Estimating the antecedents of Portuguese human resources churning: A multilevel analysisAbstract
Purpose: The purpose of this study is to estimate the antecedents of human resource churning. We identified the causes of churning, analyzed their interdependence with human resource churning, and examined the influence that retention measures have on the occurrence of human resource churning.
Originality: Develop and validate a final model through a multilevel analysis to estimate the antecedents of human resource churning, which is believed to be the first time that such a model has been designed and analyzed in Portugal.
Design/methodology/approach: This is a quantitative study using questionnaires as the data collection method. Three analyses were conducted using various statistical techniques to validate the proposed hypotheses and fulfill the defined objective.
Findings: Based on the results, positive values for professional performance and working hours contribute to human resource retention, which in turn negatively influences churning. In this sense, it is assumed that organizations should address professional performance conditions and working hours through strategic measures to retain human resources and minimize human resource churn.
Contributions: This research aims to develop a human resources churn model to test and analyze its key dimensions, identify which variables have negative effects, and propose corrective measures and human resources retention policies.
Keywords:
churning; human resources; structural equations; multilevel; retention
Resumo
Objetivo: Estimar os antecedentes do churning de recursos humanos. Procedeu-se à identificação das causas de churning, à análise da interdependência que estas exercem sobre o churning de recursos humanos, bem como à análise da influência das medidas de retenção sobre a ocorrência de churning de recursos humanos. Originalidade/relevância: Elaborar e validar um modelo final através de uma análise multinível, de forma a estimar os antecedentes do churning de recursos humanos; segundo se pressupõe, trata-se da primeira vez que se procede à sua conceção e análise em Portugal. Principais aspectos metodológicos: Trata-se de um estudo de caráter quantitativo, com recurso a questionários como método de recolha de dados. Procedeu-se à realização de 3 análises, recorrendo a várias técnicas estatísticas, com o intuito de validar as hipóteses propostas e cumprir o objetivo definido. Síntese dos principais
resultados: Considerando os resultados obtidos, os valores positivos nas condições de desempenho profissional e no horário de trabalho contribuem para a retenção de recursos humanos, o que, por sua vez, exerce influência negativa sobre a ocorrência de churning. Neste sentido, pressupõe-se que as organizações atuem sobre as condições de desempenho profissional e os horários de trabalho, por meio de medidas estratégicas de retenção de recursos humanos, de forma a minimizar a ocorrência de churning de recursos humanos. Contribuições/implicações: Por meio desta investigação, pretende-se contribuir com um modelo de churning de recursos humanos, com o intuito de testar e analisar as suas principais dimensões e, com base nos seus resultados, constatar quais as variáveis que exercem efeitos negativos, de forma a propor medidas de correção, bem como políticas de retenção de recursos humanos.
Palavras-chave:
churning; recursos humanos; equações estruturais; multinível; retenção
INTRODUCTION
To highlight the need to operationalize the concept of voluntary human resource churning Torre et al. (2017) reported a negative relationship between voluntary churning and organizational performance, highlighting three important factors that require further research: 1. measuring human capital, as this is the main factor in an organization, noting that larger organizations are the most damaged in terms of human resource churning, 2. the lack of hidden information as new hires may (or may not) compensate for the damage caused by voluntary churning, more in-depth research is needed to understand how each factor influences churning and how it is related to productivity, and finally, they mentioned 3. the scarcity of studies comparing the effects of churning, taking into account contextual factors such as technological innovation and the organization’s activity.
The importance of operationalizing the concept of human resource churning lies in its ability to enable organizations to take preventive and corrective measures to reduce replacement costs, retain top talent, generate revenue, and stand out in their business activities (Hudson, 2015). In this sense, it is pertinent to estimate the antecedents of employee turnover. In this study, the causes of churning were identified, along with their interdependencies and the influence of retention measures on human resource churning. To demystify this subject, which remains underexplored and underdeveloped in Portugal, we sought to contribute to this body of literature by increasing both the number of empirical studies and the overall literature. To alert organizations to the need to implement this concept to anticipate potential employee departures and reduce costs through preventive measures.
This study employs a quantitative approach to analyze the antecedents of churning in the Portuguese human resources sector. This study aimed to highlight the importance of operationalizing the concept of churning in the context of human resources management to promote the retention of its best workers and to minimize costs associated with the loss of investment in departing workers and the costs of new hires to fill replacements resulting from voluntary departures.
