This chapter demonstrates the result of the data analysis, which was based on the framework of this study. All of the variables and control variables were examined in correlation analysis. Furthermore, the moderator effect and hypothesis were tested at the same time. This study used Statistical Package for the Social Sciences (SPSS) version 22 for windows and AMOS version 24 to analyze the data.
Pearson’s Correlation Analysis
Correlation analysis is a statistic way to explore the strength and direction of the relationship between variables. Correlation value can illustrate the relationship by use of r
> 0 for a positive association and r < 0 for a negative association. The positive and negative correlation will be indicated by (+) and (-) respectively.
The mean, standard deviations, reliabilities, and correlations among all variables are shown in Table 4.1. According to Table 4.1., self-efficacy was positively correlated with all of the variables with values from .40 to .62, such as, career development (r = .62, p
< .01), psychosocial support (r = .40, p < .01), role modeling (r = .52, p < .01), in-role performance (r = .62, p < .01), helping (r = .53, p < .01) and voice (r = .57, p < .01).
Furthermore, career development and psychosocial support and role modeling were proved to be positively correlated with in-role performance, helping and voice behavior. The value between career development and in-role performance was r = .48, p < .01, helping was r
= .49, p < .01, and voice was r = .44, p < .01. Psychosocial support also showed strong relationship with in-role performance (r = .21, p < .01), helping (r = .26, p < .01), and voice (r = .33, p < .01). Moreover, role modeling function has significant correlation with in-role performance (r = .39, p < .01), helping (r = .46, p < .01), and voice (r = .34, p < .01). In conclusion, all three functions of mentoring are significantly associated with individual performance.
47
Furthermore, there were some interesting findings. Gender has negative relationship with in-role performance (r = -.24, p < .01) and helping behavior (r = -19, p < .01). This study coded female into 0, male into 1. Hence, the result can be interpreted as female employees have better in-role performance and are more willing to help others during working. Another finding is that first job experience has positive effect on career development, psychosocial support, role modeling and voice behavior. First job experience was coded, 0 as “No”, 1 as “Yes”. This research focuses on the early-career employees, and the data shows that respondents in their first job have more positive perception of mentoring functions. They felt they received more mentoring help than others, in career development (r = .11, p < .05), psychosocial support (r = .16, p < .01), and role modeling (r = .13, p < .01). In addition, it could be inferred that early-career employees are brave to speak their suggestions for organizations.
48
Table 4.1.
Mean, Standard Deviations, Correlations, and Reliability (N = 395)
Note. Numbers in parentheses represent Cronbach’s alpha value. *p< .05, **p< .01
Code. Age: 1 = Below 20, 2 = 20~29, 3 = 30~39, 4 = Above 40, Gender: 0 = Female, 1 = Male, Education: 1 = Under high school, 2 = High school, 3 = Bachelor Degree, 4 = Master degree, 5 = Doctorate, Year after Graduation: 1 = Still studying, 2 = Below 1 year, 3 = 1~below 2 year, 4 = 2~below 3 year, 5 = 3~below 4 year, 6 = 4~below 5 year, 7 = More than 5 years, Tenure:1 = Below 1 year, 2 = 1~below 2 year, 3 = 2~below 3 year, 4 = 3~below 4 year, 5 = 4~below 5 year, 7 =
49
Regression Analysis
Hierarchical regression analysis was conducted to test the hypotheses. The following Table are the result of regression analysis. Table 4.2., Table 4.3. and Table 4.4. show the result of the effect of predictor variables on in-role performance, helping and voice, respectively. Each Table contains four models. Control variables were tested in the first model, as the baseline of this research. Since the correlation analysis shows only gender and first job have significant relationships with the dependent variables, these two variables were entered in the first model as control variables. Second, the independent variable was entered, to see the effect of self-efficacy. The moderator variables were added to the third model to examine the direct effect. Lastly, the researcher utilized mean centering method for standardizing variables before calculating the interaction term of self-efficacy with moderators. Interaction terms resulting from the product of the independent variable and the moderators were added to the fourth model, for the purpose of testing the moderating effect. Model 3 and 4 were repeated for the second and the third moderators proposed in the research framework. Therefore, for the regression analysis of each dependent variable, eight regression models will be presented.
Relationship between In-role Performance and Mentoring Functions
The result of hierarchical regression is shown in Table 4.2. From model 2, it proves that self-efficacy has a strong positive effect (β = .623, p < .001) on in-role performance.
