Subjective well-being (SWB) is defined as ‘a person’s cognitive and affective evaluations of his or her life’ (Diener, Lucas, and Oshi, 2002, p. 63) both right now and for longer periods, for example, for as far back as a year (Diener, 1984; Diener et al., 1999; Kahneman et al., 1999; Argyle, 2001). These evaluations include emotional reactions, moods and judgements that are formed by the individuals about their life satisfaction, for example, marriage and work (Diener, 1984; Diener et al., 1999; Kahneman et al., 1999; Argyle, 2001). In this manner, SWB concerns the investigation of what individuals may call bliss or fulfillment (Diener, 1984; Stock, Okun and Benin, 1986; Diener et al., 1999; Kahneman et al., 1999; Argyle, 2001). Historically, the terms ‘subjective well-being’ and ‘happiness have been used interchangeably by the scientists to narrate how the overall quality of life is evaluated by the individuals (Veenhoven, 1984; Veenhoven, 1987; Diener et al., 2003). However, Bradburn (1969) defined subjective well-being as having distinct negative and positive affects and stated that “an individual will be high in subjective well-being in the degree to which he has an excess of positive over negative affect and will be low in subjective well-being in the degree to which negative affect predominates over positive” (Hills & Argyle, 2001a).
Subjective well-being has been explored by the researchers through a wide range of different methods, for example, experiments, surveys (Diener, 1984). The history and understanding of subjective well-being comes from several different theoretical traditions that have been reviewed by the researchers (Diener, 1984; Veehoven, 1984). This understanding comes from the quality of life researchers and sociologists who are concerned about how subjective well-being is influenced by the demographic factors, for example, income and marital status (Bradburn 1969, Andrews & Withey 1976, Campbell et al. 1976). It comes from the mental health researchers who want to extend their ideas and embed life satisfaction and fulfilment into their research (Jahoda 1958). Finally, it comes from the personality, social and cognitive psychologists who study individuals differences of happy and unhappy people and how their feelings of well-being are influenced by these variations (Wessman & Ricks 1966; Brickman & Campbell 1971; Emmons 1986; Lykken & Tellegen 1996; Parducci 1995; Diener & Lucas 2000; Lucas et al. 2002).
Subjective well-being is measured, by psychologists, by measuring how people think and feel about their lives (Diener et al.,1999). There are three components of subjective well-being (Andrews and Withey, 1976; Lucas et al., 1996; Diener et al., 1997; Diener et al.,1999). These components are described as ‘Life satisfaction’, ‘Positive Affect’ and ‘Negative Affect (Andrews and Withey, 1976; Lucas et al., 1996; Diener et al., 1997; Diener et al.,1999) and these should be studied and measured independently (Andrews & Withey, 1976, Lucas et al., 1996).
Life Satisfaction (LS) is defined as the variation to which the overall quality of life is positively evaluated by an individual as a whole (Pavot et al., 1991; Lucas et al., 1996; Veenhoven, 1996; DeNeve & Cooper, 1998; Diener et al., 2003). Negative Affect (NA) is a general measurement o f subjective pain and unpleasurable engagement that include an assortment of aversive disposition states, for example, fear, blame, guilt, anger, anxiety with low negative affect being a state of calmness and quietness in individuals (Watson et al., 1988). Positive Affect (PA) mirrors the degree to which an individual feels energetic, dynamic and cautious (Watson et al., 1988). Individuals are said to be in the state of high positive affect if they are highly energetic, fully concentrated and pleasantly engaged, whereas, individuals are said to be in the state of low positive affect if they lack energy, enthusiasm and are sad (Watson et al., 1988).
Subjective well-being is by and large thought to be one of the most important goals by the individuals that they seek to achieve throughout their lives (Tay et al., 2015; Diener, King & Napa, 1998; Sapyta, & Suh, 1998; Diener, 2000). Progressively, factors that contribute to workers’ overall well-being have been understood and reconginsed by the organisational psychologists in recent years (Ilies, Schwind, & Heller, 2007; Weiss & Rupp, 2011) and these factors have led to an increased emphasis on workers’ overall subjective well-being (Tay & Harter, 2013). Organisational psychologists have shown interest in both work and non-work domains, for example, family that affect individual’s overall subjective well-being (Erdogan, Bauer, Truxillo, & Mansfield, 2012) and have suggested that health and the quality of life is enhanced through subjective well-being (Boehm et al., 2011; Diener & Chan, 2011; Steptoe & Wardle, 2011). Furthermore, social relationships are also improved by subjective well-being (Lyubomirsky, King, & Diener, 2005).