In this sense, based on the study carried out by Pirrolas et al. (2023a), after analyzing the variables that make up the causes of human resource churning, the following factors emerged: Conditions of Professional Performance (CPP), Satisfaction with Working Conditions (SWC) and Schedule (S) In terms of the composition of each of these factors, the Conditions of Professional Performance (CPP) factor was composed of leadership, the Satisfaction with Working Conditions (SWC) factor was composed of working environment and salary, and the Schedule (S) factor was composed only of Schedule, which did not change its name (Pirrolas et al., 2023a).
Having selected and named these factors and considered the main causes of churning, it is pertinent to briefly discuss the influence of human resource retention in mitigating churning. In view of the above, the following points address the causes of human resource churning and strategic human resource retention measures, and propose hypotheses that are analyzed below.
THEORETICAL FRAMEWORK
Causes of human resource churning
The occurrence of human resource churning is a problem not only for individual organizations, but also for the economy and society as a whole, as it can negatively affect long-term growth. Human resource turnover can affect workplace morale and motivation and jeopardize productivity (Punnoose & Ajit, 2016). A high rate of churning increases human resource costs (Berengueres et al., 2017) and adversely affects productivity, requiring additional time and effort to replace skilled workers who have left (Ma et al., 2019).
Leavers come from a variety of perspectives, i.e., there can be a variety of reasons for employees to leave their organizations, whether these are related to better job offers, including bigger and better career opportunities and more attractive locations, or for negative reasons that can be related to damaging personal relationships with supervisors or colleagues and adverse and unsafe working environments (Saradhi & Palshikar, 2011).
A good relationship between supervisors and workers can promote a positive working environment, reflecting trust, respect, and loyalty, producing a stronger sense of team and greater job satisfaction, thereby reducing intentions to leave (Trevor, 2001). However, job satisfaction is considered a key predictor of human resource churning (Dickter et al., 1996).
Other factors may also contribute to human resource turnover, such as age, job stability, pay, overall job satisfaction, and fairness. At the same time, supervision, recognition, and potential career progression are considered key factors influencing human resource churning.
Given the multidimensional nature of human resource churning, it has been gaining relevance in organizations, prompting greater attention and the need for intervention. However, little work has been done on predicting human resource churning (Berengueres et al., 2017), and this is a point that requires organizational intervention, given the need to predict churning and thus prevent it.
Accurately predicting human behavior is a difficult task. However, by leveraging information technology and workers’ accumulated data, it is possible to analyze this data, map behavioral patterns within certain categories, and predict workers’ intentions to leave the organization (Ma et al., 2019).
To obtain the information, organizations can collect data through an informal interview (a friendly conversation) or by collecting feedback from employees after each workday (Berengueres et al., 2017). However, it is important to mention Walton’s (1973) approach in presenting the dimensions of quality of life at work as contextual variables that in practice are the responsibility of human resource management, such as concern for humanistic and environmental values, considering eight criteria for this purpose: fair and adequate compensation; healthy and safe working conditions; opportunity for skills development; opportunity for continuous growth; social integration at work; constitutionalism; work and living space and social relevance of life at work.
Conditions of Professional Performance (CPP)
As mentioned above, the professional performance conditions factor, comprising leadership, recognition, and work environment, is considered a cause of churning.
Leadership has many definitions, but there is no real consensus. Essentially, it is a relationship through which one person influences the behavior or actions of others. In everyday speech, leadership and management are wrongly used synonymously. While management is concerned with the development, planning, and control of organizational resources, leadership is concerned with aligning people with the results expected from an organizational perspective. To lead, you need the ability to manage; therefore, the two are interlinked (Gwavuya, 2011).
Leadership is defined as the ability to lead, inspire, guide, and treat workers impartially; this factor has a major impact on employee turnover. If the organization or team is led by poor leadership, the imbalance it creates in the relationship between the manager and the worker will lead to dissatisfaction and demotivation, promoting employee turnover. In this sense, a leader should promote well-being, motivation, and organizational effectiveness (Alkhawaja, 2017; Kamalaveni et al., 2019; Miranda et al., 2016).
McNeese-Smith (1995) state that leadership has a major impact on employee retention in an organization, with a leader’s attitude influencing employee commitment and leadership style determining employee retention by influencing the organization’s success or failure (Budhiraja & Malhotra, 2013; Kaye & Evans, 2002).
In this sense, the type of leadership is one of the decisive factors in whether to stay or leave an organization. Good leadership provides satisfaction with remaining in the organization, promotes empathy between leaders and workers, and fosters their permanence in the organization (Kaye & Evans, 2001; Sisodia et al., 2017).
