Abstract
Using life role salience theory, we investigated the extent to which occupational, parental, marital, and home care roles explained mental health outcomes among female expatriate spouses. Participants (N = 86) were from English-speaking Northern American or Western European countries; the average age was 38. Results of a two-way within-subject analysis of variance suggested that the occupational role was least important. In a comparison of role expectations, participants had greater value for (than commitment to) their parental role; the reverse was true for the marital role. Results of a path analysis explained 10% of the variance in psychological distress via a direct path from career role importance (β = .14) and an indirect path from the marital role importance through marital satisfaction (β = −.12). Three paths (parental role importance, β = .14; home care role importance, β = .15; indirectly from marital role importance through marital satisfaction, β = .18) explained 24% of the variance in psychological well-being.
Keywords
Among the factors contributing to expatriate adaptation, the adjustment of the expatriate’s spouse stands out as the most salient (e.g., Bhaskar-Shrinivas, Harrison, Shaffer, & Luk, 2005; Hechanova, Beehr, & Christiansen, 2003). Unfortunately, spousal adjustment has been primarily investigated as an antecedent of expatriate adjustment and as a resource for the expatriate employee (e.g., Lazarova, Westman, & Shaffer, 2010); few attempts have been made to understand the variables that contribute to the successful adaptation of the spouse. Those that have explored expatriate spouse transition found difficulties to be associated with loss of employment/career (e.g., Black & Stephens, 1989), changes in life roles (Mohr & Klein, 2004), and redefinitions of self-worth and identity (Kupka, Everett, & Cathro, 2008). We utilized life role salience theory as a framework for investigating the well-being in the female expatriate spouse.
Life Role Salience Theory
Super’s life span, life space theory (for a review see Super, Savickas, & Super, 1996) postulated that people live in multiple-role environments where work, family, educational, and community roles change in intensity and importance throughout the life cycle. In articulating future directions for Super’s approach, Herr (1997) identified a number of workforce trends such as globalization that were ripe for investigation within the life-span, life-space frame.
Our project investigated mental health outcomes (i.e., psychological well-being [PWB] and distress) in expatriate spouses as a function of differing dimensions of life role salience for four life roles (i.e., occupational, parental, marital, and home care). In the creation of the Life Role Salience scales (LRSS), Amatea, Cross, Clark, and Bobby (1986) developed the constructs used in our study. For each of the four life roles assessed, participation is defined along two dimensions. The first is role reward value: the degree to which that role is an important part of self-definition and personal satisfaction. The second is role commitment level: the extent to which a person is willing to commit personal time and energy resources to the role. Value and commitment dimensions likely vary across individuals. For example, an expatriate spouse who left a career defining position might have strong value for the occupational role, but have low commitment if she is unable to locate similar employment in her host country. Role importance (Campbell & Campbell, 1995) is the combination of the value and commitment dimensions. With regard to these predictors, the purpose of our study was to (a) investigate the degree of difference in value and commitment dimensions of role salience and (b) test the effect of role importance on mental health outcomes in a sample of female expatriate spouses.
Role-Related Research Regarding the Female Expatriate Spouse
Three sets of researchers (Bikos et al., 2007a, 2007b; Mohr & Klein, 2004; Shaffer & Harrison, 2001) have used mixed-method designs to investigate the adaptation of the expatriate spouse; these studies were a rich resource for our project. Shaffer and Harrison examined spousal adjustment through in-depth interviews with 10 female expatriate spouses, and then with the questionnaire developed by the authors, surveying 221 couples in 37 countries. Similarly, Mohr and Klein utilized a two-stage approach, first conducting individual and group interviews with expatriate spouses of both genders and then employing a questionnaire to survey 43 American expatriate spouses in Germany. Bikos et al. (2007a, 2007b) followed 32 American female spouses residing in Turkey, during their first year of expatriation. Interviews and measures of role salience and well-being were administered to all participants within 2 weeks after arriving in country and then at 3-month intervals through the first year. Next, we review findings from these studies in light of the occupational, parental, marital, and home care roles.