Overall quality of workers’ subjective well-being helps organisations achieve their goals (Diener, Nickerson, Lucas, & Sandvik, 2002). Research suggests that organisations benefit from increased employee happiness as increased employee happiness increases their success, performance and productivity and is profitable for the organisation (Berkman, 1971; Organ & Ryan, 1995; Staw & Barsade, 1993; Staw et al., 1994; Wright & Bonett, 1997; Wright & Cropanzano, 2000a; Wright et al., 2002; Diener, Nickerson, Lucas, & Sandvik, 2002; Fisher, 2003; Diener, 2004; Lyubomirsky, King, & Diener, 2005; Layard, 2006; De Neve et al., 2013; Bartolini, Bilancini, Bruni, & Porta, 2016). Furthermore, people with higher subjective well-being are happier, committed to the organisation, satisfied with their jobs, get along with their colleagues, show up on time, show organisational citizenship behaviours and are better employees (Diener, 1984; Headey, Veenhoven, & Wearing, 1991; Feist, Bodner, Jacobs, Miles, & Tan, 1995; Pelled & Xin, 1999; Fisher, 2013; De Neve, Diener, Tay and Xuereb, 2013). Wright and Cropanzano (2000a) suggest that workers’ performance is one of the few things that organisation can benefit from through improved subjective well-being. They suggest that subjective well-being increases happiness and improve workers’ health which in turn lowers absenteeism, employee turnover and increases performance (Wright et al., 2002).
Personality “refers to an individual’s characteristic patterns of thought, emotion, and behavior, together with the psychological mechanisms – hidden or not – behind those patterns” (Funder, 2004, p. 5). The Big Five or the five factor model, by Costa and McCrae 1992, is the most common personality model that is used as a starting point to explain personality (Jennifer and Smith, 2007; Vollrath, 2001; Feizi et al., 2015). The Big Five has traits of Extraversion (E), Neuroticism (N), Openness to Experience (O), Conscientiousness (C) and Agreeableness (A) (John & Srivastava, 1999; McCrae et al., 2000; Rothbart & Bates, 1998; Leandro and Castillo, 2010).
Extraversion is defined as individuals (high on extraversion) who are sociable, enthusiastic, assertive, active, excitement seeking, energetic, cheerful, talkative and emotionally positive (Costa & McCrae, 1992; McCrae & John, 1992; Rothbart & Bates, 1998; DeNeve & Cooper, 1998; Zhao and Seibert, 2006), whereas, individuals who score low on extraversion are reserved, independent, quiet and prefer to spend time on their own (Costa & McCree, 1992; Zhao & Seibert, 2006). Neuroticism is defined as individuals, that score high on neuroticism, who experience negative emotions, are self conscientious, less emotionally stable and more likely to experience stress (Costa & McCrae, 1992; Zhao & Seibert, 2006). On the other hand, individuals who score on neuroticism are relaxed, calm and self-confident (Zhao & Seibert, 2006). Openness to experience describes individuals who are curious, like to explore ideas and seek new experiences (McCrae, 1987; Zhao and Seibert, 2006). Individuals who score high on openness to experience are creative, reflective, analytical and innovative (Zhao and Seibert, 2006). Individuals who score low on openness to experience tend to be conventional and have a narrow interest in trying new things (Zhao and Seibert, 2006). Conscientiousness describes individuals, high on conscientiousness, who are motivated, organised, persistent, and hard working in order to achieve their goals (Barrick & Mount, 1991; Zhao & Seibert, 2006). The final personality trait of the big five is agreeableness. Individuals who score high on agreeableness are described as caring, trusting, friendly and forgiving (Zhao and Seibert, 2006). These individuals are cooperative and try to maintain positive relationships with everyone (Zhao and Seibert, 2006). Individuals who score low on agreeableness are described as selfish, ruthless and manipulative (Zhao and Seibert, 2006). These individuals care about themselves and are self-centred (Costa & McCrae, 1992; Digman, 1990; Zhao and Seibert, 2006).
