vietnamese parents lacked a high school degree. Only 4.1% of the Cambodians and 16.8% of the Vietnamese had some college or higher education, either in their country of origin or in the U.S. The majority of Cambodian mothers (88.6%), and one-half of Vietnamese mothers (47.5%) were Buddhist; 37% of Vietnamese mothers were Catholic. The two groups differed significantly by age (F 158, 158 = .6743, p < .01), single parenthood (χ 2 = 16.291, p < .001), education (completed in the native country [t 217 = −3.72, p < .001] but not in the U.S. [t 235 = −.7729, p = .440]), and year of arrival to the U.S. (F 158, 155 = 1.583, p < .01).

About 49% of the child sample were male (79 Vietnamese, 81 Cambodian) and 51% were female (84 Vietnamese, 83 Cambodian). In 2004 (wave 4), the average age of the children in both groups was 15.2 years (SD = 1.05 in both groups). Among Cambodian children, 88% were born in the U.S., in contrast to 34% of Vietnamese children. Of those not born in the U.S., the average age of arrival in the U.S. was 4.6 for Cambodian and 4.3 for Vietnamese children (SD = 3.9 and 2.4, respectively). Hence, the bulk of this subsample’s schooling and socialization occurred in the U.S., and these youngsters would likely be more similar to U.S.-born, second-generation youth than first-generation newcomers.

Measures

The model has five constructs. The four predictor variables (ICD, reports of conflicts by parent and child, and parent–child bonding) are drawn from wave 4 and the outcome variable (youth problem behaviors) from wave 5 to establish a temporal order. The survey questionnaires were translated to Vietnamese and Khmer and back-translated and tested for validity. For example, the measurement equivalence of several major constructs was examined by testing several forms of invariance, including conceptual, functional, item and scalar equivalence. Results are reported elsewhere (e.g., Choi and Harachi 2002; Choi et al. 2006; Harachi et al. 2006a). In summary, findings from these studies suggest many equivalent items across groups and even when two groups (Vietnamese and Cambodians) or three groups (Vietnamese, Cambodian, and European Americans) endorse items differently, examining correlations and relationships (but not mean scores) would be appropriate.

In testing a path or structural model, it is ideal to follow a two-step estimation process (Anderson and Gerbing 1988) to first estimate the fit of the measurement model while creating latent constructs (Shumow and Lomax 2002), and then to test the fit of the structural model. However, creating latent constructs requires much larger sample sizes than the present data can provide. Thus, we created “measured constructs” rather than latent constructs. In other words, we defined the constructs as an unweighted sum of the response scores of the individual items. We normalized each scale with the minimum of zero and the maximum of 10 to make them comparable by using the linear transformation of values; within each construct, we coded the items such that high scores reflect a high value of the construct. A limitation of using measured constructs is an inability to account for measurement errors in estimating the fit of the structural model. To address this limitation, prior to fitting the proposed structural path model, we applied a series of exploratory statistical analyses at the item and scale levels to examine the reliability of measurements and to detect patterns in the data that may not be explicit in a path analysis fit. We describe some of these findings when necessary in the text. Table 1 shows the correlations among constructs and descriptive statistics of each construct in this study.

Table 1 Correlations among constructs
Full size table

Intergenerational Cultural Dissonance (wave 4) was assessed by 10 items adopted from the Asian American Family Conflict Scale—Likelihood (Lee et al. 2000). Youth were first asked “How likely is this type of situation to occur in your family?” followed by, for example, “Your parents tell you that a social life is not important at this age, but you think it is,” and “Your parents don’t want to bring shame upon the family, but you feel that your parents are too concerned with saving face.” Response options ranged from (1) “never” to (5) “always.” The reliability coefficient was .86. There was no significant difference in the rate of ICD between Vietnamese and Cambodian youth (mean 4.56 vs. 4.88, SD 2.14 vs. 2.02, χ 2 = 35.30, p = .549).

Parent’s Report of Conflict (wave 4) consisted of eight items adopted from the Conflict Behavior Scale (CBQ, Prinz et al. 1979), in which mothers reported the level of conflict with the child. Example questions include: “My child and I almost never seem to agree” and “At least three times a week, we get angry at each other.” The response options were (0) “false” and (1) “true.” We summed the items to indicate the construct (range from 0 to 8), and the reliability coefficient was .83. We found significant differences in the rate of conflict across Vietnamese and Cambodian families (mean 2.63 vs. 2.87, SD 3.15 vs. 3.00, χ 2 = 17.11, p < .05); Cambodian mothers reported a higher rate of conflict than Vietnamese mothers.