Attention to the occupational role has largely been around change in role and role adjustment, in relation to employment transitions. For example, Shaffer and Harrison (2001) reported that qualitative participants who experienced a change in employment status also struggled to transport their self-definitions to the new environment. Curiously, the results of their quantitative investigation indicated no relationship between self-reported change in employment status and personal adjustment of the female expatriate spouse. In contrast, there was a significant negative correlation between the spouse-reported change in employment status of the expatriate and her adjustment. Moreover, change in employment status did not account for any unique proportion of variance in more complex regression equations predicting expatriate adjustment. Similarly, Mohr and Klein (2004) reported that participants in the qualitative portion of the study were frustrated because they left careers, returning to the role of a housewife and becoming economically dependent. Curiously, the results of their subsequent quantitative investigation were counter to the themes voiced in the qualitative interviews. Specifically, they found a positive correlation between the role change scale and general living adjustment (defined as adapting to environmental aspects of the international transition such as differences in housing, shopping, and transportation) and between the role change scale and interpersonal interactions with members of the host culture. Finally, Bikos et al. (2007a, 2007b) reported that female spouses who strongly valued their career roles but could not perform them (i.e., were unable to commit to them) experienced psychological distress (DIS).
The parental role has generally had a positive effect on spouse adaptation (Shaffer & Harrison, 2001). The presence of children provides stability in social expectations and serves as a venue for forming relations with others (Mohr & Klein, 2004). Participants in the Bikos et al. (2007b) study reported increased and somewhat stressful parental demands at the beginning of the expatriation (e.g., locating safe play areas and age-appropriate playmates) but reported their decline as the family settled into the new community.
Less is known about the marital role. In the study by Bikos et al. (2007b), some participants reflected on seemingly normal marital difficulties that were exaggerated by the international adjustment. In contrast, Shaffer and Harrison (2001) reported that the adjustment of the expatriate employee was the strongest predictor of the spousal adjustment.
Increased responsibilities in the home care role were viewed as a result of employment loss and were often perceived as a negative consequence (Mohr & Klein, 2004; Shaffer & Harrison, 2001). Bikos et al. (2007b) reported that early in the expatriate transition, women were eager to manage domestic responsibilities (e.g., arranging for lodging, utilities, banking) but were unable to do so because of legal and contractual limitations. After settling in, while some women experienced a decrease in daily chores, some experienced an increase in hospitality-related responsibilities because of the husband’s employment. These new responsibilities left the women wondering about their ability to succeed in this new role and the potential negative impact that inadequate performance might have on their husband’s career.
A quantitative comparison of the four life roles (Bikos et al., 2007a) revealed no change over time in the importance of each of the four life roles during female participants’ first year of expatriation to Turkey (e.g., the parental role was consistently, highly important). There were differences between roles. The parental role was the most important (and statistically significantly more important than the marital role) and the occupational role was the least important (and statistically significantly less important than the remaining three). In addition, participants had greater value for than commitment to their occupation and greater commitment to than value for their marriage. A limitation of the Bikos et al. (2007a) study was that no analyses tied aspects of life roles to indicators of adaptation.
Study Purpose and Analytic Approach
In light of these findings, we investigated the extent to which occupational, parental, marital, and home care roles explained mental health outcomes among female expatriate spouses living in Ankara, Turkey. We analyzed our data in two phases. In the first phase, we hypothesized statistically significant differences between role categories (occupational, parental, marital, and home care) and role dimensions (value and commitment). In the second phase, we used path modeling to evaluate the degree to which role importance of the four life roles predicted DIS and PWB (Figure 1). Based on the foregoing literature, we predicted that those who placed greater importance on the occupational role (e.g., the women who had strong value for and commitment to employment but were unemployed or underemployed) would experience increased DIS and decreased PWB. In contrast, we predicted that those who placed greater importance in each of the family-related roles (e.g., mothers who valued the parental role and were actively engaged in it during the expatriation; wives who valued the marriage and were committed to maintaining it; homemakers who now had time to commit to home care activities) would report reduced DIS and increased PWB.
Hypothesized, specified model of the importance of life roles on mental health outcomes. Marital roles are hypothesized to be mediated through marital satisfaction. + = predicted positive relationships. − = predicted negative relationships.
