This chapter presents the findings obtained after the analysis of the data collected to study the relationship between the variables: Leader-Member Exchange, Employee Voice Behavior, Psychological Safety, and Gender. This chapter is divided into four sections. The first section includes the results of the Descriptive Analysis carried out, the second section presents the results of the Pearson’s Correlation Analysis, the third section shows the validity of the study presenting the results from the Confirmatory Factor Analysis, and the fourth section presents the findings from the Hierarchical Regression Analysis focused on hypotheses testing.
Descriptive Statistics
The demographic information included were gender, age, level of education, and job tenure.
The sample of this study consisted of 100 male participants (50%) and 100 female participants (50%). The majority of participants fell in the categories of 18-24 years of age (67%) and 25-34 years of age (80%). 160 of the respondents (80%) reported to have a bachelor’s degree as their highest level of education and 29 respondents (14.5%) reported to have a master’s degree as their highest level of education. Regarding their job tenure in their current organization, 72 participants (36%) reported to have a job tenure of between 1-2 years and 60 participants (30%) reported to have a job tenure of 5 or more than 5 years. The frequency and percentage of the demographic information are presented in Table 4.1.
32 Table 4.1
Descriptive Statistics (N=200)
Item Frequency Percentage
1. Gender
Pearson’s Correlation Analysis was used to understand the relationship between two variables. Table 4.2 presents the means, standards deviations, correlations among the variables, and the Cronbach’s Alpha value of each variable.
All the variables in this study showed significant correlation coefficients. Leader-Member Exchange was positively correlated with Employee Voice Behavior (r= .65, p<.01), and Psychological Safety (r= .76, p<.01). And, Employee Voice Behavior was positively correlated with Psychological Safety (r= .59, p<.01). The Cronbach’s Alpha value for all variables is above
33
0.70. Leader-Member Exchange with a Cronbach’s Alpha of .89, Employee Voice Behavior with .92, and Psychological Safety with .93.
Table 4.2
Mean, Standard Deviation, Correlation and Reliability (N=200)
Variable Mean S.D. 1 2 3
1.Leader-Member Exchange
3.88 0.77 (.89)
2.Employee Voice Behavior
3.63 0.89 .65** (.92)
3.Psychological Safety
5.31 0.98 .76** .59** (.93)
Note. N= 200; **p <.01.; Number in parentheses represents Cronbach’s alpha value.
Confirmatory Factor Analysis
Before testing all of the hypotheses, confirmatory factor analysis was carried out by running the data in AMOS 24.0. The main objective of this test is to evaluate the validity of the data and the fit of such to a previous hypothesized measurement model.
This study received 200 valid responses. As chi-square (x2) test has proven to be influenced by the size of the samples in different studies, other fit indices were also used for evaluating the model fit. Including the ratio of chi-square to degrees of freedom (x2/df), Comparative Fit Index (CFI), Normed Fit Index (NFI), Incremental Fit Index (IFI), Tucker-Lewis Index (TLI), Goodness of Fitness Index (GFI), and the Root Mean Square of Approximation (RMSEA).
A good model fit should follow the next: x2 must be less than .05, x2/df between 2-5, CFI, NFI, IFI, and TLI greater than .90, GFI greater than .80, and RMSEA less than .08. Table 4.3 provides a summary of the thresholds for a good fit model.
34 Table 4.3
Summary of Fit Indexes
Fit Index Threshold Reference
x2 >.05 x2/df 2-5 CFI >.90
NFI >.90 Hooper, Coughlan, & Mullen, 2008 IFI >.90
TLI >.90 GFI
RMSEA
>.80 <.08
Carmines & McIver, 1981
Brown & Cudeck, 1993; MacCallum, Brown, & Sugawara,1996
Originally, the number of observed items for Leader-Member Exchange was 7, Employee Voice Behavior was 7, and Psychological Safety was 21. However, after running AMOS, CFA results showed a poor model fit due to the number of items in each variable. Scholars have argued that item parcel is a successful way to handle and improve a poor model fit reducing the measurement error and also increasing the stability of the parameter estimates (Bagozzi &
Edwards, 1998).
Therefore, the original 7 items of Leader-Member Exchange were divided into 3 parcels, Employee Voice Behavior 7 items were divided into 3 parcels, and Psychological Safety 21 items were divided into 7 parcels. After conducting parceling, CFA results reported a better model fit than the first time. The validity of the full model was tested. The three-factor model presented acceptable fit indexes. Refer to Table 4.4, and Figure 4.1 present the CFA results of the model fit and the research CFA Measurement Model, respectively.