Child’s Report of Conflict (wave 4) was assessed by four items, also adopted from the CBQ. Examples include “My parents never listen to my side of the story,” and “My parents nag me a lot.” The response options were (0) “no” and (1) “yes.” The reliability coefficient was .64. We found no significant difference in the rate of youth-reported conflict between Vietnamese and Cambodian families (mean .72 vs. 1.51, SD 1.72 vs. 2.40, χ 2 = 6.34, p = .175).

We did find a discrepancy between parental and youth reports of family conflict. For example, 16.8% of the Vietnamese and 32.3% of the Cambodian youth responded that they and their parents get angry with each other at least three times a week, while their mothers denied such frequency based on the same item. Conversely, 17.6% of the Cambodian mothers and 8% of the Vietnamese mothers reported arguing a lot about rules, while their children reported no such conflict. Consequently, the correlation between parent and youth reported conflict was low (r = .12 for Vietnamese and r = −.18 for Cambodians).

Bonding with Parents (wave 4) was assessed by 14 youth-reported items, seven of which pertained to their relationship with their mother, and another seven to relationships with their father (Armsden and Greenberg 1987; Hawkins and Catalano 1990). We used the sum of seven items for single-parent families, and we used the average of summed scores of bonding with both parents (14 items) for two-parent families. Questions include “Are you close to your mother (or father)?” and “I wish I had different mother/father (reverse-coded).” Response options were (1) “NO!” (2) “no” (3) “yes” and (4) “YES!” for six items, and (1) “never,” to (5) “always” for eight items. Using the same method as described above, we recoded the response options to make them comparable. The reliability coefficient was .88 for the scale. We identified no group differences between Vietnamese and Cambodian families in the rate of bonding with parents (mean 6.18 vs. 5.56, SD 2.05 vs. 2.01, χ 2 = 135.01 p = .484).

Youth Problem Behaviors (wave 5) was measured with 15 items, which included questions on the frequency in the year prior to the interview of being arrested, trespassing, drawing graffiti, shoplifting, setting fire, carrying a weapon to school, hurting others, initiating a fight, running away from home, going out late without parental permission, damaging other’s property just for fun, and hitting parents. The scale also included items that asked whether respondents tease, lie (responses ranged on a scale of 1–4, from emphatic NO! to emphatic YES!), or belong to a gang (no (1)/yes (2)). The response options to the frequency questions were (1) “0 times,” (2) “1–2 times,” (3) “3–4 times,” and (4) “5 or more times.” We adopted the items from several sources, including the Seattle Social Development Project (Hawkins and Catalano 1990) and Seattle Personality Questionnaires (Kusche et al. 1988) to cover a range of problem behaviors.

Because only a few youth reported committing a behavior more than one or two times, we first dichotomized each item so that responses were either 0 for no incidence of the behavior or 1 for any incidence of the behavior, and then we summed the items (range from 0 to 15). Among the youth, 114 (45.7%) reported no problem behaviors, and most of youth reported one (n = 66, 20.9%), two (n = 36, 11.4%), three (n = 30, 9.5%) or four (n = 23, 7.3%) problem behaviors. The most common problem behaviors in the sample were teasing, lying, fighting, shoplifting, and having gone out late. The alpha reliability was .67. In addition, Cambodian youth reported significantly higher counts of problem behaviors than Vietnamese youth (Cambodian youth mean 1.76, SD 1.39 vs. Vietnamese youth mean .93, SD 1.97, χ 2 = 22.84, p < .01). Among respondents who reported more than four problem behaviors, only two respondents were Vietnamese while 14 were Cambodian. Because nearly one-half of the respondents reported having no problem behaviors, we further dichotomized the scale into binary variable, 0 for no problem behaviors and 1 any problem behaviors. A total of 43% of Vietnamese youth and 65% of Cambodian youth reported instances of any problem behavior. We conducted subsequent analyses with the dichotomized outcome. In addition, we adjusted the youth behaviors for gender and age in the model owing to the significant differences in the outcome variable by these control variables.

