Research Article |
Corresponding author: Kris Hardies ( kris.hardies@uantwerpen.be ) Academic editor: Anna Gold
© 2023 Kris Hardies.
This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY-NC-ND 4.0), which permits to copy and distribute the article for non-commercial purposes, provided that the article is not altered or modified and the original author and source are credited.
Citation:
Hardies K (2023) A survival analysis of organizational turnover in the auditing profession. Maandblad voor Accountancy en Bedrijfseconomie 97(1/2): 5-16. https://doi.org/10.5117/mab.97.90216
|
This study uses survival analysis to examine employee turnover in the auditing profession. Building on the Job Demands-Resources model, I analyze the impact of job characteristics (demands and resources) and personal characteristics on organizational turnover. The study is based on a survey among a sample of 309 employees who either were or had been employed in the Belgian auditing profession. At any particular point in time, excess job demands (e.g., workload) increase and job resources (e.g., organizational support for alternative work arrangements) decrease the risk of organizational turnover. Higher family involvement (personal demands) is also associated with increased turnover risk.
Employee turnover, organizational turnover, turnover, survival analysis, auditing profession
It is important for audit firms to keep job demands at tolerable levels in order to retain their employees for the auditing profession. Workload is a stronger determinant of turnover in the auditing profession than in other contexts, but support for alternative work arrangements (AWAs) reduces the risk of employee turnover. Results may inform audit firms about employee retention.
Retaining highly qualified staff is a critical issue in professional service firms. This is especially true for auditing firms where organizational turnover peaks at 15–20% (
There is extensive research on employee turnover, both within the accounting literature (for a review, see
Specifically, I apply survival analysis to determine the survival rates of employees in their first job in the auditing profession and to examine factors related to their turnover. Survival analysis is a statistical technique that analyses the time duration until a specific event of interest happens, also referred to as “survival time”. In the present study, the event of interest is leaving the initial firm of employment. The survival time is the length of stay within the initial firm of employment (i.e., organizational turnover). This analytical approach has a number of advantages compared to more conventional analyses such as logistic regression or linear regression. Whereas logistic regression analysis only focuses on the occurrence of an event (i.e., leaving the firm or not), survival analysis also accounts for the time elapsed until an event occurs (i.e., the length of stay in the firm). The length of stay in the firm can be modeled as the dependent variable in a linear regression model. However, the sample contains responses from participants still employed at their first employer (and for whom there is thus not yet an exact survival time). In other words, the data are “right-censored”, as some respondents did not (yet) experience the event of interest (i.e., leave their firm). Survival analysis can adequately accommodate the loss of observations when censoring occurs, whereas traditional regression models cannot resolve this issue (
Finally, the current study focuses on how long audit firm employees stay at their first employers. Studying initial employment is important because audit firms play a crucial role in auditors’ identity formation (
I draw upon the Job Demands-Resources (JD-R) model to examine the effects of job characteristics (demands and resources) and personal characteristics on organizational turnover in the Belgian auditing profession. The JD-R model was originally developed to explain burnout (
According to the JD-R model (
Job demands are the physical, social, or organizational aspects of the job that require sustained physical and/or psychological effort or skills and are therefore associated with certain physical and/or psychological costs; job demands potentially evoke strain if they exceed the employee’s adaptive capability (
Hypothesis 1a: Perceived workload is positively associated with organizational turnover.
Hypothesis 1b: Work-life conflict is positively associated with organizational turnover.
First, auditing is characterized by high workloads, especially during the ‘busy season’ (
Second, as job demands may interfere with family demands, they can create work-life conflicts. The auditing profession is infamous for its high levels of work-life conflict, with auditors experiencing higher levels of work-life conflict than accountants in industry (
Job resources are the physical, psychosocial, social, or organizational aspects of the job that are either instrumental to achieving work goals, reducing job demands, or stimulating personal growth and development (
Hypothesis 2a: Job content (challenge and variety) is negatively associated with organizational turnover.
Hypothesis 2b: Perceived organizational support for alternative work arrangements is negatively associated with organizational turnover.
