The relationship between job satisfaction and performance at work: case of Tunisian accounting firms

Article history: Received February 2 2019 Received in revised format February 27 2020 Accepted March 8 2020 Available online March 8 2020 Research on the impact of satisfaction on job performance has proliferated over the last few years; however, the results are deemed inconclusive. Here, a total of 240 accounting professionals participated in this study through a questionnaire survey. The four career stages were integrated into our model as a moderating variable of the relationship between job satisfaction and performance. The results prove that the further we move through career stages, the stronger the relationship between satisfaction and performance at work is. by the authors; licensee Growing Science, Canada 20 © 20


Introduction
The effect of job satisfaction on performance has been a debatable topic in the previous literature (Raza et al., 2015). Over the last few decades several research works have addressed this topic; however, the results are mixed. Most of the previous studies were conducted on teachers (Indhumathi, 2011), doctors (Khan et al., 2012), nurses (Platis et al., 2015) and employees in general (Almutairi et al., 2013). However, to date, the accounting profession has not attracted much interest. Accordingly, the present study aimed to fill this gap by studying the impact of job satisfaction on the performance of accounting professionals in the Tunisian context. The remainder of the paper is organized as follows: Section 1outlines the literature review that led to the formulation of the hypotheses to be tested. The methodology of our research is presented in section 2. Section 3 discusses the results. Finally, we conclude the paper by emphasizing the academic and managerial implications of our research as well as its limitations.

Literature review and the research hypotheses
The relationship between satisfaction and performance is based on two theories, namely: the social-cognitive theory and the theory of social exchange. According to Ajzen (1991), the attitude of the individual towards the profession (i.e., job 232 satisfaction) influences his behavior (i.e., performance). The latter asserts that the more unsatisfied the individual is, the inferior his performance (Organ, 1988). Thus, the interconnection between job satisfaction and performance at work has long attracted extensive interest but the results have been mixed. For example, Brayfield and Crockett (1955) concluded that there is no noticeable direct relationship between satisfaction and performance. Furthermore, Locke (1976) conducted an in-depth review of the literature dealing with satisfaction and concluded that job satisfaction has no direct effect on productivity. Nimalathasan and Brabete (2010) found that there is a positive but non-significant relationship between both variables. However, Indhumathi (2011), through a survey on 444 secondary school teachers, found that there is a significant relationship between job satisfaction and teachers' performance in different school categories. Similarly, Platis et al. (2015) carried out an analysis in the National Center of Public Administration and Local Authorities during a training session on 246 staff members (nurses) and found that there is a positive and significant relationship between job satisfaction and performance at work. In the same vein, Khan et al. (2012) examined factors that may affect the level of job satisfaction and their impact on performance using a sample of 200 doctors, nurses and administrative personnel who work in autonomous medical institutions in Punjab. They showed that salary, advancement, job security, working conditions, relationships between colleagues and with the managers along with the nature of the work influence job satisfaction and performance at work. Another study was conducted by Prasanga and Gamage (2012) on the impact of job satisfaction on performance among 219 seafarers in India. Their results showed a positive and significant relationship between the two variables. Aftab (2012) tried also to explain the nature of the relationship between satisfaction and performance among 335 employees in the banking sector and concluded that there is a positive relationship between the two. Almutairi et al. (2013) found a positive and significant relationship between satisfaction and job performance in a study conducted on a sample of 91 employees working in five-star hotels in Riyadh, Saudi Arabia. In the same vein, Dizgah et al. (2012) conducted a study on the relationship between job satisfaction and performance at work in Iran's public sector using a sample of 323 individuals. Their results showed that there is a significant relationship between the two variables. Hence our first hypothesis: Hypothesis 1: There is a positive and significant relationship between job satisfaction and performance at work.
Career development was integrated as a moderating variable to further analyze the relationship between job satisfaction and performance at work.