Analysis Strategy

We examined the relationships between ICD and parent–child conflict and whether parent–child conflict had a direct effect on youth problem behavior. We also examined whether parent–child bonding mediated the effects of family conflict on youth problem behaviors (see Fig. 1). To estimate the proposed model, we conducted path analyses with weighted least squares mean and variance-adjusted χ 2-test statistic (WLSMV) with robust standard errors estimations. WLSMV estimation is appropriate for binary dependent variables, and we used path analysis to test the mediational hypotheses (Baron and Kenny 1986; Hoyle and Smith 1994). The latter is also a useful strategy for simultaneously investigating the combined and the unique effects of each variable (Ramirez-Valles et al. 1998). We relied on MPlus 4.1 for analysis (Muthén and Muthén 2005).

In testing the path model, we examined the fit of the model in each group first and used multiple-group methods to determine whether the hypothesized relationships among the model variables were equivalent across Vietnamese and Cambodian families (Byrne 1994). Multiple-group path analyses compared two nested models (i.e., unconstrained and constrained models) (Byrne 1994). The unconstrained model allows all model parameters to be estimated freely for each group. In the constrained path model, cross-group equality constraints are placed on all hypothesized paths between constructs (Hancock et al. 1999). These constraints enable us to identify any statistically significant differences in the magnitudes of path parameters across groups.

We assessed the fit of all models by examining model chi-square (χ2), the Comparative Fit Indices (CFI, Bentler 1990), and the Root Mean Squared Error Approximation (RMSEA, Browne and Cudeck 1993). A good fit is indicated by CFI values of greater than .90 (Byrne 1994). Values of less than .05 for the RMSEA are considered evidence of a good fit, values between .05 and .08 indicate a fair fit, and values greater than .10 represent a poor fit (Browne and Cudeck 1993). We examined the statistical significance of the estimated parameters with z statistics and a .05 level of statistical significance.

We used several measures to evaluate the statistical significance of the difference between the unconstrained and the constrained models, that is, any statistically significant group difference in path models between Vietnamese and Cambodian families. The change in χ2 relative to the change in degrees of freedom indicates whether the constrained model has statistically significantly poorer fit than the unconstrained model, which is also an indication of statistically significant group difference (Byrne 1994). The DIFFTEST on the constrained models further shows which equality constraints contribute most to degradation in model fit (Muthén and Muthén 2005). We also used Modification Indices (MI) function to investigate whether data suggest adding any particular path in the model. We used MI as a purely exploratory step to inform future studies.

About 70% of the data were complete for both groups. Nearly 20% of the Vietnamese and 15% of Cambodian samples had missing data on just one variable. Only 4% of the Vietnamese and 10% of the Cambodian samples had missing data on three or more variables. The analyses were conducted both with and without missing data imputations. Mplus uses maximum likelihood (ML) to estimate the models with missing data. The results from analyses with and without missing data imputations were similar; hence, we report the results using the missing data imputations.

Results

Path Analysis in Each Group

We first estimated the path model with each group. The results (Table 2 and Fig. 2) indicate that the model using the WLSMV estimator fits the data well. Fit indices for the Vietnamese families were: χ 2(7) = 14.93, p = .003, CFI = .93, RMSEA = .009. For Cambodian families, the fit was: χ 2(7) = 15.24, p = .004, CFI = .91, RMSEA = .088. The model accounted for 21% of the total variance in the outcome variable for Vietnamese families and 28% for Cambodian families.

Table 2 Model fits and path statistics for each group
Full size table
Fig. 2
figure 2

The model with path coefficients

Full size image

The statistically significant paths found in the model were similar to the results of the pair-wise correlations, with the exception of the relationship among Cambodians between parent-reported conflict and problem behavior, which was not significant in pair-wise correlations but was significant in the full model. Results showed statistically significant paths with p values of .05 or less for both groups between ICD → Youth Report of Conflict (β = .681 for Vietnamese, and .560 for Cambodians) and Youth Report of Conflict → Parent–Child Bonding (β = −.700 for Vietnamese, and −.516 for Cambodians). The path between Parent Report of Conflict → Problem Behavior (β = .246) was significant only among Cambodians. Figure 2 shows the standardized path coefficients of the model for each group. Table 2 presents the path coefficients and corresponding statistics for the theoretical model organized by ethnic group.