First, job characteristics such as challenge and variety of skills have been identified as one of the antecedents of organizational commitment in the early stages of a career (
Second, by offering alternative work arrangements (AWAs), audit firms can potentially reduce work-family conflict (
In addition to job demands and resources, the JD-R model acknowledges the potential role of personal demands and resources to affect organizational outcomes such as turnover. Personal demands are ‘the requirements that individuals set for their own performance and behavior that force them to invest effort in their work and are therefore associated with physical and psychological costs’ (
Hypothesis 3a: Careerism is positively associated with organizational turnover.
Hypothesis 3b: Career involvement is negatively associated with organizational turnover.
Hypothesis 3c: Family involvement is positively associated with organizational turnover.
First, individuals driven by “careerism” may leave their employer sooner. Careerism refers to the view that employment within an organization is merely a stepping stone toward a career in another firm (
Second, people differ not just in their “careerism” but also in the importance they ascribe to their careers (versus family). That is, they differ in terms of their career (family) involvement. Career (family) involvement refers to the degree to which someone identifies with their work (family) or the importance they place on their work (family) (
This study relies on a sample of 309 participants (Mean age = 34.26; SD = 9.13; 61.5% male) collected through an anonymous online survey developed for the purposes of this study. Participants were recruited in November-December 2015 by e-mailing all the persons in Belgium who had ever entered into at least one course of theoretical instruction organized by the Belgian Institute of Registered Auditors (IBR) as a prerequisite to enter the practical training to become a certified auditor.
A total of 3,708 persons were contacted by e-mail and invited to participate in the online survey. In total 376 persons who either were or had been employed in the auditing profession participated in this study, corresponding to a response rate of 10%. I excluded 67 participants due to missing data, resulting in a final sample of 309. Although the response rate is comparable to other studies on turnover in the auditing profession using online surveys (e.g.,
Organizational turnover was measured as the time between employment at the initial audit firm and the date when employment at the firm ended (“leavers”) or the date of completing the survey for respondents who were still employed at their first employer (“stayers”).
Perceived workload was measured with four items adapted from
Work-family conflict was measured with the work-family conflict scale of
Job content was measured using three questions from
Perceived organizational support for AWAs was measured with the scale developed by
Careerism was measured with the scale developed by
Career and family involvement were measured with eight items from
Demographic variables were included as control variables because prior research shows that demographic variables are related to turnover. Therefore, I measured and controlled for the effects of age, sex (0 = male, 1 = female), marital status (0 = no relationship, 1 = relationship), and parenthood (0 = no children, 1 = children). Additionally, I also controlled for the type of firm where the respondent was employed (0 = small firm, 1 = Big 4 accounting firm), as prior research suggests higher work-family conflict and burnout among Big 4 employees (
I employed survival analysis as this technique provides insight into why turnover occurs while accounting for the length of time prior to turnover (
The respondents are classified into two groups based on the occurrence of the event: leavers and stayers. For leavers, organizational turnover is measured as the number of years based upon their length of stay at their first firm of employment. For stayers, the data are right-censored and their organizational turnover is measured as the number of years based upon the time between the date they started working at their first firm of employment and the date on which they participated in the survey.
I analyzed the data using Cox proportional hazards survival analysis. The hazard rate function h(t) describes the conditional probability of an employee leaving their firm (i.e., organizational turnover). This probability depends on the employee’s organizational turnover.
First, data were cleaned and screened for missing data. I excluded data from 67 respondents: 53 who had more than 50% missing data and 14 who failed to complete any item on at least one scale (e.g., left the whole AWA scale blank). For the remaining data, a missing value analysis was conducted on all items of the scales. Given that Little’s MCAR test was non-significant (χ² = 2461.62, df = 2363, p > 0.05), we can assume that data are missing completely at random (MCAR). The expectation maximization (EM) algorithm was used to impute all missing values for the scale items. All analyses were conducted with these imputed values. The final sample consists of responses from 309 respondents.
Second, although all scales in this study were conceptually distinct and had been validated previously, I further tested the validity and internal consistency of the scales in the present study. Therefore, I conducted two factor analyses.