Personality traits, described above, have been reported to be stable across different cultures (Hendriks et al., 2003; McCrae, Costa, del Pilar, Rolland, & Parker, 1998) and regardless of different positive or negative life events that may fluctuate the levels of subjective well-being, personality traits of individuals will bring these levels back to their normal levels (Ormel & Schaufeli, 1991; Ormel & Wohlfarth, 1991). It has been suggested by several subjective well-being review articles that personality is one of the strongest predictor of subjective well-being, controls most of its variance and plays a huge role in people’s subjective well-being (Costa & McCrae, 1980; Diener, 1984; Diener & Larsen, 1993; McCrae & Costa, 1991; Myers, 1992; Myers & Diener, 1995; Diener, 1996; Ferguson, 2001; Vollrath, 2001).
Historically, researchers interested in subjective well-being, were more focused on the external factors that can improve individuals’ subjective well-being and improve the overall quality of their lives (Wilson, 1967; Diener et al., 1999). A wide variety of external factors were identified by Wilson (1967) that were related to subjective well-being of individuals. However, it was later realised that these external or demographic factors, for example, income, education level only had a small impact on subjective well-being (Diener et al., 1999), whereas, personality traits were found to have a stronger influence on subjective well-being (Diener et al., 1999). This led to an increase focus on studying and understanding the relationship between different personality traits and subjective well-being (Diener et al., 1999).
Many personality traits have been found to be linked to subjective well-being of individuals (DeNeve & Cooper 1998). However, majority of the researchers, that have explored this area, in order to understand the relationship between personality and subjective well-being, have been consistent in their findings and suggest that neuroticism and extraversion are strongly correlated with subjective well-being (Costa & McCrae 1980, Watson & Clark, 1984; Tellegen 1985, Headey & Wearing, 1992, Watson & Clark 1992; Steel & Ones, 2002; Steel et al., 2008). Research has reported neuroticism to have been linked to negative affect of subjective well-being (Fujita, 1991; McCrae & John, 1992; Miles & Hempel, 2003; Rothbart & Bates, 1998; Diener & Lucas, 1999) and extraversion to have been linked to positive affect of subjective well-being (Watson and Clark, 1984; Pavot, Diener and Fujita, 1990; Diener et al., 1992). A meta-analysis published by DeNeve and Cooper (1998) reported that out of all 137 personality factors that have been studied by the researchers, neuroticism has been found to have shown the strongest correlation with subjective well-being. Lucas and Fujita (2000) reported in their meta-analysis that extraversion is strongly correlated with positive affect of subjective well-being.
Social support is defined as ‘‘various forms of aid and assistance supplied by family members, friends, neighbors, and others’’ (Barrera et al., 1981, p. 435). Social support has several factors and these are social integration, enacted support, provided support and perceived support (Barrera, 1986).
Social integration is defined as the no of times an individual has made contact with the people in their social network (Barrera 1986; Granovetter 1985). Enacted support is defined as the actual support that an individual has received from other people (Barrera, 1986). This support could be informational, emotional or tangible (Barrera, 1986). Provided support is described as the support that one could provide to others (Barrera, 1986). This support could be informational, emotional or tangible (Barrera, 1986; Brown et al. 2003; Lu 1997). Perceived support is described as how satisfied one is with the expected support and the support that they have received from other people (Siedlecki, Salthouse & Jeswani, 2013).
Humans are social animals and they have a need to create and maintain good quality social relationships with other people (Baumeister & Leary, 1995). Many researchers have shown strong positive links between social support and subjective wellbeing of individuals (Myers, 2000; Lyubomirsky, King, & Diener, 2005) as close friends, family and good social networks have positive effects on individuals health and subjective wellbeing (Wilson, 1967; Wethington and Kessler 1986; Berkman and Syme 1994; Mastekaasa, 1994; Glenn, 1996; Hendrick & Hendrick, 1997; Diener et al., 1998; Kristenson et al. 1998; Myers, 2000; Reis & Gable, 2003; Diener & Seligman, 2002; Miller et al. 2004).
