The influence of employee voice behavior on affect : a two-wave panel study into the effects and moderators of the relationship between employee voice behavior and affect

(1)

The influence of employee voice behavior on affect

A two-wave panel study into the effects and moderators of the relationship between

employee voice behavior and affect

N.E.M. Hoogenboom

Master thesis Business Studies University of Amsterdam

(2)

Master thesis N.E.M. Hoogenboom 2

The influence of employee voice behavior on affect

A two-wave panel study into the effects and moderators of the relationship between employee voice behavior and affect

University

University: University of Amsterdam

Faculty: Faculty of Economics and Business

Study: Master Business Studies, track Leadership and Management Study year: 2013 – 2014

Author

Name: Nikki E.M. Hoogenboom Student no. 10669345

E-mail: n.e.m.hoogenboom@gmail.com

Supervisor

Name: Renske E. van Geffen, MSc

Faculty: Faculty of Economics and Business E-mail: r.e.vangeffen@uva.nl

(3)

Abstract

The goal of the present study is to find out how employee voice behavior affects the emotions of the individual and if this relationship is moderated by the personality traits extraversion and neuroticism, and group voice climate. Furthermore, a distinction is made between promotive voicing and prohibitive voicing to examine if the results differed for the different types of voice. The data was collected during team meetings, using both observations and surveys (N = 213). The results showed that promotive voicing is positively related to positive affect and that prohibitive voicing is negatively related to positive affect. Weak support was found for the moderating variables, although extraversion and voice climate efficacy did show several significant relationships. Extraversion attenuated the relationship between prohibitive voicing and negative affect. In the relationship between promotive voicing and negative affect, extraversion actually strengthened the relationship. Furthermore, voice climate efficacy seemed more important for an individual’s positive affect than for negative affect. Several explanation, implications and limitations are discussed. It is suggested that future research focuses more on the concept of employee voice behavior in general, as much remains unknown about the moderating and mediating effects.

Introduction

Employees possess valuable information that can be critical for an organization’s survival or functioning (Morrison, 2011). Through the use of “voice”, employees can help organizations innovate and adapt (Liang, Farh, & Farh, 2012), since top managers depend on information held by lower-level employees in decision-making (Morrison, 2011). Likewise, in work groups it is essential that employees deliver honest input in order for the group to work effectively (Morrison, 2011). The intentional sharing of work-related ideas, opinions and suggestions is called voicing (Van Dyne, Ang, & Botero, 2003). On the other hand, employees can also intentionally withhold work-related ideas, suggestions and opinions about organizational problems, which is called organizational or employee silence (Morrison & Milliken, 2000).

(5)

Master thesis N.E.M. Hoogenboom 5 Despite the potential significant importance of employee voice behavior in organizations, there is yet not much known about the topic. Voicing is a fairly new concept, partially derived from constructs that already received more research attention, such as upward communication (Jablin, 1979), whistle blowing (Dozier & Miceli, 1985; Near & Miceli, 1985, 1995), and issue selling (Dutton, Ashford, O’ Neill, Hayes, & Wierba, 1997; Dutton & Ashford, 1993). There are indeed similarities between these constructs and voice, but there are also remarkable differences. Upward communication for example is concerned with communication from lower levels to higher levels of hierarchy in the organization (Jablin, 1979; Morrison, 2011). Voice on the other hand concerns sharing work-related ideas, suggestions, and opinions (Van Dyne et al., 2003) and it is not specified if this is a form of communication from lower to higher level employees or between employees from the same level for example. Another related construct is whistle-blowing, which is primarily focused on employees who report practices that are illegal, immoral or illegitimate (Near & Miceli, 1985). Voice on the other hand is more concerned with constructive ideas, suggestions or opinions (Van Dyne et al., 2003), of which reporting illegal practices could be an example. Voicing can thus be described as a broader concept than whistle blowing. The last construct, issue selling, is more related to employee voice behavior than whistle blowing. However, issue selling is concerned with trying to attract management’s attention or the attention of someone higher in the organizational hierarchy to ‘sell’ issues (Dutton & Ashford, 1993), whereas voicing can be targeted at the management as well as at the colleagues in your team. In other words, voice is not primarily concerned with expressing ideas to higher-order employees in contrast to issue selling.

Voicing started to receive more research attention in the late 90s and the early 00s, with the publication of articles on predicting voice behavior in work groups (LePine & Van Dyne, 1998), the multidimensionality of employee silence and employee voice (Van Dyne et al., 2003), the meaning and purpose of employee voice (Dundon, Wilkinson, Marchington, & Ackers, 2004), and group voice climate (Morrison, Wheeler-Smith, & Kamdar, 2011) for example.

(6)

Master thesis N.E.M. Hoogenboom 6 Due to the relative novelty of the construct there are still some ambiguities surrounding voice. For example, a large part of the voice literature has focused on the construct of employee silence or the effect of leadership on voice (Detert & Treviño, 2010; Liu, Zhu, & Yang, 2010; Walumbwa & Schaubroeck, 2009). However, the research on the different types of voice, voice in work groups and the influence of the (voice) climate is still limited (Liang, Farh, & Farh, 2012; Morrison, 2011; Morrison et al., 2011; Tangirala & Ramanujam, 2008a). A possible explanation for this is that suggestions and concerns are traditionally communicated to employees higher in the organizational hierarchy. Upward communication, issue selling and whistle blowing are all aimed at expressing concerns or suggestions towards higher-order members of the organization. The different types of voice, the voice climate and voice in work groups are traditionally not considered in those types of studies.

Morrison (2011) states in a review of the existing literature on voice that future research should be directed at exploring, amongst others, group voice and the effect of voice on others. Another suggestion for future research is the exploration of different outcomes for different types of voice, something that is also suggested by Liang et al. (2012). Other research gaps concerning employee voice behavior are for example the (development of) voice climate (Morrison, 2011; Morrison et al., 2011; Tangirala & Ramanujam, 2008a) and the relationship between personality traits and voice (Janssen, Vries, & Cozijnsen, 1998; Klaas, Olson-Buchanan, & Ward, 2012). This study tries to answer some of the calls for future research by focusing on the effects of different types of voice within the context of work groups. Furthermore, this study tries to find out if this relationship is influenced by the group voice climate and certain personality traits of the members in the work group. Unique in this study is its design: a two-wave panel design in combination with the observation of voice actions in meetings. The dependent variable, affect, was measured at two points in time: before and after the meeting. This allows the researcher to make more accurate inferences about the effect a moderator has, since the change in the dependent variable can now be measured. This makes it possible to make causal inferences about the effect the

(7)

Master thesis N.E.M. Hoogenboom 7 independent variables have on the dependent variable. Furthermore, the observations in the meetings were used to observe the voice behavior and to count the voice actions.

The results of this study can be helpful for organizations in different ways. First, by researching different types of voice and how individuals in work groups react to these different types of voice it is possible to state if there should be a preference for a particular type of voice. This way, organizations can adjust their practices in work groups to benefit more fully from the ideas, suggestions and opinions from their employees. Secondly, by researching personality traits and voice it is possible to make a distinction between personality traits that can enhance or discourage employee voice behavior. Companies can take this into account when creating work groups or teams. And lastly, the results may indicate if there are differences between the favorability of the voice climate and the type of voicing behavior. The voice climate may then be adapted to an organization’s preference for a particular type of voice (prohibitive or promotive).

