Ridderprint

Fostering overseas success: A meta-analysis 99 which excludes zero indicates that the population parameter is likely not zero whereas a 80% credibility interval (CR) that includes zero suggests that additional variance may be explained by moderating factors (Whitener, 1990). Higgins and Thompson’s I 2 (2002) represents the proportion of the total variability in effect sizes that can be attributed to the heterogeneity of effect sizes. Cochran’s Q (Cochran, 1950) was used to test the remaining heterogeneity within ( Q w ) and explained heterogeneity between effect sizes ( Q b ). For moderator analyses, a significant between-group homogeneity statistic ( Q b ) indicates that the moderator variable explained significant variance between effect sizes whereas non-overlapping confidence intervals indicate significant differences in effect size estimates between subgroups exist at the p = .05 level (Hunter & Schmidt, 2004). Duval and Tweedie’s (2000) “trim and fill ” method was used to obtain a publication-bias- adjusted estimate of the true mean effect size. This method can be used as a sensitivity analysis, providing an indication of how the effect size could have affected by publication bias (Rothstein, Sutton, & Borenstein, 2006, p. 142). Finally, we used the estimated effect sizes to algorithmically construct a graphical mapping of the support-success relationships (see Kamada & Kawai, 1989), with stronger relationships resulting in shorter, broader edges, and constructs with similar relationships being co-located. 5.5 Results 5.5.1 Main Effects Table 5.2 displays the overall relation found between social support and expatriate success. After pooling dependent effect sizes, the average true effect size of the 84 independent samples is estimated at ρ = .24. With significant heterogeneity remaining ( Q w = 2581.55, p < .001), we added a moderator term for the specific criteria. This improved model fit significantly ( Q b = 12.38, p < .01) and the results were in line with Hypotheses 1a through 1d with positive relationships between social support and expatriates’ adjustment (ρ = .25), commitment (ρ = .33), performance (ρ = .19), and retention (ρ = .21). Examining subdimensions of criteria only improved model fit significantly for commitment outcomes ( Q b = 11.75, p < .01), where social support had no significant effect on continuance commitment. Because of the above, further analysis were conducted on each criterion separately. 5.5.1.1 Cross-cultural Adjustment Table 5.3 displays that social support from each domain has similar positive relationships with expatriates’ cross-cultural adjustment, confirming Hypothesis 2a: ρ = .25 for community support, ρ = .32 for family support, and ρ = .24 for work-related support. Within domaims, the relationship between HCN community support and adjustment remained positive (ρ = .22) but support by expatriate community members had no significant association (ρ = .13). The estimated effect of spousal support was slightly higher (ρ = .32) than family support in general (ρ = .25). Support from different coworker groups had similar effects with ρ = .27 for mentor and peer support, and ρ = .24 for supervisory support. Regarding organizational support, the relationship between adjustment and perceived adjustment-oriented support was most notable (ρ = .36).