In addition, research suggests that in social processes, the personality of the individual, who is receiving social support, plays a huge role (Pierce et al. 1997) and that personality traits of the individuals drive their social and interpersonal behaviours (Pierce et al. 1997), for example, extroverts are more likely to seek social support, make frequent contact with their friends and family and participate in social activities which makes them more likely to have less psychological distress as social support mediates the relationship between personality and psychological distress (Krause, Liang, and Keith 1990; Larsen and Ketelaar 1991; McCrae and Costa 1991; Russell et al. 1997; Lee, Draper and Lee, 2001). Diender and Seligman (2002) found in their study that people who engage in social activities regularly are the happiest. Fleeson, Malan and Achille (2002) conducted an experiment where college students (introverts and extroverts) were asked to engage in ‘extroverted’ activities, record these activities and how they felt, after engaging in each activity, in a diary. They found that students who engaged in extroverted activities were happier than those who did not (Fleeson, Malan and Achille, 2002). Both studies support the idea that social support and engagement in social activities improve individuals’ life satisfaction as both introverts and extroverts who participated in social activities were happier than those who did not (Diender and Seligman, 2002; Fleeson, Malan and Achille, 2002). Moreover, research suggests that extroverts have a high level of subjective well-being as they are more likely to socialise, less likely to have negative encounters, more likely to seek support and engage in social activities and more likely to have less psychological distress (Yun Dai and Bidjerano, 2007).
However, individuals who score high on neuroticism have more negative interactions, prefer not to participate in social activities, avoid social situations and are less likely to seek social support (Hotard et al. 1989; McCrae and Costa 1991; Watson and Clark 1992; Russell et al. 1997; Finch 1998; Berry, Willingham, and Thayer 2000).
Personality, social support and subjective well-being has been explored in the literature by many researchers. Researchers have made efforts to define and explain relationships that exist between personality, social support and subjective well-being. However, majority of the research, that has studied these variables together, is outdated and the data has not been collected from the UK participants. Moreover, there is barely any research articles that have explored the relationships between personality, social support and subjective well-being in the customer services sector. Finally, no direct article was found that has shown social support as a mediator between personality and subjective well-being. This research would try to bridge these gaps and update the literature on personality, social support and subjective well-being.
A cross sectional study design was used to collect quantitative survey data for this study. The survey questionnaire measured extraversion and emotional stability as independent variables, social support (significant other, family, friends) as a mediator and subjective wellbeing (life satisfaction, positive affect, negative affect) as a dependent variable.
In the present study, 571 (Male = 205, Female = 366) participants were recruited online. However, out of 571 survey participants, 140 did not meet the participation criteria. Out of 140 participants, who did not meet the participation criteria, 100 participants had tried taking the survey from countries other than the UK, 35 participants did not answer any question and the remaining 5 did not answer questions related to their subjective wellbeing. These participants were removed from the study.
After removing the unqualifying participants from the current study, a total of 431 (Male = 153, Female = 278) participants, who work in the UK customer services (retail) sector employees, were included in the analysis. It was decided to focus on the UK customer services (retail) sector as majority of the previous research has been carried out in other countries such as the USA. Furthermore, majority of the research that has been conducted is outdated and focus of the previous research studies has been on a wide range of different sectors such as the healthcare and the education sector.
In the current study the participants’ age ranged from 18 to 71. Several criteria were used to calculate the total number of participants that were required for this study. Green’s (1991) formula was used first in order to calculate the sample size that would produce a statistically significant outcome. Green’s formula (N > 50 + 8m – where m is the number of independent variables) suggested that a minimum of 66 participants would be required for this study. However, as Mplus requires a large sample size, it was decided to use other methods as well. Survey monkey and Check Mark were used next. These are popular online websites that are used worldwide for data collection purposes. In order to find the sample size, the information on population size, margin of error and confidence level was required. The population size of the sector, that the current study has focused on, was estimated to be 2.9 million in October 2017 (Telegraph, 2017), margin of error 5% and the confidence level 95%. Both websites suggested that a total of 385 participants would be required for this study. It was decided not to use G Power as it cannot handle complicated models.
Finally, academic research studies have suggested that determining the correct sample size that is required for structural equation modelling (SEM) is not straightforward (Wolf, Harrington, Clark, Miller, 2016). It has been suggested that if the data is non-normally distributed, as the data is skewed in this study, and there are no missing values, the sample size should be at least 265 (Muthen and Muthen, 2009). Furthermore, research has suggested that for SEM the sample could be anywhere between 30 – 465 participants depending on the model, missing values and whether or not the data is normally distributed (Wolf et al., 2016). The sample size in this study is large (431 participants) with no missing values thus making the use of SEM suitable for this study.