Literature review

Employee voice behavior

As stated before, the construct of employee voice behavior is fairly new. Although there related constructs, such as whistle-blowing, issue-selling, and upward communication, these are traditionally concerned with communication towards higher-level employees (Morrison, 2011). The origination of voice can be dated back to Hirschman's (1970) exit, voice, and loyalty framework. In this framework, voice, or more specifically voice option, is described as the possibility for the organization’s customers or employees to express their displeasure directly to the management or other authorities within the organization, or through protest in general (Hirschman, 1970). Employees and customers can also express their dissatisfaction by leaving the company or buying products from another firm, which is called the exit option (Hirschman, 1970). As the literature around voicing evolved, so did the definition of the construct. Whereas Hirschman’s (1970) definition also described the possibility for dissatisfied customers to express their opinion, more recent

(8)

Master thesis N.E.M. Hoogenboom 8 conceptualizations of voice focus more on the expressions of opinions and concerns between employees (Tangirala & Ramanujam, 2008b). In this research, voice is defined as the intentional expression of “work-related ideas, information and opinions about work-related improvement” (Van Dyne et al., 2003, p. 1370). Although there are several definitions available, most of them include the (intentional) expression of information, ideas or constructive challenge (Van Dyne et al., 2003; LePine & Van Dyne, 1998; Tangirala & Ramanujam, 2008b) with the intention to improve the current situation (Van Dyne et al., 2003; LePine & Van Dyne, 1998). The definition does not state to whom these ideas, information and opinions are directed. The voicing behavior of employees can thus be directed towards anyone: supervisors, subordinates, group members, and so forth.

In the last decades, the focus in employee voice behavior concentrated more and more on work groups or teams (LePine & Van Dyne, 1998). Work groups or teams are becoming more important in organizations due to their flexibility. An important feature in these work groups are the shared responsibilities as a team instead of an individual responsibility (Rumsey, Walker, & Harris, 1994). Voicing then becomes an important aspect of a team: communicating ideas, suggestions and opinions is crucial when sharing responsibility (LePine & Van Dyne, 1998). However, the literature on voice in work groups and the effect this has on individuals is still limited. For example, it is still unknown if these effects are positive, negative or neutral. Next to that, there is a need for a better understanding of the outcomes of (different types of) voice (Morrison, 2011).

Before starting to discuss the possible outcomes of voice, the focus is first on the different types of voice. For example, Van Dyne et al. (2003) distinguish between defensive voice, acquiescent voice and prosocial voice. Another distinction is the one between suggestion-focused, opinion-focused and problem-focused voice (Morrison, 2011). Furthermore, voice can also be subdivided into prohibitive and promotive voice (Liang, Farh, & Farh, 2012). To differentiate between suggestion-, opinion-, and problem-focused voice can cause problems because of the relatedness of the three constructs. It may not always be clear when voicing can be regarded as expressing an opinion, a problem, or a suggestion. It is

(9)

Master thesis N.E.M. Hoogenboom 9 believed that the distinction between acquiescent, defensive and prosocial voice can also cause some equivocation. The distinction between these three types of voice is mostly based on the underlying motivation (resignation, fear or to benefit others), which may be hard to observe and is therefore hard to use in practice. Therefore, the distinction between prohibitive and promotive voicing as described by Liang et al. (2012) is seen as less problematic than the other two distinctions.

Promotive voicing can be defined as “employees’ expression of new ideas or suggestions for improving the overall functioning of their work unit or organization” (Liang et al., 2012, p. 74). This type of voicing is somewhat future-oriented; it is focused on an ideal (future) state. Prohibitive voicing on the other hand is defined as “employees’ expression of concern about work practices, incidents, or employee behavior that are harmful for the organization” (Liang et al., 2012, p. 75) and is hence more concerned with preventing possible problematic situations, either in the future or the past. These different types of voice share some commonalities; they are, for example, both aimed at helping the organization. However, although the good intentions behind promotive voicing may be easily recognized, the good intentions behind prohibitive voicing are not that easily recognized and may even be regarded as a negative or defensive emotion (Liang et al., 2012).

Affect

Because this study researches the effect voice has on individuals in work groups, the concept of affect is used to measure this. Affect can be defined as a set of dimensions that describe an individual’s psychological state (Remington, Fabrigar, & Visser, 2000; Russell, 1980). According to Russell (1980) affect can be represented in a spatial model, which is called the circumplex model of affect. Emotions are judged along the continuum of positive-negative and active-passive (or high vs. low arousal) (Remington et al., 2000; Russell, 1980). Although the circumplex model of affect has been introduced in 1980, it is still seen as a reasonable representation of perceived emotions. However, the model shows some variations. Despite these limitations, the circumplex model of affect is seen as a useful representation of affect (Remington et al., 2000).

(10)

Master thesis N.E.M. Hoogenboom 10 Generally, eight affective states or concepts are defined, namely arousal, excitement, pleasure, and contentment (all describing a positive affect state) and sleepiness, depressions, misery, and distress (negative affect states) (Russell, 1980).

Carver & Scheier (1990) describe the origins of affect based on the concepts of self-regulation and discrepancy reduction. Self-self-regulation holds that when people are moved towards goals, their behavior along the way is adjusted based on a feedback loop. As such, they compare their intended actions and their actual actions and, when necessary, adjust their behavior accordingly. Then, a process of discrepancy reduction follows in which the individuals try to minimize the discrepancy between their intended and actual actions. When an individual experiences no discrepancy, affect is expected to be neutral. However, when a negative discrepancy is experienced between the actual and intended actions, negative affect is created. The opposite is true for positive affect; a positive discrepancy between intended and actual actions leads to positive affect (Carver & Scheier, 1990). When the concepts self-regulation and discrepancy reduction are applied to employee voice behavior, the following can be expected: an employee voices for a certain reason, which can be seen as the intended voice action. It then receives a reaction from the group, based on which a perception of the actual voice action is created within the voicing individual. The discrepancy is then assessed and affect is created. This may implicate that employees who do not receive an expected reaction from their co-worker(s) may experience negative affect. Positive affect on the other hand is then experienced when an employee receives a better reaction than expected from their co-worker(s).

Employee voice behavior and affect

Although this study does not examine the perceived discrepancies between intended and actual (re)actions, Carver & Scheier’s (1990) theory of the origination of affect is still seen as useful, especially in combination with Liang et al.’s (2012) distinction between promotive and prohibitive voicing. Liang et al. (2012) state that the good intentions behind promotive voicing are more easily recognized and usually interpreted as positive. Therefore, the expected discrepancy is expected to be neutral or positive, and thus it is more likely that

(11)

Master thesis N.E.M. Hoogenboom 11 positive affect is created. The opposite is expected for prohibitive voicing, where the good intentions may not always be recognized and may even cause defensiveness (Liang et al., 2012). As this increases the risk of experiencing a negative discrepancy, it is therefore expected that prohibitive voicing may be more related to negative affect. This leads to the following hypotheses:

H1a: Promotive voicing is more positively related to positive affect than prohibitive voicing.

H1b: Prohibitive voicing is more positively related to negative affect than promotive voicing.

Personality

Another variable that is taken into account in this research is personality. One of the most common methods of describing the concepts of personality is the five-factor model. Although there have been some ambiguities surrounding the ideal number of dimensions, there now appears to be a fairly good agreement about the number of factors (Digman, 1990). One of the best-known five-factor models is the one depicted by Norman (1963). Although he was certainly not the first to come up with a five-factor structure, it did receive more research than other conceptualizations of the five personality dimensions (Digman, 1990). Norman (1963) described the five factors of personality as extraversion, agreeableness, conscientiousness, emotional stability, and culture. More recently however, the five dimensions are described as extraversion, agreeableness, conscientiousness, neuroticism, and openness (or openness to experience) (LePine & Van Dyne, 2001). Emotional stability has been replaced by neuroticism and culture by openness. Although the names of the constructs are different, the content is still the same.

LePine & Van Dyne (2001) investigated the relationship between personality traits and employee voice behavior. They found that extraversion was the personality trait that best predicted an individual’s voice behavior. The traits neuroticism and agreeableness on

(12)

Master thesis N.E.M. Hoogenboom 12 the other hand are negatively associated with voice behavior, meaning that individuals who are neurotic or tend to agree easily do not engage in voice behavior as much (LePine & Van Dyne, 2001). In other words, extraversion promotes voice behavior, whereas neuroticism and agreeableness are associated with less active forms of voice behavior. The strongest relationships between personality traits and voice behavior were found for extraversion (positive) and agreeableness (negative) (LePine & Van Dyne, 2001).