Table 4.4
Factor Measurement Model Fit Summary (N=200)
x2 Df x2∕df CFI NFI IFI TLI GFI RMSEA
215.69 62 3.48 .91 .87 .92 .89 .86 .11
35 Figure 4.1. Research CFA measure model
36
Hierarchical Regression Analysis
Hierarchical Regression Analysis was conducted on all three hypotheses of this study. It was used to analyze and test the relationship of Leader-Member Exchange on Employee Voice Behavior (H1), to examine the mediating effect of Psychological Safety between Leader-Member Exchange and Employee Voice Behavior (H2), and to study the moderating effect of Gender in the relationship between Leader-Member Exchange and Employee Voice Behavior (H3). For this study, Baron and Kenny’s (1986) three-step approach was followed. For the first block control variables were entered before any regression analysis was carried out: Age, level of education, and job tenure.
Hypothesis 1
Table 4.5 presents the effect of Leader-Member Exchange on Employee Voice Behavior. In model 1, only the demographic control variables were entered in the regression equation with Employee Voice Behavior as the dependent variable. For model 2, Leader-Member Exchange was also entered into the regression equation with Employee Voice Behavior. Model 2 shows a significant R2 change of .37 and a positive effect of Leader-Member Exchange on Employee Voice Behavior (β = .63, p< .001). Therefore, H1 is supported, Leader-Member Exchange is related to Employee Voice Behavior. It is imperative to emphasize that Leader-Member Exchange combined with the control variables explain 42% of the variances in Employee Voice Behavior.
37 Table 4.5
Effect of Leader-Member Exchange on Employee Voice Behavior
Employee Voice Behavior
Model 1 Model 2
Step 1: Controls
Age .02 .01 Level of Education .09 .01 Job Tenure .17 .03 Step 2: Main Effect
Leader-Member Exchange .63***
R2 .05 .42
Adj. R2 .04 .41
ΔR2 .05 .37
F 3.51** 35.12***
N 200 200
Note. Standardized regression coefficients are shown, **p< .01. ***p< .001.
Hypothesis 2
A similar process was followed to test H2. Hierarchical regression analysis was carried out to test H2: Psychological Safety mediates the relationship between Leader-Member Exchange and Employee Voice Behavior. In this analysis, Employee Voice Behavior was regressed on a set of control variables (age, level of education, and job tenure), Leader-Member Exchange, and Psychological Safety in such order.
According to Baron and Kenny (1986), three factors must be met to have a full mediation effect:
1. The independent variable and the dependent variable must have a significant relationship.
2. The independent variable and the mediator must have a significant relationship.
3. Original beta value must be smaller than beta value when the independent variable and mediator are entered into the model.
Following these steps, it can be said that for H2 the first two factors were met. Table 4.6 shows that Leader-Member Exchange has a significant relationship with Employee Voice Behavior (β = .63, p< .001). Table 4.7 shows that Leader-Member Exchange and Psychological Safety have a
38
significant relationship (β = .78, p< .001). However, when it comes to the third factor, the original beta value was not smaller than the beta value when the independent variable and mediator were entered into the model.
Table 4.6 presents all three models on the regression of control variables, Leader-Member Exchange, and Psychological Safety on Employee Voice Behavior. After conducting the analysis is can be seen that the F value for all three models is significant, however it is important to notice that model 2 and 3 have higher F values, model 2 F= 35.12, and model 3 F= 30.27 both p < .001.
Also, it is shown that for model 2 and 3 R2 and ΔR2 present higher values (model 2 R2 =.42, ΔR2= .37; model 3 R2 =.44, ΔR2= .39) with model 3 presenting the highest value of the three models. The value of the R2 represents the explaining power for each model.
Regarding Beta values, for Leader-Member Exchange in model 2 shows β = .63, p< .001, while for Psychological Safety in model 3 shows β = .23, p< .001. It is paramount to notice that in model 2 when only control variable and Leader-Member Exchange were entered into the analysis, beta value for Leader-Member Exchange was higher (β = .63, p< .001) than in model 3 when control variables, Leader-Member Exchange, and Psychological Safety were entered (β = .46, p<
.001). This explains that Psychological Safety has not a full mediation effect, but a partial mediation effect in the relationship between Leader-Member Exchange and Employee Voice Behavior. In model 3 beta value for Leader-Member Exchange dropped, however the value still remained significant.
39 Table 4.6
Mediating Effect of Psychological Safety on the Relationship between Leader-Member Exchange and Employee Voice Behavior (N=200)
Employee Voice Behavior
Model 1 2 3
Age .02 .01 .03
Education .09 .01 -.00
Job Tenure .17 .03 .25
Leader-Member Exchange .63*** .46***
Psychological Safety .23***
R2 .05 .42 .44
Adj. R2 .04 .41 .42
ΔR2 .05 .37 .39
𝐹 3.51** 35.12*** 30.27***
N 200 200 200
Note. Standardized regression coefficients are shown; **p <.01. ***p <.001.