A wide range of different issues were considered while creating the survey questionnaire for the current study. The researcher made sure that the measures, that are used in this study, are valid, reliable, widely used and well-established. The researcher made sure that the language that is used is clear, simple and easily understandable and that the survey items capture the information that is required to achieve the aims and objectives of the study. Furthermore, practical costs (e.g. time and money) for both the researcher and the participants were considered while designing the survey questionnaire.
The measures that are used in this study are extraversion, emotional stability, social support (significant other, family, friends) and subjective wellbeing (life satisfaction, positive affect, negative affect). A set of basic demographic questions such as age, gender were also asked to the survey participants. A copy of the full questionnaire can be found in Appendix 1.
Multidimensional Scale of Perceived Social Support (MSPSS)
Perceived social support was measured using a, free, brief 12 items multidimensional scale of perceived social support (Zimet, Dahlem, Zimet and Farley, 1988). The scale measures the perceptions of social support that individual might receive from three different sources. These are significant other, friends and family. Each sub-scale of social support has 4 items. The scale has a good validity, good test-retest and internal reliability and a stable factor structure across different samples (Zimet et al., 1988; Zimet, Powell, Farley, Werkman and Berkoff, 1990; Mitchell and Zimet, 2000; Nakigudde, Musisi, Ehnvall, Airaksinen and Agren, 2009). The scale has been adapted and translated into many different languages and is widely used in the academic literature (Nakigudde et al., 2009). The response to a question is measured on a 7-point Likert scale ranging from 1 (very strongly disagree) to 7 (very strongly agree). The cronbach’s alpha (α) was calculated in SPSS (version 23.0) for the full scale and its individual sub-factors. The cronbach’s alpha for the full scale was .94. The cronbach’s alphas (α) for the subscales were .94 for significant other, .93 for family and .94 for friends.
Satisfaction with Life Scale (SWLS)
Life satisfaction was measured using a, free, 5 items satisfaction with life scale (α = .89) (Diener, Emmons, Larsen and Griffin, 1985). The satisfaction with life scale only measures one of the three components of subjective wellbeing (Diener et al., 1985). The scale measures participants’ life satisfaction as a whole (Diener et al., 1985). The response to a question is measured on a 7-point Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree). The scale has a good validity and reliability and is widely used, in diverse populations, in the academic literature (Diener et al., 1985; Pavot and Diener; 1993; Lopez-Ortega, Toress-Castro and Rosas-Carrasco, 2016).
Scale of Positive and Negative Experience (SPANE)
Positive and negative affect, which are the other two components of subjective wellbeing, were measured using a, free, brief 12 items scale of positive and negative experience (α = .94) (Diener and Diener, 2009). The scale of positive and negative experience has two subscales (positive affect and negative affect) that measure 6 items each (Diener and Diener, 2009). The cronbach’s alpha for positive affect is .88 and .85 for negative affect. The response to a question is measured on a 5-point Likert scale ranging from 1 (very rarely or never) to 5 (very often or always). The scale of positive and negative experience has a good validity, reliability and a stable factor structure in diverse populations (Diener and Diener, 2009; Silva and Caetano, 2013).
Ethical issues were considered for this project as Diener and Crandall (1978) have recommended that a researcher must not harm their participants and their identities must be kept confidential. The ethical concerns for this study were extremely low. However, the researcher’s student email and their supervisor’s email were provided to the survey participants if they required any further information about the research project. Data collection for this study did not begin until the British Psychological Society (BPS) guidelines were followed and full ethics approval was received from The University of Manchester’s Ethics Committee.
The aims and objectives of the current study were explained to the survey participants and participants were encouraged to read and understand this information carefully before giving their consent. Participants’ details and responses were secured on a password protected computer and they were assured that their responses will not be shared with anyone and only the researcher and their supervisor will have access to the survey data. Participants were asked to create a participant code that allowed the researcher to anonymise their details. At no stage participants were asked to provide the researcher with their name or any other personal details. Participants were advised that they may wish to take a break or withdraw from the study at any time without providing a reason to the researcher. They may close their browser or send an email to the researcher should they wish to withdraw from the study. Participants were allowed to skip a question should they wish to not answer a question for any reason. Although the ethical concerns or the potential harm of the current study was low, counselling details of several organisations including The Samaritans, MIND, BACP and the NHS were provided to the survey participants. Finally, all survey participants were thanked for their time.