More than half of the respondents in the sample (52.8%) were still employed at their initial accounting firm at the time of the survey (i.e., they were stayers). Table
Full sample | Stayers | Leavers (firm) | Leavers (entirely) | |||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
N = 309 | n = 163 | n = 49 | n = 97 | |||||||||||
M | SD | min | Mdn | max | M | SD | Mdn | M | SD | Mdn | M | SD | Mdn | |
1. Organizational turnover | 6.90 | 6.69 | 0.08 | 4.42 | 34.67 | 7.95 | 7.41 | 5.25 | 6.83 | 7.13 | 4.00 | 5.17 | 4.54 | 3.58 |
2. Age (in years) | 34.26 | 9.13 | 23.00 | 31.00 | 70.00 | 32.77 | 8.79 | 30.00 | 41.65 | 11.15 | 41.00 | 33.05 | 6.46 | 31.00 |
3. Sex | 0.39 | 0.49 | 0.00 | 0.00 | 1.00 | 0.41 | 0.49 | 0.00 | 0.33 | 0.47 | 0.00 | 0.38 | 0.49 | 0.00 |
4. Marital status | 0.69 | 0.47 | 0.00 | 1.00 | 1.00 | 0.68 | 0.47 | 1.00 | 0.73 | 0.45 | 1.00 | 0.67 | 0.47 | 1.00 |
5. Parenthood | 0.29 | 0.46 | 0.00 | 0.00 | 1.00 | 0.31 | 0.46 | 0.00 | 0.41 | 0.50 | 0.00 | 0.22 | 0.41 | 0.00 |
6. Big 4 | 0.64 | 0.48 | 0.00 | 1.00 | 1.00 | 0.62 | 0.49 | 1.00 | 0.49 | 0.51 | 0.00 | 0.76 | 0.43 | 1.00 |
7. Career satisfaction | 3.84 | 0.58 | 1.60 | 4.00 | 5.00 | 3.91 | 0.54 | 4.00 | 3.75 | 0.53 | 4.00 | 3.78 | 0.67 | 4.00 |
8. Workload | 3.66 | 0.70 | 1.75 | 3.75 | 5.00 | 3.72 | 0.69 | 3.75 | 3.43 | 0.77 | 3.50 | 3.69 | 0.70 | 3.75 |
9. Work-family conflict | 3.29 | 0.91 | 1.00 | 3.40 | 5.00 | 3.25 | 0.91 | 3.20 | 3.25 | 0.96 | 3.40 | 3.38 | 0.88 | 3.40 |
10. Job content | 3.71 | 0.65 | 1.50 | 3.75 | 5.00 | 3.83 | 0.61 | 4.00 | 3.70 | 0.62 | 3.75 | 3.51 | 0.69 | 3.50 |
11. Support for AWAs | 2.75 | 0.88 | 1.00 | 2.75 | 5.00 | 3.13 | 0.70 | 3.25 | 2.36 | 0.94 | 2.25 | 2.30 | 0.82 | 2.25 |
12. Careerism | 3.25 | 0.84 | 1.00 | 3.40 | 5.00 | 3.14 | 0.78 | 3.20 | 2.96 | 0.86 | 3.00 | 3.56 | 0.82 | 3.80 |
13. Career involvement | 3.10 | 0.61 | 1.25 | 3.00 | 4.75 | 3.13 | 0.63 | 3.25 | 3.17 | 0.59 | 3.25 | 3.03 | 0.59 | 3.00 |
14. Family involvement | 3.67 | 0.64 | 2.00 | 3.75 | 5.00 | 3.66 | 0.65 | 3.75 | 3.71 | 0.63 | 3.75 | 3.68 | 0.64 | 3.75 |
Table
Table
1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
1. Organizational turnover | 1 | |||||||||||||
2. Age (in years) | 0.71** | 1 | ||||||||||||
3. Sex | –0.12* | 0.20** | 1 | |||||||||||
4. Marital status | 0.13* | 0.15** | –0.04 | 1 | ||||||||||
5. Parenthood | 0.59** | –0.56** | –0.09 | 0.35** | 1 | |||||||||
6. Big 4 | –0.12* | 0.20** | 0.07 | –0.04 | –0.05 | 1 | ||||||||
7. Career satisfaction | 0.07 | 0.00 | 0.12* | –0.03 | 0.02 | 0.07 | 1 | |||||||
8. Workload | –0.03 | 0.12* | 0.06 | –0.02 | –0.06 | 0.17** | –0.05 | 1 | ||||||
9. Work-family conflict | 0.03 | –0.01 | 0.08 | –0.07 | 0.01 | 0.16** | –0.09 | 0.65** | 1 | |||||
10. Job content | 0.18** | –0.15** | 0.26** | 0.07 | 0.21** | 0.04 | 0.46** | 0.09 | –0.07 | 1 | ||||
11. Support for AWAs | 0.25** | –0.05 | 0.15** | 0.08 | 0.25** | 0.11 | 0.16** | –0.13* | 0.16** | 0.24** | 1 | |||