Satisfaction with Working Conditions (SWC)
Considering satisfaction with working conditions as one of the causes of churning, as mentioned above, this was made up of working environment and salary. According to Rossberg et al. (2004), one of the factors influencing churning is the working environment, defined as the surroundings in which the worker performs his duties. If the work environment is poor, it affects work performance, leading workers to consider leaving the organization. In this sense, a good working environment reduces the occurrence of churning and promotes employee retention (Horwitz et al., 2003).
In an organization, employees expect a work environment where they can use their abilities and skills and satisfy their basic needs. The connectedness of the workspace has a profound impact on workers. It is considered a factor influencing workers’ decisions to stay or leave the organization, depending on their satisfaction with the organization (Brill et al., 2001).
Gouchait and Cho (2010) state that 80 percent of workers is seeking a better work environment. It is therefore vital to create a work environment conducive to organizational support to retain employees (Ghosh & Sahney, 2011). However, it should be noted that the work environment, in addition to the surroundings, includes various factors such as supervisory, organizational, and peer support (Broad & Newstrom, 1992).
Concerning salary as one of the causes of human resource churning, given that it is defined by the monetary reward for the work done, an organization that pays low salaries or uses an inappropriate salary scale will contribute to the voluntary departure of workers, increasing churning. In general, salary is one of the key factors in keeping workers in the organization (Milkovich et al., 2013; Pitts et al., 2011).
According to a study by Iqbal et al. (2017), the salary factor accounts for 38% of worker retention in organizations. Organizations that offer higher pay and benefits than other companies in the same sector are better able to attract and retain high-quality employees. To strengthen employee motivation, the relationship between pay practices and job satisfaction is very important (Iqbal et al., 2017).
Given the increasing competitiveness experienced in the labor market, the level of remuneration practiced in an organization directly influences the churning of human resources, since workers compare their remuneration with that practiced in other organizations in the same sector of activity, and this is a decisive factor in whether they remain in the organization (Nagaraju & Pooja, 2017).
Organizations with higher salaries and good benefits have lower turnover rates. In this sense, studies have shown that organizations with low salaries and poor benefits tend to increase the churning rate. Additionally, in these organizations, the workforce is smaller compared to organizations with higher salaries (Dale-Olsen, 2006; Hwang et al., 1998).
Schedule (S)
Another factor considered a cause of churning is working hours, i.e., their influence on churning, since they interfere with the balance between personal and professional life. An organization that promotes flexible working hours, enables better workload management, provides breaks during working hours, and offers the possibility of teleworking will help reduce employee turnover, thereby promoting employee retention (Kamalaveni et al., 2019).
The ability to choose companies that offer flexible working hours allows employees to better adapt to work, manage their workload, and balance responsibilities in their personal lives (Nor, 2018). The implementation of flexible working hours allows employees to reconcile work with their personal lives, providing a healthy balance and contributing to a decrease in the rate of people leaving the organization, resulting in greater employee retention (Nasir & Mahmood, 2016).
Given that work interferes with employees’ personal lives, it can lead to emotional exhaustion and higher rates of stress among employees (Hyman & Summers, 2004). The following retention strategies are suggested: flexible working hours, flexible work organization, adequate resources, adequate workload, and adequate breaks during working hours to promote an environment of organizational well-being (Deery, 2008).
The possibility of flexible working hours, the length of working hours, and the opportunity to take time off work are all factors that contribute to employee satisfaction and thus help to reduce the occurrence of churning (Kaye & Evans, 2001; Pirrolas & Correia, 2020).
Strategic measures for human resource retention
Retaining workers is currently a very critical issue. The cost of losing talented, experienced, and qualified workers is much higher than the cost of hiring new ones. It is therefore very important for organizations to retain their best workers (Iqbal et al., 2017). Strategic human resource retention measures are defined as an organization’s ability to retain its best employees. An organization that can retain its employees invests in implementing policies that provide better working conditions and promote employee wellbeing, so they feel satisfied staying with the organization.
Thus, considering working conditions as the driving factor behind churning, it can be predicted that implementing strategic retention measures will help reduce the churning rate (Saradhi & Palshikar, 2011). Organizations must act in order to understand the reason for leaving, through exit interviews, the results obtained through performance evaluation, perceiving behavioral changes, in other words, understanding the reason for dissatisfaction or the intention to leave the organization. With this insight, they can implement corrective or preventive measures to minimize human resource churn.