The survey questionnaire was created using Qualtrics. Once the questionnaire was created, a link was generated and posted on Facebook, Twitter and LinkedIn. The survey questionnaire was posted in relevant social media groups and participants were informed that the participation would be voluntary, however, they must be over the age of 18 and working in the UK customer services (retail) sector. The data collection process took place over a period of 45 days. During this period, potential participants were sent constant reminders on social media. The survey participants were informed what each section of the survey was trying to measure and what the overall aims and objectives of the study were. Finally, once the data collection had stopped, it was saved as a ‘.csv’ file and SPSS (version 23.0) was used to open the data file.
Data analysis strategy
A basic set of assumptions were tested first using SPSS (version 23.0). These include an assumption of linearity, multicollinearity, outliers, homogeneity, homoscedasticity, skewness and kurtosis. The assumption of linearity was tested and met as the relationships between the IVs and the DVs were linear and not curved. The data set did not have any multicollinearity issues suggesting that the variables in the data set were not highly related, and no independent errors were observed thus the assumption of multicollinearity was met (Coakes, 2005). No outliers were observed in the data set. The assumptions of homogeneity and homoscedasticity were tested and satisfied (Coakes, 2005). Finally, skewness and kurtosis were tested.
The test for kurtosis suggested that the data set was not kurtotic. The test for skewness suggested that the data set was slightly skewed. Research suggests that if all other assumptions are met and the sample size is large (N > 200), skewness should not affect our results as skewness only effects results in smaller sample size ( ). However, in order to avoid any potential challenges or problems, Weighted Least Squares Means and Variance (WLSMV) was used as an estimator in MPlus (Muthen and Muthen, 1998 – 2015). Weighted Least Squares Means and Variance (WLSMV) is a powerful estimator that works better than Maximum Likelihood (ML) when a Likert-type response format is used and the data set is discontinuous and non-normally distributed (Flora and Curan, 2004; Proitsi et al, 2009; Booth and Hughes, 2014; Brown, 2014).
Structural Equation Modelling
As recommended by Anderson and Gerbing (1988), a two-step approach was taken to Structural Equation Modelling in MPlus. In the first step, Confirmatory Factor Analysis (CFA) was carried out in order to verify the factor structure of the used variables (Joreskog, 1969; Bollen 1989; Hughes and Irwing, 2018). In the second step the structural relationships between the constructs (extraversion, emotional stability, social support, subjective wellbeing) were tested and determined in Mplus (version 7.4). The structural relationships between the constructs were guided by previous research evidence. The structural relationships between the constructs were determined by testing and comparing model fit indices criteria and the variance explained by each model.
In order to verify the factor structures of the used variables, model fit indices were assessed. For this assignment, model fit criteria were selected that are most commonly used in the academic research (see Hu and Bentler, 1998, 1999). These include Chi-square (χ2) goodness-of-fit, Comparative Fit Index (CFI), Tucker-Lewis Index (TLI) and Root Mean Square of Approximation (RMSEA). These criteria give a good indication of whether a model would fit. However, if the sample size is large, Chi-square can produce significant results suggesting that the observed model is different from the model implied as it is sensitive to large samples (Hair et al., 2006). As a result, it was decided to use multiple model fit indices (CFI, TLI and RMSEA) to asses a model fit.
In the academic literature, different researchers have argued what cut-off points, for model fit indices, should be used (see Chen et al, 2008). For this research project, it was decided that the cut-off points for the fit indices should be: ≥ .90 – .95 for CFI, ≥ .90 – .95 for TLI and ≤ .08 for RMSEA (Moosbrugger et al., 2003; Hu and Bentler, 1998, 1999).