It supports Hypothesis 1. Model 3, 5, 7 demonstrate the direct effect of the moderators, career development, psychosocial support and role modeling. From Model 3, it shows that career development (β = .137, p < .01) has a significant direct effect, which supports Hypothesis 7a. In addition, model 7 supports Hypothesis 7c; role modeling (β = .087, p < .05) also shows significant positive effect on in-role performance. By contrast, psychosocial support mentoring does not show a positive relationship with in-role
50
performance (β = -.037, p > .05), hence Hypothesis 7b was rejected.
However, the interaction terms are not significant in Model 4 and 6. Thus, Hypothesis 4a and 4b are rejected. Only Hypothesis 4c is supported. Role modeling (β
= -.070, p < .05) has a negative moderating effect on self-efficacy and in-role performance, which means it will weaken the effect of self-efficacy on in-role performance.
R square represents the ratio of the variation of the regression model. The larger the value, the more explanatory of the model. The result of R square change in Model 2 shows that the addition of self-efficacy significantly increased the percentage of variances that can be explained by the model (△R2 = .383, △F = 265.788, p < .001).
However, the moderators increased very little of the variances being explained. Only career development enhances 1.1% of variances (△R2 = .011, △F = 8.026, p < .01).
Other moderators, psychosocial support (△R2 = .001, △F = .795) and role modeling (△R2 = .005, △F = 3.808, p < .05), increased the influence by less than 1 % of variances.
51
Table 4.2.
Results of Hierarchical Regression Analysis – In-role Performance β
In-role Performance
Model 1 2 3 4
Control Variable
Gender -.235*** -.241*** -.235*** -.231***
First Job .036 -.004 -.016 -.010
Independent variable
Self-efficacy .623*** .539*** .521***
Moderator
Career Development .137** .134**
Psychosocial Support Role Modeling Interaction
Self-efficacy x Career Development -.066
Self-efficacy x Psychosocial Support Self-efficacy x Role Modeling
F 8.030*** 76.579*** 63.981*** 54.025***
R2 .058 .442 .453 .457
Adj. R2 .051 .436 .446 .449
△R2 .058 .383 .011 .004
△F 8.030*** 265.788*** 8.026** 2.774
Note. *p < .05, **p < .01, ***p < .001
(continued)
52
Table 4.2. (continued)
β
In-role Performance
Model 5 6 7 8
Control Variable
Gender -.240*** -.237*** -.239*** -.230***
First Job .001 .002 -.014 -.008
Independent variable
Self-efficacy .638*** .631*** .579*** .561***
Moderator
Career Development
Psychosocial Support -.037 -.034
Role Modeling .087* .085*
Interaction
Self-efficacy x Career Development
Self-efficacy x Psychosocial Support -.030
Self-efficacy x Role Modeling -.070*
F 61.390*** 51.201*** 62.470*** 52.853***
R2 .443 .444 .447 .452
Adj. R2 .436 .435 .440 .443
△R2 .001 .001 .005 .004
△F .795 .586 3.808* 3.083*
Note. *p < .05, **p < .01, ***p < .001
53
Figure 4.1. Plotting interaction effect _ role modeling
The moderating effect of role modeling is illustrated in Figure 4.1. According to the figure, for all three levels of role modeling, self-efficacy and in-role performance maintain positive relationship.There are no difference in the slopes between medium and high levels of role modeling. However, for low level of role modeling the effect of self-efficacy on in-role performance is much stronger.
Even though all three levels of role modeling show positive relationship between efficacy and in-role performance, it seems there are less differences for high self-efficacy employees. It could be inferred that role modeling has smaller effect on improving the in-role performance of high self-efficacy employees. By constrast, for early-career employees who have low self-efficacy, the groups with higher and medium role modeling have higher in-role performance level. The interaction diagram shows a cross when self-efficacy is in the range of 4.5 to 5.5. It can be interpreted that when an
Role Modeling
54
employee’s self-efficacy is lower than 4.5, having a mentor as a role model can help the employee raise the in-role performance. That is, assigning a mentor to serve as a role model can potentially be a good solution to enhance the in-role performance of those early-career employees who seem to have lower self-efficacy.
55
Relationship between Helping and Mentoring Functions
Table 4.3. displays that self-efficacy has positive relationship with helping (β
= .522, p < .001). It means that those with higher efficacy are more likely to help others.
Hence, Hypothesis 2 is supported. According to Model 3, it shows that career development (β = .248, p < .001) and role modeling (β = .253, p < .001) have significant direct effect, which supports Hypothesis 8a, 8c. Both of the moderators showed positive direct effect, which can be deduced that mentoring enhances protégés’ helping behavior in the workplace. However, Hypothesis 8b does not get support. The psychosocial support mentoring function does not have effect on helping behavior (β = .056, p > .05).