DEVELOPMENT OF THE PROPOSED HUMAN RESOURCES CHURNING MODEL AND ITS RATIONALE
Considering retention as a strategic human resource management practice, perceived by employees, and churning as the result of a set of management indicators directly related to the costs resulting from turnover (Burgess et al., 2000; 2001; Flórez et al., 2017; David et al., 2015; Perryer et al., 2010).
It is assumed that we are dealing with a Multilevel Model in which the first level represents employees’ perspectives on the causes of churning and retention (Hom et al., 2017; Irshad, 2011). Similarly, retention also corresponds to an employee’s perspective on human resource management practices, which translates into a greater or lesser intention to leave the organization (Hudson, 2015; Dhanpat et al., 2018). The second level corresponds to organizations in terms of size, the number of voluntary departures and replacements, and the wage bill.
The outline of the model of the causes of human resource churning can thus be seen (Figure 1).
If we link the concept of churning to the costs of turnover, then a high level of employee retention leads to a decrease in the intention to leave, reducing turnover and, in turn, to a decrease in churning (Yigit & Shourabizadeh, 2017; Steil et al., 2016; Burgess et al., 2000). Therefore, the following hypothesis is formulated:
-
H1: Employee Retention exerts a negative influence on human resource churning.
Conditions of Professional Performance (CPP), Satisfaction with Working Conditions (SWC) and Schedule (S), on the other hand, are positive causes of retention (Pitts et al., 2011; Katz & Kahn, 1966) and negative causes of churning (Rahimic et al., 2012; Hatane, 2015). The following hypotheses can thus be formulated:
-
H2a: Conditions of Professional Performance positively influence Retention;
-
H2b: Conditions of Professional Performance negatively influence Churning;
-
H3a: Satisfaction with Working Conditions positively influences Retention;
-
H3b: Satisfaction with Working Conditions negatively influences Churning;
-
H4a: Schedule positively influences Retention;
-
H4b: Schedule has a negative influence on Churning.
These Hypotheses support the Multilevel Structural Model shown in Figure 2.
This study consists of various data analyses. Firstly, the indicators used to measure human resource churning were validated. This was followed by validation of the retention measurement scale. In a third phase, the proposed model of the causes of human resource churning was estimated, and the hypotheses supporting it were tested (Figures 8 and 9). According to Anderson and Gerbing (1988), the measurement model, together with the structural model, enables a comprehensive, confirmatory assessment of construct validity, including convergent and discriminant validity.
Validation of the human resources churning measurement model (analysis 1)
This first analysis involved developing and administering a questionnaire to collect indicators of human resource churn. The creation of the variables selected for this work was triggered by a theoreticalconceptual study carried out by (Pirrolas & Correia, 2023b) which resulted in the selection of categorical variables referring to the organization: type of organization and sector of activity, as well as the main dimensions of human resource churning, specifically: 1. the size of the organization, 2. the number of voluntary departures, 3. the number of entries by replacement, and 4. the wage bill. These variables correspond to Level 2, i.e., the level corresponding to organizations.
The sample
The sample was a convenience sample of 12 organizations, each represented by its human resources director, across various sectors. The data collection process took place between October 2022 and January 2023.
Through descriptive analysis, it was possible to verify that the 12 human resources directors belonged to different types of organizations, with 7 to national companies, 3 to multinational companies, 1 to public administration bodies, and 1 to social solidarity bodies (IPSS, NGOs, others). The majority of participants, 7 people, belonged to the industrial sector, followed by the construction and public works sector with 2 people, the health and social action sector with 2 people, and the agriculture, fishing, commerce, transport, and communications sectors with 1 person each.
Regarding the size of the organization, 8 respondents belonged to organizations with more than 250 employees, 3 belonged to organizations with 10 to 50 employees, and one belonged to organizations with less than 10 employees.
Data collection instruments
In order to characterize the organizations and specify the components of human resource churning, i.e. the indicators related to turnover costs, we used a questionnaire made up of the following manifest variables (items): 1. size of the organization (DORG); and 2. number of voluntary departures (S); number of replacements (E); and wage bill (CSUB) (Table 1).
