Best Practices and Industry Benchmarks

Preliminary Analyses

The estimated ICC (1) was 0.14 for supervisor-rated high performance expectations, implying that around 14% of the variance in high performance expectations was attributed to supervisor-level factors. This means that the data were nested. Thus, in subsequent analyses, we controlled for the between-level variances of the supervisor-rated outcome variables that were significantly explained at the supervisor level.

Table 1 shows the means, standard deviations, and correlations of our studied variables. An examination of the zero-order correlations provided initial support for our hypotheses. As expected, high performance expectations were significantly correlated with stress (r = 0.290, p < 0.01), and stress was shown to significantly affect territoriality (r = 0.198, p < 0.01).

Table 1. Means, standard deviations, and correlations.

Variables	M	SD	1	2	3	4	5	6	7	8
1. Education	3.080	1.090	——
2. Income	7.110	4.749	0.554 **	——
3. Gender	1.510	0.501	−0.141 *	−0.259 **	——
4. Age	32.960	5.908	−0.139 *	0.178 **	−0.120 *	——
5. Highperformance expectations	3.701	0.832	−0.007	0.159 **	−0.086	0.146 *	——
6. Stress	2.858	0.582	0.167 **	0.160 **	−0.258 ***	0.096	0.290 ***	——
7. Territoriality	3.166	0.620	−0.042	−0.127 *	−0.088	−0.051	−0.024	0.198 **	——
8. Task autonomy	3.747	0.741	0.118 *	0.190 **	−0.089	0.028	0.297 **	0.078	−0.030	——

* p < 0.05, ** p < 0.01, *** p < 0.001. For all variables, n = 291. For education, 1 = junior middle school, 2 = high school, 3 = junior college, 4 = bachelor, 5 = master, 6 = doctor. For gender, 1 = male, 2 = female.

Confirmatory Factor Analysis

We carried out a confirmatory factor analysis to verify the construct validity of the scales for high-performance expectations, stress, territoriality, and task autonomy. Because the number of measurement items overstepped the suggested parameters for the sample size ratio for evaluation, we packaged the scale items for stress into three parcels following previous research. Bandalos argued that the inclusion of all measurement items as observed indicators in the original model will result in some parameter estimation bias because the recommended parameters to sample size ratio will be exceeded. Thus, following Rogers et al., our study used item parceling, which adopts a high load strategy. Stress was measured using 13 items. According to the order of load factor, we averaged four items as the first parcel (items 1–4), four items as the second parcel (items 5, 6, 7, and 10), and the remaining five items as the third parcel. The hypothesized four-factor model (χ² = 127.94; df = 59; CFI = 0.939; RMSEA = 0.063; TLI = 0.919) was superior to the other three models. The results in Table 2 showed that the distinctiveness of the four constructs in this study was acceptable.

Table 2. Comparison of measurement models.

Model	Factors	χ2	df	χ2/df	RMSEA	TLI	CFI	Δχ2
Baseline Model	Four Factors	127.94	59	2.168	0.063	0.919	0.939
Model 1	Three factors - High performance expectations and stress combined	297.01	62	4.790	0.114	0.738	0.791	169.07
Model 2	Two factors - High performance expectations, stress, task autonomy combined	468.02	64	7.313	0.147	0.641	0.563	340.08
Model 3	All four factors were combined	899.09	65	13.832	0.210	0.260	0.112	771.15

Hypothesis Testing

Snijders and Bosker's formulas were used to calculate the pseudo-R2 for the effect sizes in predicting outcomes. In particular, the two-level structural model estimates showed that after ruling out variances due to team membership, high performance expectations were positively related to stress (γ = 0.208, SE = 0.054, p < 0.001, H1) at the employee level. Stress had significant and positive effects on territoriality (γ = 0.230, SE = 0.075, p < 0.01, H2). The stress–territoriality link was moderated by stress (γ = 0.275, SE = 0.114, p < 0.05, H3). The parameter bootstrapping results are shown in Table 3.

Table 3. Multi-level structural equation modeling results.

Predictors	Stress		Territoriality
	γ	SE	γ	SE	γ	SE
	Model 1		Model 2		Model 3
Education	0.047	0.063	0.026	0.061	0.023	0.045
Income	0.001	0.018	−0.033 **	0.011	−0.027 **	0.008
Gender	−0.292 ***	0.073	−0.07	0.083	−0.028	0.071
Age	0.008	0.007	−0.008	0.008	0.002	0.007
High performance expectations	0.208 ***	0.054			−0.066	0.059
Stress			0.230 **	0.075	0.345 **	0.084
Task Autonomy					−0.013	0.055
Stress * Task Autonomy					0.275 *	0.114
Pseudo-R2	0.09		0.18		0.09

* p < 0.05, ** p < 0.01, *** p < 0.001. Pseudo-R²: Snijders and Bosker’s (1994) formulas were used to calculate pseudo-R2 for the effect sizes in predicting outcomes. The indirect effects and conditional indirect effects were tested using two-level modeling in Mplus 7.4 with 5000 times of parameter bootstrapping. γ and SE refer to the unstandardized parameter estimates and their corresponding standard errors, respectively.

From Figure 2, it can be seen that task autonomy moderated the relationship between stress and territoriality. Specifically, when task autonomy was high (1 SD above the mean), stress was more pronounced with territoriality; whereas when it was low (1 SD below the mean), stress was less positively related to territoriality.

Figure 2. Task autonomy moderates the relationship between stress and territoriality.

With respect to Hypothesis 4, a moderated mediation model was estimated, which moderated the second stage. In this study, we used the Monte Carlo simulation bootstrap procedure recommended by Preacher and Selig to estimate the unbiased confidence intervals of conditional indirect effects. The confidence level was set to 95%, and the bootstrap self-sampling number was set to 5000. The test results can be seen in Table 4 below.

Table 4. Test two-level modeling results of the moderated mediation effect.

Group Statistics	γ	SE	95% Confidence Interval
			Lower Limit	Upper Limit
Conditional indirect effect
High Task Autonomy (+1 SD)	0.102	0.038	0.028	0.175
Low Task Autonomy (−1 SD)	0.009	0.024	−0.038	0.056
DIFF	0.092	0.046	0.002	0.183

As Table 4 indicates, the negative effect between high performance expectations and territoriality through the conduction path of stress was significantly stronger for employees with higher levels of task autonomy than employees with lower levels of task autonomy. The difference between the two levels reached a significant level (γ = 0.092, SE = 0.046, p < 0.0), and its 95% unbiased interval [0.002, 0.183], excluding 0. Thus, Hypothesis 4 was supported by the observation data. The multilevel structural equation modeling results are summarized at the bottom of Figure 3.

Figure 3. Multilevel structural equation modeling results. * p < 0.05, ** p < 0.01, *** p < 0.001.