Method
Participant Characteristics
Participants were female volunteers (N = 86) who accompanied their husbands on an international assignment to Ankara, Turkey. The participants were from North American or Western European countries and expected their assignments to last from 1 to 10 years (M = 3.11, SD = 1.68). Their ages ranged from 24 to 56 (M = 38, SD = 6.09). All had children, with an average of two children accompanying them on this trip. Fifty-seven percent had household help (M = 19.73 hours of help per week, SD = 11.56), 48% volunteered (M = 3 hours per week, SD = 5.69), and 26% were employed (M = 6.55 hours per week, SD = 13.47). Examples of employment included civilian positions at the U.S. Embassy, Department of Defense, and the international schools. Participant recruitment materials stated that while the expatriate spouse could be employed, it must have been the husband’s job that brought them to Turkey. The most common industry classifications for jobs held by the expatriate husbands were military (35%), government (30%), and multinational corporations (13%).
Sampling Procedures
Participants were recruited through networking and snowballing techniques. The first author recruited participants in person, at social gatherings (e.g., Ankara Women’s Club) and events (e.g., the annual craft bazaar at an international school), and by e-mail invitation. Advertisements were also posted in the newsletter created by the U.S. Embassy.
Sample Size, Power, and Precision
We used G*Power 3.1 (Faul, Erdfelder, Buchner, & Lang, 2009) to estimate the sample size needed to sufficiently power the analysis. We specified a linear model regression with fixed effects, a two-tailed test, an effect size ( f 2) of .15, α = .05, 1 − β = .95, with 5 predictors. The recommended sample size was 89. Because our model was more complex (two dependent variables and with one mediating variable) than we could specify in G*Power, we proceeded with the understanding that we ran the risk of being underpowered. In recent years, researchers (e.g., Mallinckrodt, Abraham, Wei, & Russell, 2006; Shrout & Bolger, 2002) have recommended that bootstrap resampling methods be used to test for the significance of indirect effects in mediated models. These authors argue that the bootstrapping technique can maximize statistical power by computing nonsymmetric confidence intervals (CIs) and thus reduce the probability of Type II error; an effect is considered to be statistically significant if the interval does not contain zero. These procedures are recommended for research involving small samples.
Measures
Participants completed the packet of measures via e-mail or paper and pencil. The packet (in English only) contained an information letter, two copies of the informed consent, and the measures. Except for format (to ease computer completion), the packets were identical. To protect privacy, participants were cautioned about the risks associated with the e-mail option and advised to delete the completed copy of measures they had saved to their computer.
Life role salience
The LRSS (Amatea et al., 1986) assessed the value and commitment dimensions of the occupational, parental, marital, and home care roles. Each role is evaluated in terms of role reward value and role commitment level expectations. The role value dimension is indexed by statements in which the individual agrees that the role is an important source of self-definition and/or personal satisfaction. Example items include, “Having a successful marriage is the most important thing in life to me” (marital role value) and “Having a comfortable and attractive home is of great importance to me” (home care role value). Statements assessing the role commitment dimensions describe the extent to which the individual is willing to commit personal resources to ensure success in, or to develop, that role. Examples include, “I expect to devote a significant amount of my time and energy to the rearing of children of my own” (parental commitment) and “I expect to devote a significant amount of my time to building my career and developing the skills necessary to advance in my career” (occupational commitment). The LRSS consists of 40 Likert-type items measured on a 1 (disagree) to 5 (agree) scale, with 5 statements for each of the value and commitment subscales.
A series of three studies by Amatea and colleagues (1986) established the validity and reliability of the LRSS among separate samples of college undergraduates (N = 435), university faculty (N = 192), and married couples expecting a baby (N = 150). Coefficient α estimates of internal consistency were adequate and ranged from .79 (home care role commitment scale) to .94 (marital role value scale). Subsequent psychometric support for the LRSS was offered in a study of 94 women who were employed in nonprofessional jobs (Campbell & Campbell, 1995). Among the results, Campbell and Campbell (1995) reported correlations between value and commitment subscales for each of the four roles that were higher (.55–.71) than those reported by Amatea et al. (1986; .38–.59). Consequently, they recommended a combined role importance scale. Regarding the psychometric performance of the LRSS in our study, αs ranged from .61 (parental role value scale) to .89 (marital role value scale). αs for all scales are reported in Table 1.
Variable Means, Standard Deviations, and Correlations.
Note. Pearson correlations are presented below the diagonal. LRSS scales are abbreviated as follows: occ_val and O_V are occupational role value; occ-comm and O_C are occupational role commitment; par_val and P_V are parental role value; par_com and P_C are parental role commitment; mar_val and M_V are marital role commitment; mar_com and M_C are marital role commitment; hom_val and H_V are home care role value; hom_com and H_C are home care role commitment. MHI scales are abbreviated as follows: PsyDis and DIS are psychological distress; PsyWell and PWB are psychological well-being. The Kansas Marital Satisfaction Scale is abbreviated as MarSat and KMS.