Table 4.7
Effect of Leader-Member Exchange on Psychological Safety
Psychological Safety
Model 1 Model 2
Step 1: Controls
Age -.06 -.09 Level of Education .18 .05 Job Tenure .22 .04 Step 2: Main Effect
Leader-Member Exchange .78***
R2 .07 .62
Adj. R2 .05 .61
ΔR2 .07 .54
F 3.89** 58.53***
N 200 200
Note. Standardized regression coefficients are shown, **p< .01. ***p< .001.
40
Hypothesis 3
H3 proposed: Gender moderates the relationship between Leader-Member Exchange and Employee Voice Behavior such that this relationship is stronger when employees are male than when they are female. The regression results in Table 4.8show that in model 3, once the interaction term was added to the analysis, it did not add a significant incremental variance for the model (R2 change = .00, p>.05), which suggests that Gender does not have a moderating effect on the relationship between Leader-Member Exchange and Employee Voice Behavior. Therefore, H3 is not supported.
Table 4.8
Results for Regression Analysis for Moderating Effect of Gender (N=200)
Employee Voice Behavior
Model 1 2 3
Step 1: Controls
Age .05 .02 .03
Education .14 .05 .05
Job Tenure .16 .02 .00
Step 2: Main Effect
Leader-Member Exchange .59*** .60***
Gender -.06 -.06
Step 3: Interaction term
Leader-Member Exchange x Gender .05
R2 .07 .38 .38
Adj. R2 .05 .35 .35
ΔR2 .07 .31 .00
𝐹 3.49** 17.36*** 14.55***
N 200 200 200
Note. Dependent variable = Employee Voice Behavior; **p <.01. ***p <.001.
41
Finding Summary and Discussion
Table 4.9
Research Hypotheses Results
Hypotheses Results
Hypothesis 1 Leader-Member Exchange is related to Employee Voice Behavior.
Supported
Hypothesis 2 Psychological Safety mediates the relationship between Leader-member Exchange and Employee Voice Behavior.
Partially Supported
Hypothesis 3 Gender moderates the relationship between Leader-Member Exchange and Employee Voice Behavior such that this relationship is stronger when employees are male than when they are female.
Not Supported
The present study was conducted to analyze the relationship among the following variables:
Leader-Member Exchange, Employee Voice Behavior, Psychological Safety, and Gender. After conducting the proper analyses, it was found that Leader-Member Exchange, in fact, has a positive relationship with Employee Voice Behavior. Employees who are able to build stronger relationships with their leaders are more likely to engage in Employee Voice Behavior (Ilies, Nahrgang, & Morgeson, 2007). Such findings agree with the rationale of Blau’s (1964) social exchange theory explaining that whenever an individual receives beneficial treatments from another individual, that specific individual will feel the need to behave in the same manner in return. Following this theory, relating it to the workplace, whenever employees are able to build high LMX relationships with their leaders, relationships in which employees receive support and resources from leaders, they will be more likely to return those beneficial treatments with favorable and beneficial work-related behaviors such as voicing out ideas, issues, thought, or comments that may improve their job, their leader’s job, or the organization’s overall performance. This explains why this study was able to find a positive relationship between Leader-Member Exchange and Employee Voice Behavior.
This study also found that Psychological Safety partially mediates the positive relationship between Leader-Member Exchange and Employee Voice Behavior. This finding is compatible with previous studies stating that those employees who build high-quality relationships with their
42
leaders will feel more psychologically safe to perform and engage in work-related behaviors (Moss, Sanchez, Brumbaugh, & Borkowski, 2009; Van Dyne, Kamdar, & Joireman, 2008). In addition, this finding contributes to the understanding of the existent literature that Psychological Safety can act as a mediator between Leader-Member Exchange and a number of in-role behaviors such as turnover intention and innovation (Agarwal, Datta, Blake-Beard, & Bhargava, 2012), but also act as a mediator in the relationship between Leader-Member Exchange and extra-role behaviors such as Employee Voice Behavior.
Lastly, the results of this specific study showed that Gender does not moderate the relationship between Leader-Member Exchange and Employee Voice Behavior. Surprisingly, this finding is not consistent with previous studies that argue men and women behave differently when engaging in Employee Voice Behavior (Kidder, 2002; Wang, Weng, McElroy, Ashkanasv, &
Lievens, 2014). One of the main reasons for the incongruence of the finding might be the size of the sample in this study. Due to limited resources and a lower response from participants, this study only collected 200 valid responses. If a sample size is too small it may not be able to identify the relationship among variables and their moderating effects, consequently providing non-significant results (Hair, Ringle, & Sarstedt, 2011). Therefore, it is advised to test such moderating effect by obtaining a larger sample size. Refer to Table 4.9 for research hypotheses results.
43