Indirect Effects

MPlus provides information about the combined and unique, as well as direct and indirect, effects of the modeled constructs on the outcome variable. ICD had significant indirect effects on problem behaviors (p ≤ .05) in both groups (β = .240 for Vietnamese, and .173 for Cambodian families) but no direct effect. This means that the effect of ICD is fully mediated by other variables, as posited in the model. For the Vietnamese subgroup, youth report of conflict (β = −.700) was significantly associated with its immediate dependent variable (parent–child bonding), but it did not explain any variance in the outcome variable.

Group Differences in Paths

The unconstrained path analyses using the WLSMV estimator fit the data well (χ 2 [14] = 30.170, p = .007; CFI of .921; RMSEA of .087). The constrained model also showed a fair/good fit (χ 2 [22] = 39.012, p = .001; CFI of .917; RMSEA of .071). The difference between the two models showed very little improvement over the unconstrained model (Δχ 2 [8] = 8.842, .500 < p < .250, ΔCFI = .004). Thus, the conceptual model overall did not differ significantly by ethnic group.

Model Testing with Full Sample

Because there was no significant group difference in the overall fit of the conceptual model, as a next step we combined the groups to test the model with the full sample. The results are presented in Table 3 and Fig. 2. The model fit the data fairly well with the full sample, (χ 2 [7] = 27.710, p = .0002; CFI of .907; RMSEA of .098). The model accounted for 26% of the total variance in the outcome variable.

Table 3 Model fits and path statistics with combined samples
Full size table

The statistically significant paths were ICD → Youth Report of Conflict (β = .628, p < .001), and Youth Report of Conflict → Parent–Child Bonding (β = −.617, p < .001), Parent–Child Bonding → Problem Behavior (β = −.246, p < .01), and Parent Report of Conflicts → Problem Behavior (β = .242, p < .01). Similar to the results from the subgroup analyses, ICD had significant indirect effects on problem behaviors (β = .228, p ≤ .05) but no direct effect with the full sample.

The MI provides information regarding paths that may be added to the estimated model that result in a reduction of the χ 2 value of the model fit. Paths that would significantly improve the fit of the model would have a chi-square value of 3.84 (p = .05 for df = 1) or greater. The only path that satisfies this criterion was the path between Parent–Child Bonding → Youth Report Conflict.

Discussion

Significant life-changing events such as migration are likely to exact physical and emotional tolls on the family (Chung 2001; Sluzki 1979). Immigrant and refugee families face challenges to reestablish family roles and patterns in an unfamiliar society with a new language and sociocultural environment. Although many immigrant families adjust well and make a positive transition (Chan and Leong 1994), many other immigrant parents lose their authority and confidence as parents in the process of renegotiating and adjusting (Foner and Kasinitz 2007; Kibria 1993). This challenging adjustment process may result in ICD and subsequent family conflict. Parent–child intergenerational cultural conflicts are the most frequent problem among Asian American youth and young adults who seek counseling (Lee et al. 2005), and they are found among all Asian American subgroups (Ying et al. 1999). The high level of psychopathology among refugee parents may further confound family relationships and exacerbate cultural dissonance (Go and Le 2005; Ying and Chao 1996).

To better understand this common difficulty among immigrant/refugee families, this study examined whether ICD was associated with parent–child conflict, and whether such conflict led directly to youth problem behaviors or indirectly to such behaviors by weakening parent–child bonding. The results show that the proposed conceptual model is not significantly different across Vietnamese and Cambodian families, indicating that the mechanisms by which ICD leads to problem behaviors are similar across these two groups. Thus, the model was tested again with the combined samples. The following discussion focuses on the results from the full sample unless otherwise noted.