The impact of the career stages on satisfaction-performance relationship
Recently, the notion of career development has gained momentum and has been used extensively in explaining the impact of the working environment on professionals' behavior. Indeed, the career development theory helped emphasize the various conceptual models that identify not only the different phases of the professional career development but also their impact on the professionals' behaviors and attitudes.
According to Super (1957), a professional employee goes through four stages, namely the exploration, the establishment, the maintenance and the disengagement stages, in his professional career. The transition from one career stage to another may change a professional's behavior. It is thus more convenient to examine the relationship between the professional's attitudes using a moderating variable, namely career (Luttmann et al., 2003). In fact, previous studies dealing with the relationship between job satisfaction and career development stages have reached mixed results. Some researchers found that the level of job satisfaction rises as one moves along the career stages, i.e. from the exploration phase to the disengagement one (Slocum and Con, 1985;Con and Slocum, 1986;Ornstein et al., 1989). Several studies in the field of accounting have dealt with the effect of career stages on job satisfaction. For example, Alder and Aranya (1984) found that the level of intrinsic and extrinsic job satisfaction increases during the first three years of the career and then decreases. Rebele et al. (1996) found that the auditors' level of job satisfaction is very low during the exploration phase and very high over the subsequent phases. As a result, the further one advances in his career, the higher his job satisfaction becomes. This may be due to the progressive and optimistic improvement in viewing the work environment. As a result, this enhancement in the level of satisfaction boosts the performance of professionals along their career development phases. Nevertheless, several researchers, such as Churchill et al. (1976), Mount (1984) and Luttmann et al. (2003), did not find a significant relationship between career phases and the level of job satisfaction. In the same vein, Norris and Niebuhr (1983) studied the relationship between job satisfaction and career development phases among the accounting professionals belonging to the "Big Eight" U.S. accounting firm and found no significant correlation. Stumpf and Rabinowitz (1981) used three career development phases, namely establishment, advancement and maintenance, to study the relationship between the performance indices and the job satisfaction facets, role ambiguity and role conflict. Using a sample of 102 full-time faculty members, the authors showed that career stages have an important moderating effect on the aspects of job satisfaction as well as the relationships between role perception and role-related performance. In the same vein, Gould and Hawkins (1978) found a positive and significant correlation between performance and job satisfaction during the establishment and maintenance phases. In addition, by examining the moderating effects of the career stage on the relationship between job complexity, Age Seniority Position satisfaction and performance at work, Gould (1979) found that job satisfaction is more strongly related to perceived job complexity in the exploration stage than in the establishment or maintenance phases. Indeed, new hires who face recurring career issues are considered to be underperforming. In the light of the aforementioned considerations, we formulate this sub-hypothesis as: Hypothesis 1-1: The relationship between job satisfaction and performance varies according to the different career development phases.
Our research model ( Fig. 1) is presented as follows:

The method of data collection
The sample used in the present work consisted of 240 professionals working in accounting firms in Tunisia.

Variables Measurement
The first variable of our model dealing with performance at work (variable to be explained) was measured by 12 items according to Choo (1986) and Fisher (2001).The second variable consisting in job satisfaction (explanatory variable) was measured by four items developed by Hong and Waheed (2011).The third variable focused on the career stages (moderating variable) consisting of 4 steps measured by 12 items used by Kristin et al. (2003). Finally, the control variables, which were measured according to a disjunctive table are three-fold, namely: age of the professional, seniority of the professional and position at firm level.

Data analysis methods
Before processing the questionnaire data using the SPSS statistical software PCA (Principal Component Analysis (Lebart et al., 2000), we codified all the variables in a well-structured database to facilitate the necessary calculations, then we used the