The factors’ structures of all scales, that are used in this study, have been verified by many researchers in the past. However, as the purpose of this study and the sample collected for this study is different from previous research studies, it was decided to use a Confirmatory Factor Analysis (CFA) in order to test whether the factors’ structures, that have been well-established, are corresponding with the current sample (Joreskog 1969; Bollen, 1989). The purpose of carrying out a confirmatory factor analysis is to test and establish that the survey items accurately measure their underlying latent variables. Other data analysis techniques and software such as SPSS were not used as the relationship between the latent variables are better represented using SEM and the measurement error variance is accounted for which is a problem with other methods as they do not account for this. Furthermore, SEM allows to test multiple relationships or an entire model simultaneously (Hair, Black, Babin, Anderson, Thatham, 2006).
Confirmatory Factor Analysis was carried out in multiple steps using Mplus (version 7.4). The reason for this was that improvements could be made and problems could easily be identified and resolved if there were any issues with the scales. In the first step the factor structure of each scale was verified independently. The researcher did not experience any issues with the extraversion, emotional stability, SPANE (Scale of Positive and Negative Experience) and the life satisfaction scale. The factor structures of these scales were verified and the model fit indices criteria were met (see Appendix 1). However, the social support scale, that has 12 items, did not meet the model fit indices criteria and the model fit was poor (χ2 = (54) 1805.607, p < 0.001; CFI = .91; TLI = .89; RMSEA (90% C.I) = .27 (.26 – .29).
There are many different options that a researcher may decide to use in order to improve the model fit indices. However, they must justify which option is being used and why. A researcher may delete items with the (most) large modification indices. They may allow cross-factor loadings. They may decide create parcels. They can correlate the variances of the survey items. They may decide to split the factor.
In the present study, the researcher decided not to implement parceling as the use of parceling has been considered inappropriate by many researchers especially when the model fit indices are not met in the first place (Marsh and O’Neil, 1984; Yang et al., 2010, Ludtke, Marsh, Morin and Nagengast, 2013). Furthermore, the researcher did not experience any issues with the use of Mplus thus the use of parceling was considered inappropriate. In the present study, the social support factor was improved by splitting the factor into three sub-factors as many researchers have suggested that it has three sub-factors (significant other, family, friends). The factor structure, of these sub-factors, were assessed independently and the model fit indices criteria were met (see Appendix 1). In addition, previous researchers have suggested that the scale of positive and negative experience has two sub-factors (positive affect and negative affect). As it is not best practice to ignore recommendations that have been made by previous researchers, a decision was made to split the SPANE scale. The factor structure, of these sub-factors, were verified separately and the model fit indices were met (see Appendix 1).
In the second step, full scales (extraversion, emotional stability and life satisfaction) and sub scales of social support (significant other, family, friends) and SPANE (positive and negative affect) were added one by one in order to create a full CFA model. The model fit indices were checked each time a new scale was added on. At this stage, no issues were experienced by the researcher at this stage and the model fit indices criteria were satisfied (see Appendix 1).
In the final step, full Confirmatory Factor Analysis (CFA) model was tested in MPlus. The full CFA model met the model fit indices criteria (χ2 = (1099) 2536.087, p < 0.001; CFI = .95; TLI = .95; RMSEA (90% C.I) = .05 (.05 – .06). The Confirmatory Factor Analysis (CFA) confirmed that all factor loadings were significant at p < 0.001 and high. The Squared Multiple Correlations (SMC) and the pattern matrix is presented in Table 1.
Table 1. Measurement model items, Factors, Loadings and Squared Multiple Correlations (SMC)
|Am the life of the party.
Feel comfortable around people.
Talk to a lot of different people at parties.
Don’t mind being the center of attention.
Don’t talk a lot.
Keep in the background.
Have little to say.
Don’t like to draw attention to myself.
Am quiet around strangers.
|Am relaxed most of the time.
Seldom feel blue.
Get stressed out easily.
Worry about things.
Am easily disturbed.
Get upset easily.
Change my mood a lot.
Have frequent mood swings.
Get irritated easily.
Often feel blue.
|There is a special person who is around when I am in need.
There is a special person with whom I can share my joys and sorrows.
I have a special person who is a real source of comfort to me.
There is a special person in my life who cares about my feelings.
My family really tries to help me.
I get the emotional help and support I need from my family.
I can talk about my problems with my family.
My family is willing to help me make decisions.
My friends really try to help me.
I can count on my friends when things go wrong.
I have friends with whom I can share my joys and sorrows.
I can talk about my problems with my friends.