*p < .05. **p < .01.
Mental health outcomes
The Mental Health Inventory (MHI; Stewart, Ware, Sherbourne, & Wells, 1998) is a 39-item measure of DIS and well-being designed for use in the general population. Likert-type items range from 1 (always) to 6 (never). Example items include, “During the past month, how much of the time have you been anxious or worried?” and “During the past month, how much of the time have you been a happy person?”
The MHI was fielded in four large samples (N = 5,489), and the results supported a flexible interpretation (Stewart et al., 1998). That is, researchers can employ a single index of mental health functioning, a 2-factor solution measuring DIS and PWB, or a 5-factor solution consisting of anxiety, depression/behavioral–emotional control, belonging/loneliness, positive affect, and cognitive functioning. We used the 2-factor solution. In this configuration, a 22-item subscale assessed DIS and a 10-item subscale assessed PWB. In our study, the coefficient α for DIS was .92, and the PWB α was .90.
Marital satisfaction
The Kansas Marital Satisfaction scale (KMS; Schumm, Milliken, Poresky, Bollman, & Jurich, 1983) is a 3-item scale, with scores ranging from 1 (extremely dissatisfied) to 7 (extremely satisfied). Sample items include, “How satisfied are you with your marriage?” and “How satisfied are you with your relationship with your husband?”
The psychometric properties of the scale have been established in a number of studies with a variety of sample characteristics. Regarding internal consistency, investigations have resulted in strong α coefficients with values frequently above .80 (e.g. Bikos et al., 2007a; Schumm, Nichols, Schectman, & Grigsby, 1983). Test–retest reliability coefficients of .71 were reported over a 10-week interval (Mitchell, Newell, & Schumm, 1983). Reports of concurrent and discriminant validity have been demonstrated in KMS comparisons to the Quality Marriage Index for Wives (Callahan, 1996) and in self-reports of marital stability (Jeong, Bollman, & Schumm, 1992). Additionally, the KMS has adequately distinguished between married and recently separated females (Schumm et al., 1985). In our study, the coefficient α was .97.
Background questionnaire
We collected traditional demographic information about the participants: age, ethnic and/or racial identity, nationality, number of children, and education level. In addition, we gathered information related to expatriation such as previous international experience, languages spoken, ability to speak and read the language of the host country, current occupational status, occupational status prior to expatriation, current volunteer involvement, and volunteer involvement prior to expatriation.
Results
Evaluating the Relative Importance of Life Role Salience Variables
We utilized a two-way within-subject analysis of variance to evaluate the differences among role expectations (i.e., value and commitment) and role categories (i.e., occupational, parental, marital, and home care). Means (Ms), standard deviations (SDs), and correlations among these variables are presented in Table 1. The Role Expectation × Role Category interaction effect was significant, Wilks’ Λ = .23, F(3, 80) = 90.088, p < .001, multivariate eta squared = .005. The role expectation main effect was not significant, Wilks’ Λ = .99, F(1, 82) = .434, p = .512, multivariate eta squared, = .772. However, the role category main effect was significant, Wilks’ Λ = .23, F(3, 80) = 90.146, p < .001, multivariate eta squared, = .772. In light of the significant main effect, we wondered whether there were differences in the overall levels of importance among the role categories. Therefore, we created a role importance variable by combining the role and commitment items for each of the four role categories. Three of the six paired sample t tests were significant (Table 2, upper half). In each of these, the importance of the occupational role was more than 1.5 SD units lower (the d statistic was our estimate of effect) than each of the other three. Type I error was managed with the Bonferroni method.
Means, Standard Deviations, and Results of Paired t tests for Role Importance and Role Expectations.
Note. Occ is occupational, Par is parental, Mar is Marital, Hom is home care.
*Indicates statistical significance according to the Bonferroni Method for control of Type I error (p < .008 for role importance comparisons; p < .0125 for role value vs. commitment comparisons).
To further explore the significant interaction effect, we tested a simple main effect. Specifically, within each of the four role categories, we conducted paired samples t tests to compare the value and commitment dimensions. Although all four comparisons were statistically significant, the effect was strongest for two categories: parental value was stronger than parental commitment; marital commitment was stronger than marital value. Results and an illustration are in the lower half of Table 2 and in Figure 2, respectively.