Performance at work Job satisfaction
The career stages (Exploration, establishment, maintenance and disengagement) 234 interactive approach using Partial Least squares (PLS) (Lohmoller, 1989;Jöreskog & Wold, 1982;Lohmoller, 1989;Tenenhaus, 1998). The purpose of this section is to describe the descriptive analyses and the homogeneity tests of the constructs. First, we carried out crossed tabulations of the socio-demographic characteristics. Then, we reviewed the validity and reliability concepts through a factorial analysis and the Cronbach alpha test, to aggregate and, if necessary, adjust the items that make up the different scales of the questionnaire. Before starting the necessary analyses for the validation of the hypotheses, a few comments should be made. Regarding the factor analysis, if the (orthogonal) initial Principal Component Analysis (PCA), as is the case for some scales, does not show a clear factor structure due to items having contributions greater than 0.30 to several factors, it is recommended to introduce a rotation to adjust the proposed structure. This rotation would help us interpret the factors that increase the value of the correlation coefficients of some items with the new axes of representation. The items of each scale initiating the constructs that have been defined previously are not completely independent from one another. Therefore, the oblique rotation is maintained because it has the advantage of accounting for situations where several factors are composed of items that contribute to more than one factor. The second level of the analysis focuses on testing the moderating effect of the "career stages" on the level of job satisfaction and performance at work. The identification of the moderating role of each career stage has to be based on a specific procedure. Several groups of observations should be established. In practice, the median is often used, accordingly the question is to identify the significance and value of the correlation between "job satisfaction" and "performance at work" in each of the formed groups. Thus, on the basis of the median value of every career stage "Exploration-establishment-Maintenance-disengagement", we could establish two groups of observations.

Career stages
The first PCA using all the items reached mixed results. It was required to retain three axes that explain 71.83% of the variance. The share of each variable on the different items is unclear: for example, "finding the most appropriate working experience" has a contribution greater than 0.30 (0.339 and 0.313) on two different axes (2 and 3), respectively. As a consequence, we executed a rotation of the axes, which helped us find a more suitable structure. Subsequently, we compared these results to those of the PCA conducted on every career stage. Therefore, some significant axes have been identified. Indeed, the analysis of the exploration stage shows that this variable is one-dimensional (with 35.3% of the total explained variance) and can be condensed by the items: "Finding the line of work that I'm best suited for, Finding a line of work that interests me, Getting started in my chosen career field". The establishment stage also has a one-dimensional structure (for 19.87% of the total explained variance) and may be restored by "being stable in a job in which I prefer to become especially competent, talented and plan my promotion at work". Similarly, the maintenance stage is a one-dimensional scale (with 16.6% of the total explained variance). Table 2 below presents an in-depth analysis of the career of this scale by executing an Oblimin rotation. The interpretation of the three factorial axes becomes more significant and reinforces the three previous dimensions. The matrix clearly presents a triple facet of the career stages of the accounting professionals (Table 2). Therefore, we can remove the disengagement stage.

Performance at work
Before deciding to combine the 12 items, we carried out the internal consistency test. The α score equal to 0.621 is acceptable.
However, Table 3 shows that the elimination of the item "maintaining the amount of work" can improve this score by 0.1 point (0.723). In fact, performance at work is expressed through two dimensions: task performance and contextual performance. The former is based on the performance of tasks, such as maintaining quality, communicating orally, communicating in writing, accepting responsibility, initiating action, exercising professional skills and care as well as adapting to new or different job situations. The latter consists of following policies and procedures, planning and organizing work, getting along with others within the firm, dealing with clients outside the firm and supervising others. Therefore, the scale we built should lead to a twodimensional factor structure. In fact, the original PCA within the sample revealed two factorial axes. However, the matrix of the components is not clear. The two items «Accepting responsibility» and «following policies and procedures» have factorial contributions greater than 0.30 to the two dimensions. In order to interpret these factors more clearly, the angular rotation was necessary. This rotation clearly showed the links between both dimensions of performance at work and their components because the factorial contributions have higher values (see Table 2). With a contribution greater than 0.7 and no contributions greater than or equal to 0.30 on another factor, all the items contribute strongly and essentially to the dimension they compose. Moreover, each axis restores almost a third of the collected data and the total of their variance is equal to 64.38%. Thus, the scale representing the variable to be explained of the model that we aim to test is homogeneous. (Table 3). Then, we examined whether the correlations between the constructs and their measurement variables differ from one group to another, therefore justifying, if required, the use of a constrained multi group analysis (MGA). Finally, a constrained or unconstrained MGA was carried out in order to test the moderating impact of the "career stage" on the intensity of the studied relationship. This impact is emphasized when we observe differences in the correlation values between the formed groups. To this end, the comparative perspective of the research enabled us to simultaneously distinguish three samples during the calculations; the first concerns seniority at work below 2 years (inexperienced), the second deals with seniority between 2 and 10 years (quasi experienced) and the third is related to seniority at work over 10 years (experienced). Before developing homogeneity tests, which are essential for the hypothesis validation, it is appropriate to specify the properties of the data collected through the data sheet of the questionnaire. Therefore, we should describe the characteristics of the three samples through the following variables: age, previous background, seniority and position.