In most ways my life is close to my ideal
The conditions of my life are excellent
I am satisfied with my life
So far I have gotten the important things I want in life
If I could live my life over, I would change almost nothing
Note. N = 431. SMC = Squared Multiple Correlations. All factor loadings are significant (p < 0.001).
The table above shows that all factor loadings are high, suggesting that the items used in this study are accurate and reliable. All scales, that have been used in this study, are reliable and valid. All scales have been validated and their factor structure has been verified by previous researchers. However, it was decided to calculate the cronbach alphas for verification purposes. The factors correlations, reliability coefficients (cronbach alpha), mean and standard deviation are presented in Table 2.
Table 2. Zero order correlations, construct reliability for the CFA model.
Note. N = 431. ** = p < .001
The degree of change between variables is determined by a correlation (Creswell, 2002). A correlation can range between -1.00 – +1.00 (Cooper and Schindler, 2014). According to Cohen (1988), the magnitude of a correlation is small if it is up to .3, it is medium if it is between .3 and .6 and it is large if it is above .6. It can be seen in the table that all factors are reliable and have high correlations, however, they are all specific. It can be seen that extraversion has a small positive correlation with significant other (r = .19, p < .001) and family (r = .20, p < .001) but a moderate positive correlation with friends (r = .39, p < .001), positive affect (r = .48, p < .001) and life satisfaction (r = .40, p < .001). On the other hand, emotional stability has a large positive correlation with negative affect (r = .75, p < .001), moderate negative correlation with life satisfaction (r = -.53, p < .001) and a small negative correlation with significant other (r = -.21, p < .001), family (r = -.27, p < .001) and friends (r = -.24, p < .001). All aspects of social support (significant other, family and friends) have moderate positive correlations with positive affect and life satisfaction and can be seen in the table above.
In this step, the structural relationships between extraversion, emotional stability, social support (significant other, family, friends) and subjective well-being (life satisfaction, positive affect, negative affect) were explored, and the role of social support as a mediator between personality (extraversion and emotional stability) and subjective wellbeing was tested. These structural relationships were guided by previous research studies ( ). In order to establish and examine these structural relationships between constructs, two models (see Figure 1 and Figure 2) were tested in Mplus: (a) A fully mediated model (b) A partially mediated model. These models were compared by comparing the variance explained by each model and their model fit indices (see Table 3).
Figure 1 – A fully mediated model
Figure 2 – A partially mediated model
Table 3. Fit indices for compared models.
|Model||df||χ2||CFI||TLI||RMSEA (90% C.I)||Variance accounted for
Well Being Support
|1||A fully mediated model||1117||4133.901**||.90||.90||.08 (.08 – .08)||94.7% 48.8%|
|2||A partially mediated model||1115||2956.228**||.95||.95||.06 (.06 – .07)||83.7% 15.5%|
The analysis showed that both, partially and fully mediated, models suggest an acceptable fit to the data [a fully mediated model = χ2 = (1117) 4133.901, p < 0.001; CFI = .90; TLI = .90; RMSEA (90% C.I) = .08 (.08 – .08) and a partially mediated model = (χ2 = (1115) 2956.228, p < 0.001; CFI = .94; TLI = .94; RMSEA (90% C.I) = .06 (.06 – .07). However, a partial mediated model showed a slightly better fit to the data. Furthermore, as all direct and indirect pathways from extraversion to social support [β2 = .27**], extraversion to subjective wellbeing [β2 = .14**], neuroticism to social support [β2 = -.21**] and neuroticism to subjective wellbeing [β2 = -.50**] were significant, partial mediation was supported.
In the final step of structural equation modelling, the model fit indices and the variance explained by each model was compared to each other. Although the fully mediated model explained more variance in subjective wellbeing (94.7%) and social support (48.8%) that the partially mediated model, the partially mediated model had a better fit and theoretically made the most sense. the fully mediated model, it explained less variance in subjective wellbeing and social support. Thus, a decision was made that the fully mediated model is better than the partially mediated model as it explained more variance in subjective wellbeing (94.7%) and social support (48.8%).
Cite This Work
To export a reference to this article please select a referencing stye below:
Related ServicesView all
DMCA / Removal Request
If you are the original writer of this dissertation and no longer wish to have your work published on the UKDiss.com website then please: