4.4. Hypotheses Testing
This paper used the multiple regression method of Baron−Kenny (1986) to test the main effect, the mediating effect, and the complementary effect. The hierarchical regression results in
Table 4 shows that (1) compared with M3, M4 had a positive impact on entrepreneurial opportunity identification after the influence of fixed control variables (β = 0.212,
p < 0.001); β = 0.534,
p < 0.001) and can additionally explain the entrepreneurial opportunity recognition variation of up to 40.8% (ΔR
2 = 0.408). The results show that absorbed slack resources and unabsorbed slack resources had a significant positive impact on entrepreneurial opportunity identification. Hence, H1a and H1b were supported. (2) Compared with model 3, after the influence of model 7 on fixed control variables, the regression coefficient of resource bricolage was significantly positive (β = 0.610,
p < 0.001), and an additional 35.9% of (ΔR
2 = 0.359) entrepreneurship opportunity identification can be explained. The results show that resource bricolage had a significant positive impact on entrepreneurial opportunity identification. H2 was supported. (3) Compared with model 4, after the influence of fixed control variables, absorbed slack resources and unabsorbed slack resources, the regression coefficient of resource bricolage was significantly positive (β = 0.288,
p < 0.001) and can extra explain 3.4% (ΔR
2 = 0.034) of entrepreneurial opportunity identification. Model 4, model 6, and model 7 provided preliminary support for the mediation test of the resource bricolage.
To further test the median effect, this paper used the mediation effect test procedure proposed by Preacher and Hayes [
60], and used the Bootstrap method to test the mediation effect of resource bricolage through the process plugin of SPSS. Mackinnon et al. [
61] found that the asymmetric confidence interval method abandons the premise that the sampling distribution of the mediating effect was a normal distribution and did not limit the sampling distribution of the mediating effect. The percentile method of bias correction provided the most accurate confidence interval estimation and reduced the probability of statistical error, and had higher statistical power. In this paper, the Bootstrap sample size was 2000, the confidence interval was set to 95%, and the mediation effect and confidence interval are shown in
Table 5.
The results of Bootstrap mediation analysis show that at the 95% confidence level, (1) the median direct effect of the resource bricolage between the absorbed slack resources and the entrepreneurial opportunity identification was 0.116, and the confidence interval (LLCI = −0.021, ULCI = 0.253) includes 0, indicating that the resource bricolage variable was added, the effect of absorbed slack resources on the identification of entrepreneurial opportunity was not significant; the indirect effect was 0.323, and the confidence interval (LLCI = 0.222, ULCI = 0.456) did not include 0, indicating that the resource bricolage played a mediating role between absorbed slack resources and entrepreneurial opportunity identification. H3a was supported. It can be seen from the direct effect that resource bricolage played a fully mediating role between absorbed slack resources and entrepreneurial opportunity identification. (2) The direct effect of resource bricolage between unabsorbed slack resources and entrepreneurial opportunity identification was 0.386, and the confidence interval (LLCI = 0.244, ULCI = 0.527) did not include 0, indicating that after the resource bricolage variable was added, the impact of resource bricolage on entrepreneurial opportunity identification was still significant. Its indirect effect value was 0.237, and the confidence interval (LLCI = 0.117, ULCI = 0.369) did not include 0, indicating that resource bricolage played a mediating role between unabsorbed slack resources and entrepreneurial opportunity identification. H3b was supported. It can be seen from the direct effect that resource played partial mediating role between unabsorbed resources and opportunity identification.
In this paper, the multivariate regression method of Baron–Kenny [
62] was used to examine the moderating effects (
Table 6).
(1) The moderating effect of absorbed slack resources on entrepreneurial opportunity identification. After adding the business ties variable in model 8, the regression coefficient of the business ties was significant (β = 0.341,
p < 0.001). The results show that the business ties had a significant positive impact on entrepreneurial opportunities identification. Based on model 8, after adding the interaction term of absorbed slack resources and business ties in model 9, the regression coefficient of the interaction term between absorbed slack resources and business ties was positive (β = 0.136,
p < 0.050). This shows that business ties played a positive moderating role between absorbed slack resources and entrepreneurial opportunity identification. H4a was supported. To test H4a more comprehensively, this paper calculated the simple slope of low business ties and high business relationship by using the mean value of the moderating effects variable (business ties) plus or minus one standard deviation as the grouping criterion. The results show that under low business ties, the simple slope value was −0.071, the t value was −0.328, and the p value was 0.746 > 0.050. Under high business ties, the simple slope was estimated to be 0.178, the t value was 0.811, and the
p value was 0.426 > 0.050. Thus, the moderating effect diagram of the business ties was drawn. As can be seen from
Figure 2: Compared with the low level of business ties, the positive impact of absorbed slack resources on entrepreneurial opportunity identification was stronger; the possibility of high business ties influencing opportunity identification raised faster than the low business ties. In other words, the slope was larger. After adding political tie variables in model 10, the regression coefficient of political ties was not significant (β = −0.045,
p > 0.050). The results show that political ties had no significant impact on entrepreneurial opportunities identification. Based on model 10, after adding the interaction term of absorbed slack resources and the political ties in model 11, the regression coefficient of the interaction term between absorbed slack resources and political ties was not significant (β = 0.092,
p > 0.050). It shows that the moderating role of business ties between absorbed slack resources and entrepreneurial opportunities identification was not significant. H4b was not supported.