Age
The crossing of age and seniority reveals a difference of 2.2 points in favor of the quasi experienced sample (from 2 to 10 years). Indeed, in the detailed composition of age, this gap can be explained in two ways: The first is that the sample contains more individuals with an average age under 30, that is 52 (65%) versus 7 (8.8%) having a seniority between 2 and 10 years. The second is that seniority has a lower concentration of accountants between 30 and 45 years, 9 (11.1%) versus 35 (43.2%) having an experience between 2 and 10 years. (Table 4)

Position
The distribution by position is largely dominated by juniors and, at a lesser degree, managers. However, it was noticed that the inexperienced population (<2years) includes 57.7% of juniors and for a negligible proportion, it does not exceed 10.3% in the 236 quasi experienced population (between 2 and 10 years). The numerical superiority of juniors contributes to the homogeneity of the three populations, which can mitigate the possible effects of the specific career stages on the positions of the accounting professionals (Table 6). The moderating effect of the career stages on performance at work

The distribution per position is largely dominated by juniors and less frequently by managers
The moderating role of the exploration stage To make the mediating effects "visible", we used interaction-based regression following a series of statistical analyses. The model is written in the following form: Task Performance= α 0 +α 1 Satisfaction+ α 2 Exploration+ α 3 Satisfaction × Exploration +α 4 Age+ α 5 Seniority+ α 6 Position+ζ (M1).
Contextual performance = α 0 + α 1 Satisfaction+ α 2 Exploration+ α 3 Satisfaction × Exploration + α 4 Age+ α 5 Seniority + α 6 Position+ζ (M2)² The interaction between job satisfaction and the exploration stage on performance at work is modeled. The tested model is relatively well specified. The explanatory power indicators and Fisher statistics are satisfactory. As a result, our model, which explains 15.27% and 12.96% of the variance of respectively task performance and contextual performance, is acceptable. As expected, satisfaction significantly contributes to performance in twofold dimension. The "satisfaction × exploration" interaction effects on task performance (p= 0.56) ( Table 6) and contextual performance (p =0.845) ( Table 7) are not significant. If, in our context, satisfaction appears as the most important explanatory variable of performance at work, the exploration phase fails to play a moderating role in the model.  The moderating role of the establishment stage The second stage consists in testing the moderating role of the establishment stage on the intensity of the "satisfactionperformance" relationship. The model was written in the following form: Task performance=α 0 +α 1 Satisfaction+α 2 establishment+α 3 Satisfaction×establishmentt+α 4 Age+ α 5 Seniority + α 6 Position+ζ (M1) Contextualperformance=α 0 +α 1 Satisfaction+α 2 establishment+α 3 Satisfaction×establishment+α 4 Age+α 5 Seniority+ α 6 Position+ζ (M2) The results presented below show a positive and significant relationship between satisfaction and performance at 10% threshold, for task performance (Table 8) and at 5% threshold, for contextual performance (Table 9). The interaction between "satisfaction×establishment" and "contextual performance" has a satisfactory practical significance at 5% threshold. Therefore, the effect of this interaction on task performance proves that it is statistically non-significant. It can then be concluded that there is a partial moderating effect of the establishment stage on the relation between "job satisfaction and performance at work". Consequently, the establishment stage, which is the second career phase, appears to play a partial moderating role in the model.  The moderating role of the maintenance stage The last moderating relationship that should be empirically verified is connected to the effect of the maintenance stage on the "job satisfaction-performance at work" relationship. As can be seen in Tables 11 and 12 below, there is a satisfactory correlation between the different variables of the model and the double dimension of performance. The intensity of this relationship is estimated at 17.5% of the variance of the task performance (Table 10) and 20.07% of that of the contextual performance ( Table  11). The quality of fit of the relationship obtained by the regression is therefore acceptable and the link is significant. For the first regression, the observed value of the F coefficient (3.119 for a sig.=0.0059) is much greater than the critical value at 10% threshold. The test shows that there is a positive and significant correlation coefficient at 10% threshold (p=0.062) between the "job satisfaction×maintenance" and "task performance at work" variables. In a second analysis of contextual performance, the satisfaction-maintenance variable positively and significantly affects contextual performance (p=0.042) and, consequently, the moderating effect of the career stage in the "job satisfaction-performance at work" relationship is completely validated.
