Ridderprint

Chapter 5 104 Table 5.7 also demonstrates that the physical proximity of agents did not explain significant variance. Similarly, the average association between adjustment and social support was higher when agents resided in the host country (ρ = .27) rather than in the home country (ρ = .21), but this difference was not significant. The confidence intervals of the effects of social support by host and home country agents on commitment, performance and retention outcomes strongly overlapped, leading us to reject hypotheses 4a through 4d. 5.5.2.2 Criterion Proximity Hypothesis 5 expected that studies would report a stronger relationship between social support and expatriate success when they had measured the latter with proximal (i.e., cross-cultural adjustment and commitment) rather than distal criteria (i.e., performance and retention). Table 5.8 confirmed this hypothesis with a significantly larger association between social support and proximal criteria (ρ =.27) compared to distal criteria (ρ =.19; Q b = 8.21, p < .01). Table 5.8: Random effects models of moderator effect of outcome criterion k N r ρ se CI95 CR80 I 2 Q w Q b Proximal criteria 67 13570 .23 .27 .02 [ .23, .31] [ .08, .46] 97.53% 2791.55*** 8.21** Distal criteria 51 10622 .17 .19 .02 [ .16, .23] [ .05, .34] 90.79% 567.54*** Note: k = number of independent studies; N = total independent sample size; r = estimated correlation in the population; ρ = estimated true correlation in the population; se = standard error of the true correlation; CI95 = 95% confidence interval of the true correlation; CR80 = 80% credibility interval of the true correlation; I 2 = variance in true effect sizes as proportion of total variance; Q w = test of within-group homogeneity; Q b = test of between-group homogeneity. * p < .05; ** p < .01; *** p < .001 5.5.2.3 Rater Effects The use of self-evaluations of performance is the norm in expatriate management literature but our results in Table 5.9 demonstrate this has not inflated the association between social support and performance outcomes. Research designs where the same respondent rates multiple variables can introduce artefactual covariance between these variables, as respondents follow response tendencies based on their personal dispositions, their beliefs regarding specific correlations, and their desires to be consistent, lenient, and compliant with social norms (Podsakoff et al., 2003). However, studies did not report significantly stronger associations when they used expatriates’ self- evaluations (ρ = .19) instead of performance ratings by supervisors or peers (ρ = .19). T able 5.9: Random effects models of moderator effect of raters k N r ρ se CI95 CR80 I 2 Q w Q b Performance Same rater 22 3596 .16 .19 .03 [ .13, .26] [ .00, .39] 91.61% 278.89*** 0.03 Different rater 15 2537 .16 .19 .02 [ .14, .23] [ .08, .30] 74.80% 66.93*** Note: k = number of independent studies; N = total independent sample size; r = estimated correlation in the population; ρ = estimated true correlation in the population; se = standard error of the true correlation; CI95 = 95% confidence interval of the true correlation; CR80 = 80% credibility interval of the true correlation; I 2 = variance in true effect sizes as proportion of total variance; Q w = test of within-group homogeneity; Q b = test of between-group homogeneity. * p < .05; ** p < .01; *** p < .001