12. Careerism | –0.21** | 0.29** | –0.29** | –0.04 | –0.29** | 0.15** | –0.13 | 0.10 | –0.04 | –0.26** | –0.13* | 1 | ||
13. Career involvement | 0.17** | –0.18** | 0.06 | –0.05 | 0.07 | –0.06 | 0.21** | 0.02 | 0.11* | 0.27** | 0.07 | 0.15** | 1 | |
14. Family involvement | –0.20** | 0.20** | 0.15** | 0.12* | –0.13 | 0.12* | –0.04 | 0.12* | –0.00 | –0.08 | 0.01 | 0.06 | –0.48** | 1 |
Table
Variables | Hazard ratio | 95% CI | p-value |
---|---|---|---|
Perceived workload | 0.533 | 0.383–0.742 | 0.000*** |
Work-family conflict | 1.111 | 0.878–1.405 | 0.382 |
Job content | 0.764 | 0.564–1.036 | 0.083* |
Perceived support for AWAs | 0.387 | 0.306–0.488 | 0.000*** |
Careerism | 1.160 | 0.918–1.465 | 0.215 |
Career involvement | 1.258 | 0.911–1.738 | 0.163 |
Family involvement | 1.594 | 1.146–2.217 | 0.006*** |
Age (in years) | 1.060 | 1.028–1.092 | 0.000*** |
Female (vs. male) | 1.171 | 0.793–1.729 | 0.428 |
In a relationship (vs. not) | 1.308 | 0.882–1.940 | 0.181 |
Children (vs. not) | 0.542 | 0.340–0.863 | 0.010** |
Big 4 (vs. non Big 4) | 1.509 | 1.034–2.204 | 0.033** |
Career satisfaction | 0.893 | 0.654–1.219 | 0.475 |
The estimate for job content is 0.76. This result means that, at any particular point in time, two-thirds as many respondents left their initial audit firm if they did not consider their job to be challenging and varied. This estimate is somewhat precise but is still compatible with the effect being zero (95% CI [0.56–1.0], p = .08). This provides some support for Hypothesis 2a. The estimate for perceived support for AWAs is 0.39, with only a very narrow range of plausible true effects being compatible with the data (95% CI [0.31–0.49], p < .01). This suggests that, at any point in time, there is a 61% reduction in the risk of turnover for employees who perceive their organization to support alternative work arrangements. This result supports Hypothesis 2b. Overall, these results provide fairly strong evidence that job resources reduce the risk of organizational turnover.
The point estimate for careerism is 1.16, but this estimate is not very precise (95% CI [0.92–1.47], p = 0.22). Thus, the current data do not support Hypothesis 3a, as both negative and positive values of careerism are compatible with the data. Likewise, the estimate for career involvement is 1.26, but this estimate is highly imprecise (95% CI [0.91–1.74], p = 0.16). The current data do thus not provide evidence for Hypothesis 3b, with a wide range of plausible true effects being compatible with the data. Finally, the data provide relatively strong evidence that family involvement is negatively associated with organizational turnover (hazard ratio = 1.59, p < .01), supporting Hypothesis 3c. A wide range of plausible true effects, however, is compatible with the data (95% CI [1.15–2.22]), suggesting that the increase in turnover as a result of increased family involvement may be anything from 15% to 122%. Overall, this provides some evidence that audit employees’ personal demands may influence turnover.