Role Expectation × Role Category interaction effect.
Modeling Mental Health Outcomes as a Function of Life Role Importance
Our approach to fitting the structural model assumed a model generating approach (Jöreskog, 1993; see also Byrne, 2010). That is, we began with an a priori specified model. Assuming that our model might be incorrectly specified, our goal was to identify significant paths we had not included a priori and trim nonsignificant paths. Our goal was to produce a model that was theoretically meaningful, statistically well fitting, and parsimonious (Byrne, 2010; Kenny, 2009; Kline 2005). Life role salience theory guided our specification. We began by specifying our hypothesized model (Figure 1). This model included (a) four correlated exogenous variables (i.e., the LRSS role importance scores for all four role categories), (b) a mediating variable (martial satisfaction) between the marital role importance score and the outcome variables, and (c) two correlated endogenous variables: DIS and PWB.
In a manner consistent with Byrne’s (2010) recommendations, assessment of model fit was based on three criteria. These included the chi-square (χ2) likelihood ratio statistic, the comparative fit index (CFI), and root mean square error of approximation (RMSEA). Our goal was to have a nonsignificant χ2 value, indicating that variances and covariances of our hypothesized model (i.e., the one we specified) did not differ significantly from those in the data set. Additionally, we sought a CFI value greater than .95. CFI values range from .00 to 1.00 and provide a comparison of the hypothesized model with the independence (or null) model where correlations among all variables are zero. Finally, we desired an RMSEA value of less than .05. The RMSEA value is expressed per degree of freedom (making it sensitive to model complexity) and provides an estimate of error of approximation in the population. The RMSEA scores are accompanied by a 90% CI, allowing us to gauge the precision of the RMSEA value in describing model fit in the population.
The specified model accounted for 18% of the variance in the mediating variable marital satisfaction, 15% of the variance in the endogenous variable DIS, and 27% of the variance in the endogenous variable PWB, χ2 (3, N = 82) = 2.505, p = .474; CFI = 1.000; RMSEA = .000 (90% CI [.000, .175].
To modify the model, we employed a model trimming approach (Kenny, 2009; Kline, 2005), using structural equation modeling (SEM) with maximum likelihood estimation procedures in AMOS 18.0. Model trimming involved re-specifying the model as just identified (i.e., with all possible paths and no degrees of freedom) and eliminating each nonsignificant regression weight, one at a time, on the basis of the largest associated p value. The goal of trimming was to (a) identify any previously unspecified significant paths and (b) trim nonsignificant paths. Because the trimmed path is nested within the saturated model, neither the proportion of variance accounted for in the endogenous variables nor the fit indices would be improved as a result of trimming (Byrne, 2010). To guide the trimming process, we monitored the magnitude and statistical significance of the regression weights and proportions of variance accounted for in the endogenous variables. In our trimmed model, no previously unspecified paths were identified and there were five fewer (e.g., paths deleted) paths than in the specified model.
Because the trimmed model retains indirect effects, we followed the recommendations by Mallinckrodt et al. (2006) to determine whether they were statistically significant. Using the bootstrap option in AMOS, we created 1,000 bootstrap samples drawn by default with replacement from the full set of 82 cases with nonmissing data, with 90% bias-corrected CIs and bootstrap estimates of indirect, direct, and total effects. The choice to use 90% CIs was made because (a) results from the 90% CI were more powerful and their results supported our theoretical model, (b) moving from 95% to 90% CIs is not uncommon in exploratory research, and (c) it is the default in AMOS.
When nonsignificant paths were removed, the trimmed model (Table 3, Figure 3) included a positive relationship between occupational role importance and DIS (β = .14), a positive relationship between parental role importance and PWB (β = .14), and a positive relationship between home care role importance and PWB (β = .15). Additionally marital satisfaction mediated the relationship between marital role importance and the endogenous variables DIS (β = −.12) and PWB (β = .18). Details regarding the statistics of the direct paths, indirect paths, and covariances are presented in Table 3. These paths accounted for 18% of the variance in marital satisfaction, 10% of the variance in DIS, and 24% of the variance in PWB. Regarding fit statistics: χ2(8) = 7.122, p = .523; CFI = 1.000; RMSEA = .000; 90% [.000, .121]. We compared the trimmed model to the saturated model with chi-square and CFI difference tests of the nested models (Byrne, 2010). For the chi-square difference test, models are not significantly different when the p value is greater than .05; for the CFI difference test, models are considered to be invariant when the difference is less than .01. Results indicated that the trimmed model was not statistically different from the fully saturated model, Δχ2 (5, N = 81) = 4.617, p > .05; ΔCFI = .000. Consequently, we prefer the trimmed model because it explains maximal variance in endogenous variables with the fewest paths. The final, trimmed model is shown in Figure 3.