Personality and affect

Multiple studies have investigated the relatedness of personality traits to affect (Costa & McCrae, 1980; Larsen & Ketelaar, 1989; Remington et al., 2000; Rusting & Larsen, 1997). Especially the personality traits extraversion and neuroticism have been studied in relation to affect (Costa & McCrae, 1980; Rusting & Larsen, 1997). Individuals who score high on extraversion are more predisposed towards a positive affect, whereas individuals who score high on neuroticism are more susceptible towards experiencing negative affect (Costa & McCrae, 1980; Rusting & Larsen, 1995, 1997). Even when controlled for the prior mood of the individual, extraversion is still related to positive mood susceptibility and neuroticism to negative mood susceptibility (Rusting & Larsen, 1997). To summarize, extraversion and agreeableness show the strongest relationships with employee voice behavior (LePine & Van Dyne, 2001), whereas extraversion and neuroticism are strongly related to affect (Rusting & Larsen, 1997). Based on these results, personality is expected to moderate the relationship between voice behavior and affect. Because affect is the dependent variable in this research, the choice is made to include extraversion and neuroticism as the moderators since these traits show the strongest relationship with affect. More specifically, extraversion is expected to strengthen the relationship between voicing and positive affect. The effect will be stronger for promotive voicing than for prohibitive voicing. Neuroticism on the other hand is expected to strengthen the relationship between voicing and negative affect. Here, the effect is expected to be stronger for prohibitive voicing than for promotive voicing. Thus, the second hypothesis states:

(13)

Master thesis N.E.M. Hoogenboom 13 H2a: Extraversion will strengthen the relationship between voicing and positive affect.

The effect is expected to be stronger for promotive voicing than for prohibitive voicing.

H2b: Neuroticism will strengthen the relationship between voicing and negative affect. The effect is expected to be stronger for prohibitive voicing than for promotive voicing.

Voice climate

Another factor that could influence the voice behavior of employees is the voice climate of the organization or the group. Organizational climate can be defined as “the shared perceptions of and the meaning attached to policies, practices, and procedures employees experience and the behaviors they observe getting rewarded and that are supported and expected” (Schneider, Ehrhart, & Macey, 2013, p. 362). The literature on organizational climate originated in the late 1960s and focused primarily on individual level differences in perceptions of the organizational climate. More recently however, the focus shifted towards facet-specific climates instead of organizational climates as a whole (Kuenzi & Schminke, 2009), such as an innovation climate (Ahmed, 1998; Ekvall, 1996), a safety climate (Neal, Griffin, & Hart, 2000; Zohar, 1980) or a voice climate (Morrison et al., 2011). Morrison et al. (2011) were one of the first to research voice climate at the work group level. Although Morrison & Milliken (2000) already described the existence of organizational voice climates, little research has since then focused on studying the voice climate (Frazier & Bowler, 2012). Next to that, the focus of Morrison et al. (2011) is specifically on group voice climates, whereas Morrison & Milliken (2000) focused more on a climate of silence. In line with the definition of organizational climate, Morrison et al. (2011) defined the more precise construct of group voice climate as “the shared beliefs about speaking up on voice behavior within work groups” (Morrison et al., 2011, p. 184).

According to Morrison et al. (2011) group voice climate has two dimensions. The first dimension describes the voice safety beliefs: do the members of the group believe that speaking up is safe or dangerous? This dimension specifically focuses on the perceived

(14)

Master thesis N.E.M. Hoogenboom 14 psychological safety of speaking up. The second dimension concerns the group voice efficacy. Group voice efficacy refers to the “shared belief about whether group members are able to voice effectively” (Morrison et al., 2011, p. 184). Morrison et al. (2011) found that group members indeed form a shared belief about voice safety and voice efficacy that goes beyond individual level differences. Furthermore, a favorable group voice climate positively influences employee voice behavior, such that an employee’s voice efficacy is higher when the group voice climate is favorable. This effect goes beyond the individual-level differences of satisfaction and identification with the group (Frazier & Bowler, 2012; Morrison et al., 2011).

Voice climate and affect

Employees are thus more likely to voice when the group voice climate is favorable compared to when the group voice climate is unfavorable (Morrison et al., 2011). It is thought that when the group voice climate is favorable, the employees are less susceptible to the reactions of the team members on their voicing behavior (Morrison et al., 2011). Following the same logic, they are also thought to be less susceptible towards experiencing negative affect following voicing. On the other hand, when the group voice climate is unfavorable, the employees are thought to be more susceptible towards the reactions of the team members on their voicing behavior and hence more susceptible towards experiencing negative affect. As such, the third hypothesis states:

H3a: When the voice climate is favorable, employees are less susceptible towards experiencing negative affect and more susceptible towards experiencing positive affect. This relationship is stronger for promotive voicing.

H3b: When the voice climate is unfavorable, employees are more susceptible towards experiencing negative affect and less susceptible towards experiencing positive affect. This relationship is stronger for prohibitive voicing.

(15)

Master thesis N.E.M. Hoogenboom 15 Based on the hypotheses, the following conceptual models are proposed:

Voice Promotive Prohibitive Personality (extraversion, neuroticism) Affect Positive Negative

Figure 1. Conceptual model (personality as moderator)

Voice Promotive Prohibitive Group voice climate Affect Positive Negative

Figure 2. Conceptual model (voice climate as moderator)

Research questions

The focus in this study is primarily on employee voice behavior and affect in work groups. Different types of employee voice behavior have been distinguished, namely prohibitive and promotive voice (Liang et al., 2012). For the concept of affect, the distinction has been made between positive and negative affect, even though there are numerous other affective states available (Russell, 1980; Watson, Clark, & Tellegen, 1988).

Prohibitive voicing is experienced as a more delicate form of voice behavior than promotive voicing (Liang et al., 2012). Prohibitive voice shows some similarities with the sensitive concept of whistle-blowing for example. This perceived sensitivity of the subject may make individuals experience negative affect faster when the actual reactions from co-workers are not as expected than is the case with promotive voicing (Carver & Scheier, 1990;

(16)

Master thesis N.E.M. Hoogenboom 16 Liang et al., 2012). On the other hand, it should be taken into account that personality traits do influence an individual’s susceptibility towards positive or negative affect and that this may bias the results.

Another factor that could influence affect is the voice climate. When the voice climate is favorable (the shared beliefs about voice safety and voice efficacy are high), individuals in the work group may be less susceptible towards experiencing negative affect. The other way around may also be true, such that when the shared beliefs about voice safety and voice efficacy are low, individuals may be more susceptible towards experiencing negative affect. Based on the review of the literature and the identified research gaps, the following research question has been drafted:

What are the moderators in the relationship between voicing and affect?

 How does voicing in a meeting influence an individual’s affect afterwards?

o Are the effects different for the different types of voice (prohibitive and promotive)?

 Is this relationship moderated by the personality traits of the individual?  Is this relationship moderated by the group voice climate?

Method

Sample and procedure

The survey data were collected from different organizations, active in different industries, in the Netherlands. To test the hypothesized relationships a two-wave panel design was employed in combination with observations during the team meetings. The two-wave panel design was used to assess the dependent variable affect before and after the team meetings. In total, 37 meetings were observed and 213 employees filled in the surveys. 585 voice actions were observed, with an average of 16 voice actions per meeting. The average duration of the meetings was 100 minutes (minimum = 30, maximum = 180). The average age of the respondents was 39.33 years (SD = 4.83); 43% were male, and 66% was higher

(17)

Master thesis N.E.M. Hoogenboom 17 educated (university of applied sciences or university). The average organizational tenure was 10.32 years (SD = 8.85).