Supplementary analyses were conducted to examine whether my results are sensitive to why audit professionals left their firm. First, I excluded respondents who did not leave their firm voluntarily and ran a survival analysis on the sample of employees who left their first employers voluntarily (n = 288). Employees leaving their firm voluntary may do so for different reasons (e.g., better career opportunities) than employees who are dismissed by their employer (e.g., low-performing or unmotivated audit staff). The results of this analysis are similar to those reported in Table
Second, I focused on whether employees left the auditing profession altogether (occupational turnover) rather than just their firm (organizational turnover). There is much less research on occupational turnover than on organizational turnover, but their determinants and consequences may differ. For example, organizational turnover may be more affected by specific workplace characteristics (e.g., job demands and resources) while occupational turnover may be more affected by broader career aspects and goals (e.g., personal demands and resources). For this analysis, I performed a survival analysis in which the event of interest was leaving the auditing profession, and the survival time was the length of stay in the auditing profession. This analysis shows that the risk of leaving the auditing profession is positively associated with perceived workload and negatively associated with job content and perceived organizational support for AWAs. Family involvement is not associated with the probability of leaving the auditing profession. Higher levels of careerism are, however, positively associated with the probability of leaving the auditing profession (hazard ratio = 1.37, p < 0.05).
Third, I examined whether there were any interaction effects present (e.g., between family involvement and sex). The data did not provide support for the existence of any interaction effects.
The auditing profession provides a context in which turnover is both highly frequent and very costly. As is the case for other professional service firms, the primary asset in auditing firms is human capital (the knowledge, skills, and connections of its professionals). This makes (voluntary) turnover of skilled professionals highly costly to audit firms. Retention of professionals is, therefore, typically considered one of the most important issues in audit and other professional service firms (
Understanding the causes of employee turnover can help audit firms to develop effective retention plans and to retain talented personnel in the auditing profession. The results of the current study thus provide some valuable insights for audit practice. First, (perceived) workload has a large effect on both organizational and occupational turnover. This finding is in line with claims from practice (e.g.,
The result on perceived workload aligns with the theoretical predictions of the JD-R model that excess job demands lead to negative outcomes such as employees quitting their job (i.e., voluntary turnover). At the same time, however, results of the current study also suggest that not every conceivable job demand is necessarily negative. While work-family conflict is often claimed to be an important reason for employees leaving the auditing profession (e.g.,
Second, in line with the JD-R model, the results of the current study suggest that high levels of job resources can reduce employee turnover. Audit firms can thus mitigate undesirable employee turnover by providing their employees with the necessary resources to achieve work goals, reduce job demands, or stimulate personal growth and development. Specifically, the current study’s data suggest that perceived support for AWAs substantially reduces the risk of employees voluntarily walking away from their audit firm or the auditing profession entirely. These results stress the importance for audit firms to create appropriate work climates that their employees consider supportive. While such AWAs are in place in most audit firms nowadays, their use is often undermined by organizational expectations about availability and commitment (
Perhaps surprisingly, the current study does not provide strong support for the idea that job characteristics, such as job-related challenge and variety, play an important role in understanding employee turnover in the auditing profession. Such characteristics typically show a moderately negative relationship with employee turnover in other contexts (
Third, the results are relevant to employee selection and hiring. Employee turnover is costly to firms, especially if highly competent employees leave at a higher rate than preferred. Such concerns may be more important to audit firms than most other firms, including other professional service firms. After all, due to their offering of extensive training, continuing professional education, and exposure to a wide variety of clients, audit firms are very attractive to career-minded individuals who use initial employment at these firms as a learning experience and stepping stone to other employment opportunities (