Task performance at work=α 0 + α 1 Satisfaction+α 2 Maintenance+ α 3 Satisfaction×Maintenance+α 4 Age+ α 5 Seniority+ α 6 Position+ζ (M1) Contextual Performance at work = α 0 + α 1 Satisfaction+ α 2 Maintenance + α 3 Satisfaction×Maintenance +α 4 Age+ α 5 Seniority + α 6 Position+ζ (M2)  It is worth noting that this result is in line with those of a previous work by William et al. (1986), whose results indicate that sellers belonging to their sample are very satisfied and more efficient in the establishment and maintenance phases than in the exploration and disengagement phases. In this context, some researchers found that the level of satisfaction at work increases while advancing in the career development stages, that is, from the exploration to the disengagement phase (Slocum & Con, 1985;Con & Slocum, 1986;Ornstein et al., 1989). This last finding is in line with the results reported by Alder and Aranya (1984) who found that the level of intrinsic and extrinsic job satisfaction rises during the first three years of the career and drops at the end of the career. Initially, it can be said that the results support Super's career development model (1959) since the attitudes at work differ greatly across the career stages. According to a distribution drawn from career concerns, it seems that all the studied attitudes are distributed over the professional trajectory in a curvilinear way. Therefore, regarding the attitudes at work, it was concluded that there are two classes of professionals; the new and the old (the exploration and the disengagement phases). The former has weak attitudes towards work while the latter (establishment and maintenance phases) has more positive attitudes towards their work. These findings led us to reconsider our research hypotheses. Therefore, our first hypothesis targeting job satisfaction seems to be partially improbable since individuals are less satisfied and less efficient at work during both the exploration and disengagement periods. Furthermore, they are notably less positive towards work, which is contrary to what has been previously formulated. In this regard, it should be noted that the vast majority of researchers who studied the relationship between the career stages and attitudes towards work used only three career stages, i.e. exploration, establishment and maintenance, which suggests that people's attitudes towards work during the disengagement period is still a little-known topic. In fact, far from presenting a positive linear relationship with the career development, attitudes towards work have a ceiling effect in the mid-career. In fact, the further we advance in the career stages, the stronger the relationship between job satisfaction and performance at work is. Moreover, when accounting professionals are satisfied at work, through intrinsic and extrinsic factors, they become more powerful. Therefore, these factors have an impact on the studied relationship with the exception of the disengagement phase.

Conclusion
To validate the hypothesis dealing with the moderating effect of the career stages, the exploration, the establishment and the maintenance stages were integrated into our model. To present this effect, we opted for the interactive approach using the Partial least squares (PLS) regression. In fact, it seems more relevant to explain the relationship between job satisfaction and performance at work based on the moderating effect of the establishment and the maintenance phases rather than the exploration one. This choice helped refine the moderating role of career stages in the relationship between job satisfaction and performance at work. Our results showed the importance of age, seniority and position in the relationship between job satisfaction and performance at work. We also found that the more we advance in the career stages, the stronger the relationship between job satisfaction and performance at work is, with the exception of the disengagement phase.
Our research has some limitations. The first is related to the reduced sample size and the second concerns the results which are contextual and cannot be generalized given the specificity of the Tunisian context. The third one relates to the imbalance in the constitution of our sample and in the selection of accounting professionals. It is hence necessary to neutralize the effect of this imbalance. Finally, new research lines can be traced by extending our research and conclusions, comparing our results with those of similar works in other contexts and limiting the analysis to a particular category of firms (to avoid the selection effect of professionals).