Affect and voice climate were measured before the team meetings; affect and personality were measured after the team meetings. Respondents were numbered to link the two surveys and to link them to the observed behavior. Because respondents could perceive this as being a less anonymous method, which in turn increases the risk of social desirability (Randall & Fernandes, 1991), a confidentiality form was signed at each meeting.

Measures

Scales ranging from 1 (“not at all”) to 7 (“very extensively”) were used for all variables.

Voice. The researchers measured voice during the team meetings using a form in which they entered the data themselves. Voice could either be promotive (aimed at improving the overall functioning of the team, work unit, or organization) or prohibitive (aimed at expressing concerns about practices or procedures that could harm the team, work unit, or organization) (Liang et al., 2012). Because multiple researchers conducted the data independently, the researchers used two meetings to reach consensus about when speaking up should be seen as voicing and when voicing should be seen as promotive or prohibitive.

Furthermore, voice was also measured in the survey (α = .66). This scale was developed specifically for this study, with a pilot performed on 500 respondents. Respondents were asked to what degree work-related suggestions were made during the meeting. Promotive voice was measured by asking if there were any suggestions made during the meeting to improve current processes. Prohibitive voice was measured by asking to what degree suggestions were made to prevent future harm.

The value of Cronbach’s alpha for the measure of voice in the survey is below .7, and thus below an acceptable value (Pallant, 2007). This could be due to the number of items that are used in the scale, as the value Cronbach’s alpha partially depends on this (Cortina, 1993; Field, 2013). For scales with a lower number of items, a higher correlation between the items is needed to reach the same internal reliability as for a scale with a higher number of

(18)

Master thesis N.E.M. Hoogenboom 18 items (Cortina, 1993). Voice was measured using only three items. Therefore, the mean inter-item correlation may be a more appropriate method to test the reliability. The mean inter-item correlation for voice is .39, which is between .2 and .4 as suggested by Briggs & Cheek (1986). Thus, the reliability of this scale is still seen as acceptable.

Affect. The independent variable ‘affect’ was measured using the Positive and Negative Affect Schedule (PANAS) scale as developed by Watson, Clark, & Tellegen (1988). The PANAS scale consists of 20 items that describe different feelings and emotions. A selection of items was made. Eleven of these items describe a positive affect state (α = .89), whereas the other nine items describe a negative affect state (α = .85). For example, respondents are asked if they feel enthusiastic, nervous, or interested (Watson et al., 1988).

Personality. The moderating variable ‘personality’ was measured using the 20 items of the Mini-IPIP Scales as developed by Donnellan, Oswald, Baird, & Lucas (2006). For this study, extraversion (α = .80) and neuroticism (α = .80) are of special interest. Example items include “I am the life of the party” or “I have frequent mood swings”.

Voice climate. The second moderating variable ‘voice climate’ was measured using 12 items as developed by Morrison et al. (2011), which measures both voice safety (α = .90) and voice efficacy (α = .90). Six items covered voice efficacy. Respondents were asked “To what extent do members of your team feel they are capable of effectively doing each of the following”, such as “develop and make recommendations concerning issues that affect the team”, or “speak up with ideas for new projects or changes in procedures” (LePine & Van Dyne, 1998; Morrison et al., 2011). The other six items covered voice safety. Respondents were asked “To what extent do members of your team feel it is safe to do each of the following”. The same six listings as used for voice efficacy were used for voice safety.

Control variables. The control variables in this study are demographic variables, such as age, gender, education, and tenure.

(19)

Data analysis

Testing normality

Normality was assessed for all the variables. This was done by inspecting skewness and kurtosis, the 5% trimmed means, the normal Q-Q plot and the histogram. The Kolmogorov-Smirnov test and the Shapiro-Wilk test were not used for this, because of the sensitivity of the tests and the associated distortion of normality (Field, 2013). Skewness and kurtosis suffer from the same problem, but this is more common for larger samples (Pallant, 2007; Tabachnick & Fidell, 2007). Since this sample is not very large, skewness and kurtosis are inspected.

Affect was not completely normally distributed. The items were very skewed, both positively and negatively, which could also be seen in the normal Q-Q plots and the histograms. Although this is not uncommon with scales measuring emotions (Pallant, 2007), the variables were transformed to show a more normal distribution. This did not solve the problem; rather, the variables were now skewed to the other side. Therefore, the choice was made to retain the original scores for affect.

The IPIP scale showed some items where both the kurtosis and the skewness were very high (>1 and >2, respectively). However, these items did not measure extraversion and neuroticism and were therefore ignored. The items measuring extraversion and neuroticism showed an acceptable normal distribution.

Voice climate showed an approximately normal distribution. The values of both skewness and kurtosis fall between -1 and 1. Furthermore, the histograms and the normal Q-Q plot displayed an acceptable normal distribution.

Factor analysis

Factor analysis was conducted for the items affect, personality and voice climate. First, it was assessed if the data was appropriate for running a factor analysis by inspecting the correlations, which should show some correlations of r = .3 or higher (Pallant, 2007). Furthermore, Bartlett’s test of sphericity (Bartlett, 1954) and the Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy (Kaiser, 1970, 1974) were used to assess the suitability of a

(20)

Master thesis N.E.M. Hoogenboom 20 factor analysis. The KMO-test have a value of .6 or more and Bartlett’s test of sphericity should reach significance (Bartlett, 1954; Kaiser, 1970, 1974). The KMO values for affect, voice climate, and personality were respectively .87, .87, and .75. Bartlett’s test of sphericity reached significance for all the variables. Then, the components with Eigenvalues above 1 were inspected, followed by the scree plot and the pattern matrix to determine the ideal number of components.

First, affect was inspected. The correlation matrix showed the presence of many correlations of .3 and above and thus factor analysis is suitable for this variable. There were four components with Eigenvalues above 1, explaining 32.3%, 17%, 5.6% and 5.2% of the variance. The scree plot showed a break after the second component, which supported the choice to remain two components. Combined, these components explained 49.3% of the variance. An oblimin rotation was performed to aid the interpretation. There was a weak component correlation (r = -.28), which supported the use of the oblimin rotation. The rotated solution revealed that the two components showed strong loadings and that all variables substantially loaded on one component. Positive affect items showed a strong loading on Component 1 and negative affect items showed a strong loading on Component 2. The component and pattern matrix can be found in Appendix 2.

Then, a factor analysis for voice climate was conducted. The correlation matrix revealed a number of correlations of .3 and above, which supported the suitability of factor analysis. There were two components with Eigenvalues exceeding 1, explaining 52.2% and 15.5% of the variance. The scree plot showed a break after the second component. Therefore, the choice was made to remain two components that combined explained 67.7% of the variance. Again, an oblimin rotation was performed. The component correlation matrix showed higher correlations than expected (r = .53), but the suggested solution was consistent with previous research on group voice climate by Morrison et al. (2011) and thus the oblimin rotation was seen as a feasible solution. The solution showed that voice climate safety items loaded strongly on Component 1 and that the voice climate efficacy items loaded strongly on Component 2. The component and pattern matrix can be found in Appendix 3.

(21)

Master thesis N.E.M. Hoogenboom 21 And lastly, a factor analysis for the IPIP scale was conducted. The correlation matrix showed the presence of coefficients of .3 and above. There were six items with Eigenvalues exceeding 1, explaining 20.8%, 14.9%, 12.0%, 8.3%, 7.4% and 5.2% of the variance. The scree plot did not show a very clear break after a certain component, but the break is most likely after five components. Even in the literature there are still some ambiguities around the ideal number of items, although there seems to be an agreement that the best solution holds five components (Digman, 1990). The analysis was first run with an oblimin rotation, but the component correlation matrix showed low correlations (all below r = .2). For that reason, the analysis was run again with a varimax rotation, which is a more common technique when the component correlations are low (Pallant, 2007). The results did not identify the five suggested components by Donnellan et al. (2006) perfectly. Extraversion loaded on Component 1, together with two items from Openness. Agreeableness, Conscientiousness and Neuroticism loaded on Components 2, 3 and 4 respectively. Openness loaded on Component 5, although two items from Openness also loaded on Component 1. The rotated component matrix can be found in Appendix 4. Despite the results from the factor analysis, the choice was made to use the components as suggested by Donnellan et al. (2006).

Multiple regression analysis

Regression analyses were conducted to test if personality and voice climate acted as moderators in the relationship between voice and affect. The product variable approach was used to test the interaction effect (Baron & Kenny, 1986). Before the interaction terms could be calculated, the independent variables were centered to minimize possible multicollinearity issues (Lance, 1988). This was done for extraversion, neuroticism, voice climate safety, and voice climate efficacy. Because promotive and prohibitive were measured during the meetings and the values for these variables were simply the sum of the voice actions, these variables were not centered.

Both two-way and three-way interaction terms were calculated before conducting the regression analysis. An example of a two-way interaction term is promotive voice ×

(22)

Master thesis N.E.M. Hoogenboom 22 extraversion. A three-way interaction term is for example promotive voice × extraversion × neuroticism.

Several hierarchical multiple regression analyses were run. The variables were entered in different blocks in a predetermined order (Field, 2013; Pallant, 2007). In general, the model looked as follows: the control variables were entered in Block 1. The independent variables were entered in Block 2, the moderating variable(s) in Block 3, the two-way interactions terms in Block 4 and the three-way interaction terms in Block 5.

Because affect was measured twice in the survey (pre-meeting and post-meeting), different methods were used to assess the change in affect. In the first method the change score for positive and negative affect was used. This method can give unbiased estimates of effect and is preferred over for example LDV (lagged dependent variable) (Johnson, 2005). In the second method the post-test scores for both positive and negative affect served as the independent variables. The pre-test scores for affect were then treated as control variables and were thus entered in Block 1.

Results

Correlation matrix

The correlation matrix can be found on the next page, including the means, standard deviations and reliabilities. Several variables show significant correlations. Some of these concern the control variables, such as age and tenure. Significant correlations for the control variables and other variables include for example the negative relationship between promotive voice and gender. This implies that that men tend to voice in a more promotive way than women. Another example is the negative relationship between prohibitive voice and age; the older a person is, the less they voice in a prohibitive way. Extraversion shows significant relationship for three out of four control variables: age (negative), education (positive), and tenure (negative).

(23)

Table 1. Descriptive statistics

M SD 1 2 3 4 5 6 7 8 9 10 11 12 1. Gendera _{0.57 0.50} 2. Age 39.33 10.82 -.006 3. Educationb _{2.90 0.89 -.022 -.162*} 4. Tenure 10.32 8.85 -.012 .638** -.282** 5. Promotive voice 12.92 6.65 -.252** -.047 -.098 -.010 6. Prohibitive voice 9.90 9.73 -.101 -.214** .033 -.130 .612**

7. Positive affect (post) 5.26 0.84 -.112 .095 -.037 .104 .117 .107 (.887)

8. Negative affect (post) 1.78 0.89 .034 .128 -.169* .061 .077 .014 -.397** (.880)

9. Extraversion 4.58 1.24 .011 -.266** .223** -.249** -.027 .097 .036 -.096 (.795)

10. Neuroticism 2.40 1.11 .151* -.011 -.219** -.052 .063 -.023 -.303** .564** -.087 (.798)

11. Voice climate safety 5.33 0.94 -.151* -.133 .148* -.048 .118 .135* .219** -.349** .137* -.406** (.897)

12. Voice climate efficacy 4.90 0.92 -.061 -.027 .084 .108 .090 .190** .377** -.276** .225** -.348** .546** (.908) ** Correlation is significant at the 0.01 level (2-tailed).

* Correlation is significant at the 0.05 level (2-tailed).

a_{0 = male, 1 = female.}b _{1 = secondary school, 2 = college, 3 = university of applied science, 4 = university, 5 = other.} Reliabilities are displayed on the diagonal.

(24)

Master thesis N.E.M. Hoogenboom 24 The post-test scores for affect are used in the correlation matrix. As expected, positive and negative affect are negatively correlated. Especially for short time intervals this correlation is common (Cote & Moskowitz, 1998; Moskowitz & Cote, 1995). These results indicate that individuals who experience positive affect tend to experience less negative affect, which is not surprising. This is in line with earlier research by Russell & Carroll (1999), who stated that affect is a bipolar construct. Other expected correlations include the negative relationship between neuroticism and positive affect, and the positive relationship between negative affect and neuroticism.

Furthermore, both voice climate safety and voice climate efficacy show a positive correlation with positive affect and a negative correlation with negative affect. In other words, a favorable voice climate is positively related to positive affect and an unfavorable voice climate is negatively related to negative affect. Although the results of the correlation matrix are not conclusive, expected relationships between variables are significant and it seems, thus far, reasonable to assume that the hypotheses may also be supported.

The personality traits extraversion and neuroticism show some significant correlations with control variables, such as age, education, and tenure. And as discussed before, neuroticism is negatively correlated to positive affect and positively correlated to negative affect. A surprising result here is that extraversion does not show a significant relationship with neither positive affect nor negative affect, even though previous research found several relationships between extraversion and positive affect (Costa & McCrae, 1980; Rusting & Larsen, 1997).

And as discussed, both measures of voice climate correlate positively with positive affect and negatively with negative affect. Furthermore, both voice climate measures also correlate positively with extraversion and negatively with neuroticism.

Voicing and affect

In order to test if prohibitive voicing and promotive voicing influence affect, a hierarchical regression analysis was run. In Block 1 the control variables were entered into the model, i.e. age, gender, tenure, and education and the pre-test scores for positive or

(25)

Master thesis N.E.M. Hoogenboom 25 negative affect. In Block 2, the sum of voicing was entered. The post-test scores for positive or negative affect served as dependent variables. The results can be found in Table 2 to Table 5.

Table 2. Summary of hierarchical regression analysis for change in positive affect (N = 207)

Model 1 Model 2 Variable B SE B β B SE B β Age -.010 .006 -.152 -.008 .006 -.121 Gender -.077 .099 -.053 -.105 .103 -.073 Education -.080 .057 -.100* -.095 .057 -.120* Tenure -.005 .008 -.057 -.005 .008 -.059 Promotive voice .014 .007 .190** Prohibitive voice -.017 .010 -.155* R2_.042_.065 F for change in R2_2.243* _2.389* * p < .1. **p < .05. ***p < .01.

Table 3. Summary of hierarchical regression analysis for positive affect (post-test) (N = 207)

Model 1 Model 2 Variable B SE B β B SE B β Age -.058 .061 -.068 -.037 .062 -.044 Gender -1.328 1.018 -.071 -1.459 1.053 -.078 Education -.684 .584 -.066 -.795 .590 -.077 Tenure -.013 .078 -.012 -.014 .077 -.013

Positive affect (pretest) 7.274 .645 .639*** 7.332 .645 .644***

Promotive voice .136 .067 .144**

Prohibitive voice -.127 .102 -.090

R2_.403_.415

F for change in R2 _{27.314*** 2.053}

* p < .1. **p < .05. ***p < .01.

The results in Table 2 show that both Model 1 and Model 2 are significant. Model 2 explains an additional 2.2% of the variance. Furthermore, the ANOVA results indicate that the model as a whole [F (6, 207) = 2.312, p < .05] is significant. Model 2 shows that promotive voice is significantly and positively related to positive affect (p < .05). Prohibitive voicing on the other hand is negatively related to positive affect (p < .1). Thus, promotive voice is related to an increase in positive affect, whereas prohibitive voice is associated with

(26)

Master thesis N.E.M. Hoogenboom 26 a decrease in positive affect. Table 3 displays the results when controlled for the prior mood. The control variables (Model 1) explain 40.3% of the variance. Model 2 on the other hand explains 41.5% of the variance. The R2_{change value is .012, which means that adding voice}

explains only an additional 1.2% of the variance in positive affect. The ANOVA results indicate that the model as a whole [F (7, 207) = 20.300, p < .001] is significant. Model 2 shows that promotive voice is positively and significantly related to positive affect. This means that the higher the score for voice, the higher the experience positive affect afterwards.

Table 4. Summary of hierarchical regression analysis for change negative affect (N = 208)

Model 1 Model 2 Variable B SE B β B SE B β Age .001 .006 .016 .002 .007 .023 Gender .029 .107 .019 .037 .112 .024 Education .015 .062 .018 .016 .063 .019 Tenure .005 .008 .054 .005 .008 .055 Promotive voice .003 .007 .034 Prohibitive voice .000 .011 .003 R2 _{.004 .006} F for change in R2 _{.222 .124} * p < .1. **p < .05. ***p < .01

Table 5. Summary of hierarchical regression analysis for negative affect (post-test) (N = 208)

Model 1 Model 2 Variable B SE B β B SE B β Age .060 .053 .080 .064 .054 .086 Gender .241 .876 .015 .489 .916 .030 Education -.281 .515 -.031 -.220 .522 -.024 Tenure .007 .066 .008 .009 .067 .010

Negative affect (pretest) 6.048 .531 .626*** 6.016 .534 .623***

Promotive voice .010 .059 .012

Prohibitive voice .062 .089 .050

R2_.416_.419

F for change in R2 _{28.881*** .531}

(27)

Master thesis N.E.M. Hoogenboom 27 Table 4 and 5 display the results for the change in negative affect and the results when controlled for prior mood. None of the tables display any relevant significant results. Hypothesis 1a stated that promotive voicing is more positively related to positive affect than prohibitive voicing. Based on the results in Table 2 and 3 it can be concluded that promotive voicing is indeed more strongly related to a change in positive affect than prohibitive voicing. Therefore, Hypothesis 1a is supported. Hypothesis 1b stated that prohibitive voicing is more related to negative affect than promotive voicing. Based on the results in Table 4 and 5, no support is found for Hypothesis 1b. Although prohibitive voicing is related to a decrease in positive affect, the results show that there is no reason to assume that prohibitive voicing is also associated with an increase in negative affect.

Personality as moderator

In the second hierarchical multiple regression analysis it was tested if personality moderated the relationship between voice behavior and affect. Results for when the change in affect served as the dependent variable can be found in Table 6 (change in positive affect) and Table 7 (change in negative affect) on the next pages. The results for when the post-test score of affect served as the dependent variable are displayed in Table 8 and 9.

Table 6 shows that the control variables explain 4.2% of the variance. Model 2 explains an additional 2.3% (p < .1) and Model 3 an additional 0.2% (p < .1). None of the interaction variables showed a significant result.

Table 7 shows the results for negative affect and personality. Model 1 only explains 0.4% of the variance, whereas Model 3 explains 3.8% of the variance. Although the values are still low, the change in variance is significant for Model 3. Model 4 already explains 6.6% of the variance. Although the increase is not significant, it is close to it. Table 7 shows that neuroticism is negatively and significantly (p < .01) related to the experienced negative affect, which was expected. Model 4 shows that the interaction term of prohibitive voicing and extraversion is significantly and negatively related to negative affect (p < .05). Thus, in meetings where there is relatively more prohibitive voicing, extravert individuals tend to experience less negative affect afterwards.

(28)

Table 6. Summary of hierarchical analysis for change in positive affect and personality (N = 207)

Model 1 Model 2 Model 3 Model 4 Model 5

Variable B SE B β B SE B β B SE B β B SE B β B SE B β Age -.010 .006 -.154* -.008 .006 -.122 -.007 .006 -.113 -.006 .006 -.094 -.006 .006 -.091 Gender -.081 .100 -.056 -.110 .103 -.076 -.107 .105 -.074 -.101 .106 -.070 -.101 .107 -.070 Education -.076 .057 -.096 -.091 .058 -.115 -.100 .060 -.125 -.097 .061 -.121 -.094 .061 -.118 Tenure -.004 .008 -.054 -.005 .008 -.056 -.004 .008 -.054 -.005 .008 -.058 -.005 .008 .060 Promotive voice .014 .007 .194** .014 .007 .189 .011 .007 .153 .011 .007 .147 Prohibitive voice -.017 .010 -.156* -.017 .010 -.152 -.015 .010 -.133 -.014 .011 -.128 Extraversion .007 .011 .045 -.006 .019 -.042 -.006 .019 -.039 Neuroticism -.010 .048 -.015 .029 .094 .044 .026 .095 .040 Prom × extra .002 .001 .226 .002 .001 .232 Proh × extra -.001 .002 -.112 -.001 .002 -.124 Prom × neuro -.005 .007 -.099 -.004 .008 -.073 Proh × neuro .002 .009 .028 .000 .009 .009

Prom × extra × neuro -.001 .002 -.109

Proh × extra × neuro .001 .002 .090

R2 _{.042 .065 .067 .080 .082}

F for change in R2 _{2.209* 2.464* .224 .667 .238}

(29)

Table 7. Summary of hierarchical regression analysis for change in negative affect and personality (N = 207)

Variable B SE B β B SE B β B SE B β B SE B β B SE B β Age .001 .006 .019 .002 .007 .024 .001 .006 .020 .002 .007 .026 .001 .007 .020 Gender .040 .107 .026 .046 .112 .030 -.005 .112 -.003 .007 .113 .004 .007 .113 .004 Education .007 .062 .009 .008 .063 .010 .041 .064 .048 .043 .065 .051 .038 .065 .045 Tenure .004 .008 .049 .004 .008 .049 .006 .008 .076 .007 .008 .084 .007 .008 .087 Promotive voice .002 .007 .027 .003 .007 .045 .000 .007 .001 .001 .008 .013 Prohibitive voice .001 .011 .005 -.003 .011 -.028 .000 .011 -.001 -.001 .011 -.010 Extraversion .001 .011 .010 .026 .020 .168 .025 .021 .162 Neuroticism .132 .051 .191*** .178 .100 .256* .182 .101 .262* Prom × extra .002 .002 .170 .002 .002 .159 Proh × extra -.004 .002 -.352** -.004 .002 -.330* Prom × neuro .005 .008 .106 .003 .008 .062 Proh × neuro -.011 .009 -.194 -.009 .009 -.160

Prom × extra × neuro .002 .002 .189

Proh × extra × neuro -.002 .002 -.162

R2 _{.004 .005 .038 .066 .072}

F for change in R2 _{.217 .086 3.399*}_{1.451 .693}

(30)

Table 8. Summary of hierarchical regression analysis for positive affect (post-test) and personality (N = 206)

Variable B SE B β B SE B β B SE B β B SE B β B SE B β

Age -.059 .061 -.069 -.038 .062 -.044 -.020 .062 -.023 -.018 .063 -.021 -.015 .063 -.018

Gender -1.385 1.020 -.074 -1.517 1.054 -.081 -1.123 1.061 -.060 -1.165 1.077 -.062 -1.164 1.081 -.062

Education -.640 .585 -.062 -.747 .591 -.073 -1.073 .604 -.104* -1.019 .615 -.099* -1.003 .621 -.097

Tenure -.010 .078 -.009 -.010 .077 -.010 -.023 .077 -.022 -.022 .079 -.020 -.022 .079 -.021

Positive affect (pre-test) 7.265 .645 .639*** 7.322 .645 .644*** 6.844 .673 .602*** 6.881 .692 .605*** 6.876 .695 .605*