Trimmed model of the importance of life roles on mental health outcomes. The marital role effect is mediated through marital satisfaction. Associated statistics are presented in Table 3.
Bootstrap Results to Test Significance of Mediation Effects.
Note. N = 82. The 90% confidence intervals (CIs) for both the unstandardized and standardized results were produced with the bias-corrected option. Occ = occupational importance role; Par = parental importance role; Mar = marital importance role; Hom = home importance role; MarSat = marital satisfaction; PsyDis = psychological distress; PsyWell = psychological well-being; Dis1 (PsyDis), Dis2 (PsyWell) = disturbances of endogenous variables.
Discussion
In our study of female expatriate spouses, dimensions of life role salience assisted in explaining mental health outcomes. Results of a main effect comparison of relative role importance indicated that women viewed their occupational role as significantly less important as parental, marital, and home care roles. When the value and commitment dimensions were compared within each of the four role categories, all were statistically significantly different. That is, there was a significant interaction effect. For both the occupational and parental roles, there was stronger value than commitment; for both marital and homemaker roles, there was lower value than commitment. These effects were strongest in the parental and marital comparisons. Results of these role comparisons were generally consistent with the Bikos et al. (2007a) findings.
Our second purpose was to test a path model evaluating effects of role importance for each of the four life roles upon the mental health outcomes of DIS and PWB; in the case of the marital life role, we hypothesized that outcomes would be mediated by marital satisfaction. In our trimmed model, the importance of the occupational role was positively correlated with DIS. That is, those who held greater value for the occupational role experienced a greater degree of DIS. In contrast, parental and home care role importance was positively correlated with PWB. Additionally, results suggested that marital satisfaction mediated the relationship between marital role importance and the endogenous variables. While the direct paths between marital role importance and the DIS and well-being measures were trimmed, the indirect paths remained significant. In causal terms, this means that marital satisfaction transmits the impact of marital role importance to DIS and PWB. Overall, our trimmed model accounted for 10% of the variance of DIS, 24% of PWB, and 18% of the mediating variable, marital satisfaction. While there are fewer paths in the trimmed model than in our specified model, the direction of each path was consistent with our hypotheses.
Our results are the first to provide quantitative support to the qualitative expressions of frustration associated with career-related role changes for the expatriate spouse. When we consider that the importance of the occupational role was statistically significantly lower than the remaining three (consistent with findings in Bikos et al, 2007a), and yet there was a positive relationship between occupational role importance and DIS, we conclude that for the majority of female expatriate spouses, the transition associated with career identity is not problematic. However, there is a subset for whom this role change is salient and distressing.
The positive effect of parental role importance on PWB and the effects of marital role importance on DIS (negative) and PWB (positive) are consistent with previous findings. That is, researchers have noted the contribution of parenting in providing social support in the home as well as in building social connections and in providing stability in social interactions outside of the home environment (Mohr & Klein, 2004; Shaffer & Harrison, 2001). For our sample of expatriate female spouses, the positive contribution of the home care role to PWB seems to fall into this bundle of family-related roles that enhance well-being and mitigate distress.
Our research findings echo results from Hall and Duvall’s (2003) investigation of females in the global mission field (not necessarily expatriate spouses, but often missionaries in their own right). Hall and Duvall reported a positive relation between self-expectations and well-being. That is, women who were functioning in their desired role (i.e., homemaker, background supporter, team worker, parallel worker) reported higher well-being scores.
Limitations
Decisions we made to enhance some elements of the study diminished others. The singular location (Ankara, Turkey) and the restriction of participants (females from North America or Western Europe) reduced the variability among participants on factors (e.g., similarity of host country characteristics and living conditions) that were not included in the statistical analysis. Such restrictions likely (a) improved our ability to interpret the findings and (b) increased the capacity to find a statistical effect. In contrast, external validity is negatively impacted in that such restrictions limit the generalizability of the results.