(2) The moderating effect of unabsorbed slack resources on entrepreneurial opportunity identification. After adding the business ties variable in model 12, the regression coefficient of the business ties was significant (β = 0.221,
p < 0.010). The results show that the business ties had a significant positive impact on opportunity identification. Based on model 12, after adding the interaction term of unabsorbed slack resources and business ties in model 13, the regression coefficient of the interaction term between unabsorbed slack resources and business ties was positive (β = 0.171,
p < 0.010). This shows that business ties played a positive moderating role between unabsorbed slack resources and entrepreneurial opportunity identification. H4c was supported. To test H6c more comprehensively, this paper calculates the simple slope of low business relationship and high business relationship by using the mean value of the regulatory variable business relationship plus or minus one standard deviation as the grouping standard. The results show that under low business ties, the simple slope value is 0.317, the t value is 1.470, and the p value is 0.157 > 0.050. Under high business ties, the simple slope is estimated to be 0.624, the t value is 3.487, and the p value is 0.002 < 0.010. Thus, the moderating effect diagram of the business ties was drawn. As can be shown in
Figure 3: Compared with the low level of business ties, the positive impact of unabsorbed slack resources on entrepreneurial opportunity identification was stronger. The possibility of high business ties influencing opportunity identification raised faster than the low business ties. In other words, the slope was larger. After adding political tie variables in model 14, the regression coefficient of the interaction term between unabsorbed slack resources and political ties was not significant. (β = −0.081,
p > 0.050). Based on model 14, after adding the interaction term of unabsorbed slack resources and political ties in model 15, the regression coefficient of the interaction term of unabsorbed slack resources and political ties was not significant (β = 0.027,
p > 0.050). It shows that the moderating role of political ties between unabsorbed slack resources and opportunity identification. H4d was not supported.
For the moderated mediation effect test of the main effect stage, this paper followed the test flow proposed by Edwards and Lambert (2007), and used Mplus 7.0 to test the moderated mediation effect of the main effect. The specific test procedure was establishing a regression equation for absorbed slack resources to resources bricolage and regression equations for slack resources, resource bricolage, business ties, and resource bricolage*business ties for entrepreneurial opportunity identification, specifically:
M stands for resources bricolage; X
A stands for absorbed slack resources; X
U stands for unabsorbed slack resources; Y stands for entrepreneurial opportunity identification; Z stands for business ties; a
0A, a
0U, b
0A, and b
0U stand for constant coefficients in the corresponding regression equation, respectively; a
xA and
xU stand for regression coefficient of absorbed slack resources and unabsorbed slack resources on bricolage; b
XA\b
xU, b
m, b
z, and b
mz respectively stand for absorbed slack resources and unabsorbed slack resources, bricolage, business ties, and bricolage*business on entrepreneurial opportunity identification. First, substituting (1) into (2) gives (3), and substituting (4) into (5) gives (6). Second, using the Bootstrap method and taking 209 valid samples as the female parent, we randomly extracted 2000 new samples with a sample size of 209, and used Mplus7.0 to calculate the effect coefficient and its confidence interval; finally determining the significance of each effect and difference based on confidence intervals. The results of the Bootstrap test are shown in
Table 7 and
Table 8.
As was shown in
Table 7, business ties significantly moderated the mediating role of resources bricolage between absorbed slack resources and entrepreneurial opportunity identification. Specifically, under high business ties, the 95% confidence interval for indirect effects is [0.026, 0.197], excluding 0, r = 0.165,
p < 0.001, indicating the impact of absorbed slack resources on entrepreneurial opportunity identification through resource bricolage was significant. Under low business ties, the 95% confidence interval for indirect effects is [−0.001, 0.064], including 0, r = 0.039,
p > 0.050, indicating the impact of absorbed slack resources on entrepreneurial opportunity identification through resource bricolage was not significant. There was a significant difference in indirect effects between the two cases. The 95% confidence interval for this difference was [0.021, 0.193], excluding 0, r = 0.126,
p < 0.050. Therefore, H5a was supported.
As was shown in
Table 8, business ties significantly moderated the mediating role of resources bricolage between unabsorbed slack resources and entrepreneurial opportunity identification. Specifically, under high business ties, the 95% confidence interval for indirect effects is [0.038, 0.203], excluding 0, r = 0.126,
p < 0.010, indicating the impact of unabsorbed slack resources on entrepreneurial opportunity identification through resource bricolage was significant; under low business ties, the 95% confidence interval for indirect effects is [−0.005,0.077], including 0, r = 0.021,
p > 0.050, indicating that the impact of unabsorbed slack resources on entrepreneurial opportunity identification through resource bricolage was not significant. There was a significant difference in indirect effects between the two cases. The 95% confidence interval of [0.014, 0.189], excluding 0, r = 0.105,
p < 0.050. Therefore, H5b was supported.
Since the moderating role of political ties was not significant, political ties did not have a moderated mediation effect between slack resources (absorbed slack redundant and unabsorbed slack resources), resource bricolage, and entrepreneurial opportunity identification. H5c and H5d were not supported.