Promotive voice .140 .067 .147** .122 .067 .128* .113 .071 .119 .107 .072 .113 Prohibitive voice -.128 .102 -.090 -.085 .103 -.060 -.080 .107 -.057 -.071 .108 -.050 Extraversion .088 .106 .047 -.044 .195 -.024 -.035 .197 -.019 Neuroticism -1.087 .507 -.128** -.646 .966 -.076 -.655 .974 -.077 Prom × extra .006 .015 .054 .007 .015 .061 Proh × extra .005 .020 .031 .003 .020 .019 Prom × neuro -.033 .075 .052 -.019 .077 -.030 Proh × neuro -.003 .089 -.005 -.017 .091 -.023

Prom × extra × neuro -.013 .017 -.095

Proh × extra × neuro .016 .022 .093

R2 _{.404 .417 .433 .436 .438}

F for change in R2 _{27.288*** 2.153 2.755* .266 .295}

(31)

Table 9. Summary of hierarchical regression analysis for negative affect (post-test) and personality (N = 207)

Variable B SE B β B SE B β B SE B β B SE B β B SE B β

Age .061 .052 .081 .064 .053 .086 .069 .050 .093 .078 .050 .105 .078 .051 .104

Gender .310 .875 .019 .549 .915 .034 -.363 .859 -.022 -.262 .859 -.016 -.278 .862 -.017

Education -.325 .515 -.036 -.264 .522 -.029 .140 .492 .016 .144 .494 .016 .166 .498 .018

Tenure .005 .066 .005 .006 .066 .006 .028 .062 .031 .030 .062 .033 .026 .062 .028

Negative affect (pre-test) 6.104 .531 .630*** 6.070 .534 .626*** 4.498 .558 .464*** 4.480 .559 .462*** 4.436 .568 .458***

Promotive voice .006 .059 .007 .031 .054 .037 -.005 .057 -.006 -.001 .057 -.001 Prohibitive voice .063 .089 .051 .004 .083 .003 .041 .085 .033 .030 .086 .025 Extraversion -.009 .086 -.005 .106 .156 .065 .089 .158 .055 Neuroticism 2.638 .438 .355*** 2.638 .777 .355*** 2.601 .781 .350*** Prom × extra .021 .012 .203* .020 .012 .195 Proh × extra -.036 .016 -.275** -.035 .016 -.262** Prom × neuro .062 .060 .113 .056 .062 .103 Proh × neuro -.074 .071 -.119 -.065 .073 -.104

Prom × extra × neuro .006 .014 .050

Proh × extra × neuro -.013 .018 -.087

R2 _{.421 .424 .514 .529 .532}

F for change in R2 _{29.416*** .482 18.244*** 1.622} _.431

(32)

Master thesis N.E.M. Hoogenboom 32 Tables 8 and 9 show the results of the regression analysis when controlled for the pre-test scores of affect. The post-test scores for affect served as the dependent variables. Models 1 and 3 are significant in Table 8, and none of the interaction terms show a significant result. Compared to Table 6, there are some new significant results. It shows that promotive voice is positively related to positive affect (p < .05). Furthermore, neuroticism is negatively related to positive affect (p < .05). The opposite effect was already observed in Table 7.

The last table that describes the relationships between affect and personality is Table 9, where the post-test scores of negative affect served as the dependent variable. This table shows the same results as Table 7. Again, neuroticism is positively related to negative affect (p < .01) and the interaction term of prohibitive voicing and extraversion is negatively related to negative affect (p < .05). However, the interaction term between promotive voicing and extraversion is now also significant, albeit at the .10 level.

Hypothesis 2a stated that extraversion would be associated with positive affect and that this relationship would be stronger for promotive voicing. Tables 6 and 8 display the results for positive affect as the dependent variable, but there are no relevant significant relationships. The results in Tables 7 and 9, where negative affect serves as the dependent variable, do provide evidence that extraversion is associated with a decrease in negative affect in meetings with prohibitive voicing. However, these results do not provide enough evidence to support hypothesis 2a. Although it can be suspected that extraversion is associated with positive affect, as it at least decreases negative affect, there is no reason to assume that this effect is stronger for promotive voicing. In fact, the results are stronger for prohibitive voicing. Thus, hypothesis 2a is not supported.

Hypothesis 2b on the other hand predicted that neuroticism would be associated with negative affect and that this effect would be stronger for prohibitive voicing. Although neuroticism indeed showed a positive, significant relationship for negative affect, there is no reason to assume that this is due to promotive or prohibitive voicing. Thus, based on these results, hypothesis 2b is also not supported.

(33)

Voice climate as moderator

The third hierarchical regression analysis tested if voice climate moderated the relationship between voice behavior and affect. The same method as for personality was used. Four multiple regression analyses were run: two where the change in positive and negative affect served as dependent variables, and two where the post-test scores for affect served as dependent variables. The results are displayed on the next pages in Tables 10 to 13.

Table 10 displays the results for the change in positive affect and voice climate. Although Model 1 and 2 are statistically significant, it is weak (p < .10). The control variables explain 4.2% of the variance, whereas the last model (Model 5) explains 9.6% of the variance. Most models are not significant. However, Model 3 does show that voice climate efficacy is positively related to the change in positive affect. In other words, the higher the score on voice climate efficacy, the higher the experienced positive affect (p < .05). The table does not show any other significant results.

Table 11 does not display any significant results, nor are any of the models in this analysis significant. The control variables explain a mere 0.2% of the variance and even the last model (Model 5) only explains 2.5% of the variance. Table 12 on the other hand does display significant results. Model 1 and 3 are significant (p < .01). Model 4 and 5, however, are not significant nor are there any significant results for the interaction terms. It does show however that voice climate efficacy is positively related to positive affect. The same result was obtained in Table 10, when the change in positive affect served as the dependent variable. The significance however is now higher (p < .01 instead of p < .05). Because none of the interaction terms are significant, it is not possible to conclude whether voice climate efficacy moderates the relationship between a specific type of voicing and positive affect.

Table 13 displays the results when the post-test scores for negative affect served as the dependent variable. Similar to the results in Table 11, where the change in negative affect served as the dependent variable, no significant results were obtained.

(34)

Table 10. Summary of hierarchical regression analysis for change in positive affect and voice climate (N = 206)

Variable B SE B β B SE B β B SE B β B SE B β B SE B β Age -.010 .006 -.152* -.008 .006 -.121 -.008 .006 -.116 -.007 .006 -.113 -.009 .006 -.131 Gender -.077 .099 -.053 -.105 .103 -.073 -.113 .103 -.079 -.133 .106 -.092 -.139 .107 -.096 Education -.080 .057 -.100 -.095 .057 -.120* -.098 .058 -.123* -.095 .058 -.119 -.096 .059 -.121 Tenure -.005 .008 -.057 -.005 .008 -.059 -.008 .008 -.092 -.007 .008 -.081 -.006 .008 -.070 Promotive voice .014 .007 .190** .012 .007 .160* .012 .007 .164* .011 .007 .154 Prohibitive voice .017 .010 -.155* -.016 .010 -.142 -.017 .010 -.159* -.018 .011 -.167* VC Efficacy .134 .065 .173** .195 .123 .251 .181 .124 .234 VC Safety -.084 .064 -.110 -.168 .125 -.219 -.218 .130 -.167 Prom × VC Efficacy .003 .008 .055 .006 .009 .111 Proh × VC Efficacy -.011 .014 -.148 -.013 .014 -.184 Prom × VC Safety .002 .009 .031 -.003 .009 -.057 Proh × VC Safety .007 .014 .103 .010 .014 .135 Prom × VCE × VCS .003 .007 .059 Proh × VCE × VCS .003 .011 .040 R2 _{.042 .065 .085 .089 .096} F for change in R2 _{2.243* 2.389* 2.203 .231 .761} * p < .1. **p < .05. ***p < .01.