Model 4 (β = -.004, n.s.), 6 (β = .010, n.s.) and 8 (β = -.037, n.s.) show no effect of the interaction terms, thus Hypothesis 5a, 5b, 5c are rejected. The self-efficacy shows too strong an effect on helping behavior, hence, the moderating effect cannot be found.
The result of R square illustrates that self-efficacy has a strong effect on the explanation of the outcome variable. In addition, the change of R square shows that self-efficacy is an important variable in predicting the helping behavior (△R2 = .270,
△F = 153.342, p < .001). Furthermore, the interaction terms do not add to the explanation of the outcome variable. In other words, the moderators do not moderate the relationship between self-efficacy and helping behavior.
56
Table 4.3.
Results of Hierarchical Regression Analysis – Helping β Helping
Model 1 2 3 4
Control Variable
Gender -.188*** -.192*** -.182*** -.181***
First Job .052 .018 -.002 -.002
Independent variable
Self-efficacy .522*** .371*** .370***
Moderator
Career Development .248*** .247***
Psychosocial Support Role Modeling Interaction
Self-efficacy x Career Development -.004
Self-efficacy x Psychosocial Support Self-efficacy x Role Modeling
F 6.162*** 44.510*** 41.988*** 34.910***
R2 .045 .315 .352 .352
Adj. R2 .038 .308 .344 .342
△R2 .045 .270 .037 .000
△F 6.162*** 152.342*** 22.198*** .006
Note. *p <.05, **p <.01, ***p < .001
(continued)
57
Table 4.3. (continued)
β Helping
Model 5 6 7 8
Control Variable
Gender -.194*** -.195*** -.188*** -4.439***
First Job .010 .010 -.009 -.149
Independent variable
Self-efficacy .500*** .502*** .393*** .392***
Moderator
Career Development
Psychosocial Support .056 .055
Role Modeling .253*** .252***
Interaction
Self-efficacy x Career Development
Self-efficacy x Psychosocial Support .010
Self-efficacy x Role Modeling -.037
F 35.943*** 29.888*** 43.669*** 36.497***
R2 .318 .318 .361 .363
Adj. R2 .309 .307 .353 .353
△R2 .003 .000 .046 .001
△F 1.463 .054 27.919*** .767
Note. *p < .05, **p < .01, ***p < .001
58
Relationship between Voice and Mentoring Functions
Almost all of the control variables do not show the influence on voice behavior.
Only first job experience has a slight effect on voice. It can be said that there is no significant effect of gender and first job on voice.
Self-efficacy no doubt has a significant impact on voice. According to Table 4.4., self-efficacy shows a direct positive effect (β = .569, p < .001). Hence, Hypothesis 3 is supported. Model 3 and 5 show that career development (β = .135, p < .05) and psychosocial support (β = .116, p < .05) have an influence on voice behavior, therefore, Hypothesis 9a, 9b are supported. It could be inferred that when mentors pay more attention on protégés’ career, the protégés would be more likely to give advices to the organization. Moreover, once protégés spend more time with the mentor in social function, they are more likely to speak out loud some suggestions and comments to the organization. Instead, Hypothesis 9c is rejected. The role modeling mentoring does not directly affect voice behavior.
Even though the moderators have direct effects on voice, the moderating effect is not found in Model 4 (β = .001, n.s.), 6 (β = -.007, n.s.) and 8 (β = -.028, n.s.). As a result, Hypothesis H6a, H6b, H6c are rejected. It appears the direct effect of self-efficacy and the moderator variables overpower the moderating effect on voice.
The result of self-efficacy’s R square change increases 32% in the variances of voice being explained (△R2 = .319, △F = 185.690, p < .001). By contrast, the interaction terms add no significant explanation of the variances. In sum, the moderators do not moderate the relationship between self-efficacy and voice.
59
Table 4.4.