Finally, the current study provides relatively strong evidence that family involvement is negatively associated with organizational turnover. In contrast, the data are consistent with family involvement not increasing the risk of leaving the auditing profession entirely. These results might be explained by the fact that most employees in the sample who left their firm without leaving the profession changed from working at a larger (Big 4) to a smaller (non-Big 4) audit firm. Employees with high levels of family involvement arguably fit better within such smaller audit firms, as such organizations are associated with less work-family conflict and burnout (
Some potential limitations to the current study need to be noted. First, even though I performed a non-response analysis based on early and late respondents, I cannot entirely rule out non-response bias. If there are systematic differences with respect to substantive constructs between those that did and did not respond to the survey, the results of the current study will not generalize to the entire population of auditors. If, for example, employees who strongly identify with their work (high career involvement) were less likely to participate in the survey and are less likely to leave their organization or the auditing profession entirely, the results for career involvement will be biased. While this cannot be ruled out, about half of the sample in the current study actually had left their initial employer, and about a third of the respondents had left the auditing profession entirely. This does suggest that the survey was able to reach a substantial number of leavers. Second, the results of the current study may also not be perfectly generalizable to the entire population of auditors because the survey was sent to all the persons in Belgium who had entered into at least one course of theoretical instruction (see footnote 4). Employees who left the auditing profession before entering these courses could thus not be identified. To the extent that employees who leave the auditing profession very early on in their careers (i.e., before entering into a single exam) differ with respect to substantive constructs from those who leave their initial audit firm at a later stage, results of the current study do not generalize to the entire population of auditors. However, this is unlikely to have much effect on the study results because, at the time of the study, the successful completion of all such exams (save for exemptions based on equivalent qualifications) was a prerequisite for entering the practical training of (at least) three years. Therefore, audit employees generally entered these entrance exams fairly quickly after starting employment at an audit firm.
Kris Hardies is professor in accounting at the University of Antwerp. Prof. Hardies studies careers in the accounting and other professions, with a special focus on gender inequalities and discrimination.
The author thanks Anna Gold (the Subject Editor) and two anonymous reviewers for valuable feedback, Diane Breesch and Kelly Steenackers for help with the initial survey development and data collection, and the Belgian Institute of Registered Auditors for their logistic support in executing this research.
Only about 10% of turnover studies use survival analysis (
The JD-R model does not provide well-defined sets of demands, resources, and outcomes, but offers an open model that allows flexibility in its application in different contexts. For this reason, the current study does not include strain and motivation as potential mediators between, respectively, demands and resources and turnover, although some versions of the model suggest that it is through these mediators that the two different processes (to wit, a health-impairment and a motivational process) affect organizational outcomes such as turnover.
EU Directive 2014/56/EC, regulating entrance to the auditing profession in the EU at that time, required a test of theoretical knowledge, related to 19 subjects, as prerequisite for becoming a certified auditor. In Belgium, this testing of theoretical knowledge was organized by way of organizing an exam for each subject matter separately (“entrance exams”). At the time of the study, the successful completion of all such exams (save for exemptions based on equivalent qualifications) was a prerequisite for entering the practical training of (at least) three years. Therefore, audit employees generally entered the entrance exams fairly quickly after starting employment as an auditor.
There were 33 items missing for the final sample of 309 respondents (i.e., 0.3% of all scale items). Dropping respondents with any missing values yields similar results to those reported.
Initially, measures of sample adequacy were carried out to see whether the data was suitable for factor analysis. The Kaiser-Meyer-Okin value was 0.83 and the Bartlett’s Test of Sphericity indicated a chi-square value of 4755.522 (p < 0.001), supporting the suitability of the data for factor analysis. I employed principal component and promax with Kaiser Normalization as the method of factor extraction and rotation, respectively. An oblique rotation (promax) was used because some of the variables were likely to be correlated (e.g., perceived workload and work-family conflict).
The proportional hazard assumptions for Cox regression were tested by means of Schoenfeld residuals and were found not to be violated. In addition, I tested the proportional hazards assumption by adding time-dependent covariates to the model. None of the interaction terms were significant and thus the assumption was met.