(35)

Table 11. Summary of hierarchical regression analysis for change in negative affect and voice climate (N = 208)

Variable B SE B β B SE B β B SE B β B SE B β B SE B β Age .001 .006 .016 .002 .007 .023 .002 .007 .024 .001 .007 .016 .001 .007 .014 Gender .029 .107 .019 .037 .112 .024 .045 .114 .029 .057 .117 .037 .051 .117 .033 Education .015 .062 .018 .016 .063 .019 .015 .064 .018 .007 .064 .008 .006 .064 .007 Tenure .005 .008 .054 .005 .008 .055 .006 .008 .067 .006 .008 .065 .005 .008 .062 Promotive voice .003 .007 .034 .004 .007 .045 .004 .007 .047 .001 .008 .008 Prohibitive voice .000 .011 .003 .000 .011 -.003 .002 .011 .015 .006 .012 .048 VC Efficacy -.061 .071 .073 -.139 .135 -.168 -.157 .137 -.190 VC Safety .052 .070 .063 .213 .138 .263 .214 .143 .263 Prom × VC Efficacy .004 .009 .067 .005 .010 .101 Proh × VC Efficacy .004 .016 .048 .004 .016 .056 Prom × VC Safety -.009 .010 -.153 -.010 .010 -.175 Proh × VC Safety -.007 .015 -.088 -.007 .015 -.096 Prom × VCE × VCS .008 .008 .157 Proh × VCE × VCS -.010 .012 -.136 R2 _{.004 .006 .010 .020 .025} F for change in R2 _{.222 .124 .420 .538 .481} * p < .1. **p < .05. ***p < .01.

(36)

Table 12. Summary of hierarchical regression analysis for positive affect (post-test) and voice climate (N = 207)

Variable B SE B β B SE B β B SE B β B SE B β B SE B β Age -.058 .061 -.068 -.037 .062 -.044 -.011 .061 -.013 -.012 .062 -.014 -.024 .063 -.028 Gender -1.328 1.018 -.071 -1.459 1.053 -.078 -1.407 1.032 -.075 -1.425 1.067 -.076 -1.432 1.067 -.077 Education -.684 .584 -.066 -.795 .590 -.077 -.940 .578 -.091 .906 .586 -.088 -.914 .585 -.089 Tenure -.013 .078 -.012 -.014 .077 -.013 .061 .076 -.057 .061 .078 -.057 -.045 .079 -.042 Positive affect 7.274 .645 .639*** 7.332 .645 .644*** 6.721 .657 .590*** 6.692 .672 .588*** 6.588 .682 .579*** Promotive voice .136 .067 .144** .091 .067 .095 .088 .068 .093 .112 .074 .118 Prohibitive voice -.127 .102 -.090 -.096 .100 -.068 -.107 .104 -.075 -.158 .111 -.112 VC Efficacy 2.299 .658 .229*** 2.350 1.229 .234* 2.392 1.241 .238* VC Safety -.491 .640 -.050 1.205 1.260 -.122 -.708 1.305 -.071 Prom × VC Efficacy -.012 .084 -.018 .001 .088 .002 Proh × VC Efficacy .012 .144 .013 -.023 .146 -.025 Prom × VC Safety .006 .087 .009 -.037 .094 -.054 Proh × VC Safety .069 .137 .076 .104 .139 .114 Prom × VCE × VCS -.053 .073 -.089 Proh × VCE × VCS .150 .113 .162 R2 _{.403 .415 .453 .454 .461} F for change in R2 _{27.314*** 2.053 6.731*** .159 1.245} * p < .1. **p < .05. ***p < .01.

(37)

Table 13. Summary of hierarchical regression analysis for negative affect (post-test) and voice climate (N = 208)

Variable B SE B β B SE B β B SE B β B SE B β B SE B β Age .060 .053 .080 .064 .054 .086 .053 .053 .071 .050 .054 .067 .047 .055 .063 Gender .241 .876 .015 .489 .916 .030 .303 .918 .019 .553 .942 .034 .472 .943 .029 Education -.281 .515 -.031 -.220 .522 -.024 -.139 .520 -.015 -.221 .524 -.025 -.242 .524 -.027 Tenure .007 .066 .008 .009 .067 .010 .027 .067 .029 .018 .068 .019 .014 .068 .016 Negative affect 6.048 .531 .626*** 6.016 .534 .623*** 5.570 .576 .577*** 5.512 .580 .571*** 5.411 .584 .560*** Promotive voice .010 .059 .012 .026 .059 .031 .022 .060 .026 -.016 .065 -.019 Prohibitive voice .062 .089 .050 .064 .089 .052 .094 .091 .076 .140 .098 .113 VC Efficacy -.724 .577 -.082 -1.820 1.092 -.207* -2.071 1.104 -.236* VC Safety -.557 .590 -.064 1.078 1.119 .125 1.119 1.158 .129 Prom × VC Efficacy .002 .074 .004 .027 .077 .048 Proh × VC Efficacy .119 .127 .146 .125 .128 .154 Prom × VC Safety -.050 .077 -.083 -.070 .084 -.117 Proh × VC Safety -.117 .121 -.147 -.123 .123 -.155 Prom × VCE × VCS .097 .064 .183 Proh × VCE × VCS -.119 .099 -.147 R2 _{.416 .419 .432 .441 .448} F for change in R2 _{28.881*** .531 2.341* .813 1.153} * p < .1. **p < .05. ***p < .01.

(38)

Master thesis N.E.M. Hoogenboom 38 Hypothesis 3a stated that a favorable voice climate would make individuals more susceptible towards experiencing positive affect, and that this relationship would be stronger for promotive voicing. Voice climate efficacy is indeed associated with positive affect, but voice climate safety did not show a significant relationship with positive affect. However, because none of the interaction terms were significant no inferences can be made about the type of voice that would strengthen or attenuate the effect. Therefore, hypothesis 3a could not be supported. Hypothesis 3a stated that an unfavorable voice climate would make individuals more susceptible towards experiencing negative affect, and that this relationship would be stronger for prohibitive voicing. However, no significant relationship were found between voice behavior, voice climate, and negative affect.

Discussion

General discussion

Based on Liang et al.’s (2012) distinction between promotive and prohibitive voicing and Russell’s (1980) circumplex model of affect, this study examined the direct relationship between employee voice behavior and affect, and the possible moderators – personality and voice climate – of this relationship. The research question that was central in this study was if the relationship between voicing and affect is moderated by personality and/or voice climate. Different subquestions were formulated. The first subquestion studied the direct relationship between voicing and affect. This also corresponds with the first hypothesis. The second and third subquestion tried to answer if the relationship between voicing and affect was moderated by personality traits (extraversion and neuroticism) and the group voice climate. The second and third hypothesis researched the above-named moderating relationships. The research used a two-wave panel design, in which the dependent variable affect was measured at two points in time. This design allows for more accurate inferences about which effect a moderator exactly has on the dependent variable. These effects were assessed through using two methods: in the first method, the change in affect served as the dependent variable. In the second method, the post-test scores for affect functioned as the

The influence of employee voice behavior on affect : a two-wave panel study into the effects and moderators of the relationship between employee voice behavior and affect

The influence of employee voice behavior on affect

N.E.M. Hoogenboom

The influence of employee voice behavior on affect

University

Author

Supervisor

Table of Contents

Abstract

Introduction

Literature review

Employee voice behavior

Affect

Employee voice behavior and affect

Personality

Personality and affect

Voice climate

Voice climate and affect

Research questions

Method

Sample and procedure

Measures

Data analysis

Testing normality

Factor analysis

Multiple regression analysis

Results

Correlation matrix

Voicing and affect

Personality as moderator

Voice climate as moderator

Discussion

General discussion