The findings provide insights into the pathways through which ICD may lead to problem behaviors. As hypothesized, ICD indirectly increases problem behaviors by disrupting family processes, that is, by significantly heightening parent–child conflict, but only conflict that is perceived by youth (but not by parent). In other words, if youth perceive gaps in cultural values with their parents, they report higher rates of disagreements and arguments with their parents. These conflicts in turn significantly weaken the sense of parent–child bonding reported by youth. Thus, if both cultural discrepancy and parent–child conflict are present, the affective relationship with parents is likely to be compromised, which predicts an increase in the level of problem behaviors among youth. Parents’ reports of conflict (but not youth reports), on the other hand, are a significant predictor of problem behaviors. Given that parent reports of conflict are not highly correlated with youth reports in the combined sample or in either group—in fact, these two reports are incongruent—it is possible that parent- and youth-reported conflict may differently and uniquely predict different constructs, or they may be assessing different constructs. Alternatively, this may be a result of a reporting artifact. If we had a parent-reported parent–child bonding measure, parent-reported conflict might have been mediated by that construct.

More critically, the findings of this study confirm the buffering effects of youth-reported parent–child bonding against youth-reported conflict. In the presence of ICD and parent–child conflict among Vietnamese and Cambodian youth, a secure sense of bonding can help mitigate the deleterious effect of conflict on problem behaviors.

The Modification Indices suggested adding a path from Parent–Child Bonding to Youth Report Conflict. The path from Youth Report Conflict to Parent–Child Bonding as proposed in the model was significant both in subgroup analyses and in full sample analyses. The findings from the MI may indicate that these two constructs are highly correlated (as shown in pair-wise correlations) and they may bidirectionally influence each other. To specifically test the directionality of the relationship, it would be necessary to examine these two constructs at different time points. However, the data did not provide these constructs at other time points; therefore, this question should be investigated in the future studies.

This study aimed to identify ethnic-specific and common predictors of problem behavior. The results show common predictors and mechanisms of those predictors. Although the commonality across the two groups is helpful in informing interventions, it is critical to identify group-specific factors as well, especially because Cambodian youth report a significantly higher rate of problem behaviors than Vietnamese youth in this sample, which is consistent with other studies (e.g., Kim 2002). The higher rate of problem behaviors among Cambodian youth does not stem from ICD because the mean of ICD was not significantly different across the two groups. This suggests that other reasons must be further explored to explain the high vulnerability of Cambodian youth.

In Lee and his colleagues’ focus groups (2000), many children of immigrants reported that ICD with their parents was not limited to adolescence but was a natural part of life beyond adolescence, and may not be necessarily harmful to the well-being of their family. However, the results from this study show that if families fail to resolve these ICDs, these gaps can result in a high level of parent–child conflict, which, in turn, may weaken secure bonding with parents and increase problem behaviors. Immigrant parents faced with new environments tend to attempt to preserve their culture regardless of its fit with their new surroundings. Time spent in the U.S. does not appear to alter parents’ core familial values (e.g., Chung 2001; Kibria 1993; Lee et al. 2005; Min 1998; Nguyen and Williams 1989). In addition, parents who perceive the mainstream culture as a threat may idealize their culture of origin and attempt to rigorously reinforce it with their children, which can increase tensions and family conflict (Chung 2001; Foner and Kasinitz 2007). Additional study of family dynamics among immigrants and refugees can shed light on how to prevent ICD and related difficulties among Vietnamese and Cambodian families. It is also important to identify the predictors of ICD and to target those factors as a way to lessen ICD among immigrant families.

The study is not without limitations. Although randomly selected and more representative than many existing studies, the CCF data are regional. Thus, the generalizability to the national population of Vietnamese and Cambodian youth is limited. The samples are also largely poor and undereducated. However, these characteristics are generally reflective of the Vietnamese and Cambodian immigrant groups in the U.S. The study did not account for measurement errors in testing the structural model, and it is hoped that future studies with larger sample sizes can test the model to provide further support for the findings. The CCF data are based on youth and parent self-reports. However, such reports are generally found to be reliable for many variables (Johnston et al. 2001). Lastly, larger sample sizes would have been ideal to examine subgroup differences in the model.

Despite these limitations, this study adds to the literature by providing information about how ICD contributes to problem behaviors among Vietnamese and Cambodian families. Although additional studies are needed to continue to contribute to the field, the findings from this study can inform interventions that aim to prevent or treat problem behaviors among Vietnamese and Cambodian youth. In sum, the findings identify common predictors and family processes, and suggest that interventions targeting youth perceptions of intergenerational cultural gaps, conflict management with families, and improved bonding with parents may help prevent problem behaviors in both Cambodian and Vietnamese families and potentially other Asian immigrant families.