Results of Hierarchical Regression Analysis – Voice β Voice
Model 1 2 3 4
Control Variable
Gender -.045 -.050 -.044 -.044
First Job .106* .069 .058 .058
Independent variable
Self-efficacy .569*** .486*** .486***
Moderator
Career Development .135* .136*
Psychosocial Support Role Modeling Interaction
Self-efficacy x Career Development .001
Self-efficacy x Psychosocial Support Self-efficacy x Role Modeling
F 1.931 46.560*** 40.723*** 33.848***
R2 .015 .334 .345 .345
Adj. R2 .007 .327 .337 .335
△R2 .015 .319 .011 .000
△F 1.931 185.690*** 6.575* .001
Note. *p < .05, **p < .01, ***p < .001
(continued)
60
Table 4.4. (continued)
β Voice
Model 5 6 7 8
Control Variable
Gender -.054 -.053 -.049 -.045
First Job .053 .053 .063 .066
Independent variable
Self-efficacy .523*** .521*** .541*** .534***
Moderator
Career Development
Psychosocial Support .116* .117*
Role Modeling .054 .053
Interaction
Self-efficacy x Career Development
Self-efficacy x Psychosocial Support -.007
Self-efficacy x Role Modeling -.028
F 40.699*** 33.835*** 39.115*** 32.618***
R2 .345 .345 .336 .337
Adj. R2 .337 .335 .328 .327
△R2 .011 .000 .002 .001
△F 6.497* .026 1.223 .423
Note. *p < .05, **p < .01, ***p < .00
61
Summary of Analysis Results
Overall, all the hypotheses were tested via hierarchical regression. Refer to Table 4.5, which summarizes the result of the hypothesis testing.
Firstly, self-efficacy has significantly positive effect on in-role performance, helping and voice (H1, H2, H3 are supported). It supported the literature review which has been proved by previous researches. This research assumed that different dimensions of mentoring function would have different moderating effect on individual’s performance.
Nevertheless, most of the hypothesis of moderating effect are rejected due to the strong relevance of self-efficacy and other variables. However, the different dimensions of mentoring do show significantly different effect.
Second, some of the moderators have direct relationships with the dependent variables.
Specifically, the career development and role modeling mentoring at workplace have direct positive effects on in-role performance and helping behavior (H7a, H7c, H8a, and H8c are supported). The career development and psychosocial support mentoring functions show positive effect on voice (H9a, H9b are supported). These indicate that mentoring may potentially be a method for enhancing individual’s performance.
Third, the R square change value from hierarchical regression illustrates that the interaction of independent and moderator variables only have weak or no influence on the dependent variables (H5, H6 are all rejected). It can be inferred that there are other important factors which may affect individual’s performance. However, role modeling does show a moderating effect on in-role performance. Role modeling mentoring function weakens the positive relationship between self-efficacy and in-role performance (H4c is supported). It can be interpreted that with a role model, protégés can still have satisfactory in-role performance without high self-efficacy. Role modeling could help protégés enhance their in-role performance, especially when they are people with lower esteem.
Moreover, gender as a control variable has a significant effect on in-role performance
62
and helping behavior; it can be inferred that female employees have better in-role performance and are more willing to help others than male employees.
Table 4.5.
Hypothesis Testing Results Summary
Hypothesis Results
H1: Self-efficacy has a positive relationship with in-role performance. Supported
H2: Self-efficacy has a positive relationship with helping. Supported
H3: Self-efficacy has a positive relationship with voice. Supported H4a: Career development mentoring function will moderate the positive
relationship between self-efficacy and in-role performance. Rejected H4b: Psychosocial support mentoring functionwill moderate the
positive relationship between self-efficacy and in-role performance.
Rejected
H4c: Role modeling mentoring functionwill moderate the positive
relationship between self-efficacy and in-role performance. Supported H5a: Career development mentoring function will moderate the positive
relationship between self-efficacy and helping behavior. Rejected H5b: Psychosocial support mentoring functionwill moderate the
positive relationship between self-efficacy and helping behavior. Rejected H5c: Role modeling mentoring functionwill moderate the positive
relationship between self-efficacy and helping behavior. Rejected H6a: Career development mentoring function will moderate the positive
relationship between self-efficacy and voice behavior. Rejected H6b: Psychosocial support mentoring functionwill moderate the
positive relationship between self-efficacy and voice behavior. Rejected H6c: Role modeling mentoring functionwill moderate the positive
relationship between self-efficacy and voice behavior. Rejected (continued)
63
Table 4.5. (continued)
Hypothesis Results
H7a: Career development mentoring function has a positive relationship
with in-role performance. Supported
H7b: Psychosocial support mentoring functionhas a positive
relationship with in-role performance. Rejected
H7c: Role modeling mentoring functionhas a positive relationship with
in-role performance. Supported
H8a: Career development mentoring function has a positive relationship
with helping. Supported
H8b: Psychosocial support mentoring functionhas a positive
relationship with helping. Rejected
H8c: Role modeling mentoring functionhas a positive relationship with
helping. Supported
H9a: Career development mentoring function has a positive relationship
with voice. Supported
H9b: Psychosocial support mentoring functionhas a positive
relationship with voice. Supported
H9c: Role modeling mentoring functionhas a positive relationship with
voice. Rejected
64