The path trimming process we utilized is not without controversy. Jöreskog (1993) identified three approaches to testing SEMs: strictly confirmatory, test of alternative models, and model generating. Byrne (2010) argued that because rejection of a hypothesized model is a common occurrence (and because its data collection is expensive), the majority of SEM work is necessarily model generating. We believe we have followed the recommendations of those (e.g., Jöreskog, 1993; Byrne, 2010; Kenny, 2009; Kline, 2005) who provide guidance for alternative model and exploratory approaches like model trimming by comparing only nested models, by evaluating the significant paths in light of our theory, and by striving for parsimony. While our results provided support for our general ideas, not all theorized paths were significant and not all of the regression weights and correlations for those retained were strong. Consequently, we likely capitalized on sample-specific characteristics. Similarly, the sample size (N = 82) for the path modeling was somewhat underpowered.
Two statistical issues may restrict the interpretation of our findings. First, because of the low αs of some of the LRSS subscales (e.g., .61 for the parental role value scale), the relationship between these subscales and other variables in the model may be distorted. While numerous studies report an adequate α coefficient for the parental role value scale (e.g., Amatea et al., 1986), it may be that among expatriate spouses different parental role values come into play, undermining the unidimensionality of the scale. Additionally, there was a strong negative correlation between the disturbances of our endogenous variables (r = .69; DIS, PWB). A likely explanation for this strong relationship is a common methods factor; these are subscales from the same instrument. Their strong covariance may also contribute to the strength of the fit indices. In light of these limitations, researchers contemplating replication should consider including multiple host countries, an international diversity among participants, and larger sample sizes.
Implications for Practice and Research
For the majority of women in our study, the importance of the family-related roles was high and contributed to positive mental health outcomes. However, there appears to be a subset for whom the occupational role is important but role expectations are unmet; these unmet expectations may contribute to DIS. Consequently, professionals who screen, select, and support expatriates may wish to (a) inquire about role expectations, particularly in the area of career/occupation, (b) assist potential sojourners in assessing the degree to which their role expectations will be met, and (c) cultivate alternate plans for meeting expectations around unmet life role expectations. That is, those who screen and counsel expatriate couples should have frank conversations about how the spouse plans to spend his or her time, find fulfillment and meaning, and meet career-related expectations in the host country. For example, while it may be reasonable for a spouse of a U.S. Department of Defense or Department of State employee to acquire a civilian post in the host country, the ability of a spouse of a multinational corporation to find work will depend on host country characteristics and requirements such as work permits, credentialing (e.g., training, certification, and licensure), and job availability.
Although Hall and Duvall (2003) did not reference social learning theory (i.e., Bandura, 1986) in their work, their findings about the importance of role expectations echo an increasing introduction of social learning theory to modeling expatriate adjustment (e.g., Black, Mendenhall, & Oddou, 1991) and sociocultural adaptation (e.g., Searle & Ward, 1990). Consequently, researchers with specialization in career development/vocational psychology may benefit from conducting research in this area from the perspective of social cognitive career theory (SCCT; Betz & Hackett, 1986, 2006). SCCT extended Bandura’s work to career development by suggesting that the interaction between cognitive–personal variables (e.g., self-efficacy, outcome expectations), contextual factors (e.g., perceived support and barriers), and overt behaviors (e.g., career decision) promote or impede career development processes (e.g., interests, choice, and performance). In particular, investigating the relationship between role-related outcome expectations and goals, self-efficacy for meeting those expectations and goals, and the host-country-specific capacity for doing is a logical next step in this line of research.
Conclusion
Our study provides support for the contribution of life role salience theory in predicting mental health outcomes of female expatriate spouses. Particularly since (a) foregoing expatriate research has suggested that spousal adaptation is the most salient predictor of employee success (e.g., Bhaskar-Shrinivas et al., 2005; Hechanova et al., 2003) and (b) most research involving the expatriate spouse focuses on his or her contribution to employee success (e.g., Lazarova et al., 2010), we suggest that there is a business case for continuing this program of research and utilizing life role salience theory in evaluating, preparing, and supporting female expatriate spouses. Specifically, we suggest that in research and practice, assessment of role importance and the ability to meet role expectations be considered and evaluated.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.