Items in the questionnaire used to assess the factors used to measure human resources churning
Methodological options for data processing and analysis
Initially, the variables under study were calculated as indices based on the average of the responses, resulting in a churning model composed of indices and continuous variables. However, the absolute value of the manifest variables - number of voluntary departures (S); number of entries by replacement (E); and wage bill (CSUB) is misleading as to the impact of this manifest variable on the latent variable churning; for example: consider the value of departures of five workers that occurred in two organizations. This figure has a greater impact on the smaller organization. Therefore, to reflect the impact (greater or lesser) on churning and retention, we chose not to use the absolute value but rather its percentage relative to the organization’s size.
So, if in the example five workers left, in an organization with 20 workers we are talking about a 25% departure, while in an organization with 100 workers the departure would be 5%. It was found that, despite being included in the questionnaire, the wage bill (CSUB) would be irrelevant for calculating turnover costs, i.e., the churning of human resources, not only in relation to the size of the organization but also as a variable related to the structure of human resources. In addition to the dimensions mentioned above, the variables Turnover (Rot) and Replacement Costs (RC) were also considered.
Turnover (Rot) was defined as the difference between workers who left (S) and those who were replaced (E). As for the Replacement Costs variable (CSUB), it was determined by adding up the costs of a) recruitment and selection, b) integrating the new worker, and c) training them. Thus, human resource churning was measured by these five manifest variables, which correspond to human resource management indices (Figure 3).
Structural Equation Modeling (SEM) was used to validate the proposed model, as it allows estimation of structural relationships among the variables, with retention as the latent variable. Various goodness-of-fit indicators were also used to identify the model that best fit the data.
Validation of the human resources churning measurement model
This validation was based on indicators for 12 organizations, provided by their human resources managers. The majority of these organizations are eight national companies, and the dominant sector of activity is industry (Table 2).
There is also variability in replacement costs resulting from the sum of recruitment costs (from 0-5000 €), integration costs (from 0-2000 €), and training costs for the new worker (from 0-1500 €) (Table 3).
The 5 manifest variables of human resources churning will take different values across organizations (Table 4).
The human resources churning was then validated using Confirmatory Factor Analysis (CFA). The first extracted measurement model indicated an absolute goodness-of-fit below the acceptable threshold. However, the relative goodness of fit already showed index values considered acceptable (Table 5).
Goodness of fit indicators of the initial measurement model of the causes of human resource churning
Three of the four manifest variables had very high factor loading values, suggesting a strong relationship with churning (Figure 4).
However, the variable Organizational Dimension (DORG) has a factor loading of 0.28, which is below the established reference of 0.50 (Hair et al., 2014; Salgueiro, 2012), so it was excluded, and the measurement model was re-specified. The resulting measurement model has an acceptable goodnessoffit (Table 6).
Goodness of fit indicators of the final measurement model of the causes of human resource churning
The four manifest variables showed high factor loadings, with the lowest, 0.54, for replacement costs (Figure 5).
It should also be noted that the average variance extracted from the churning was 81% (0.81) and the construct reliability was 0.94, well above the reference values of 0.50 and 0.70, respectively (Hair et al., 2014; Salgueiro, 2012). The results obtained support these 4 indicators (% of Leavers; % of Entrants; % of Turnover and Replacement Costs), which provide a good measure of human resource churning.
VALIDATION OF THE RETENTION MEASUREMENT MODEL (ANALYSIS 2)
Sample
The sample consisted of workers from the companies where the aforementioned human resources directors worked, specifically 204 workers from different organizations and different sectors of activity. Their characterization is presented later through descriptive analysis. This questionnaire was also administered to employees of the participating companies between October 2022 and January 2023, as mentioned above.
Data collection instrument
A 10-point Likert scale was used for data collection. This scale consisted of 6 items. The scale was drawn up using the scale previously applied (study 3), to which items related to human resource retention were added and adapted (Table 7). These items were drawn from an existing, validated questionnaire on human resource retention (Kyndt et al., 2009). It should be noted that the decision to add these items was intended to select retention measurement indicators.
Data processing options
The validation of the retention measurement model (specifically, the human resources retention strategy) follows the pattern already described and used previously; it will be carried out using Confirmatory Factor Analysis (CFA). Based on the first extracted model, items with factor loadings of less than 0.50 and/or high modification indices were eliminated (Hair et al., 2014). Convergent validity was assessed using the average variance extracted and construct reliability (Hair et al., 2014; Salgueiro, 2012).
Data treatment
The first Confirmatory Factor Analysis (CFA) yielded a Retention (RET) measurement model with an unacceptable goodness-of-fit (Table 8).
All items had a loading factor value greater than 0.50 (Figure 6) as established by Hair et al. (2014) and Salgueiro (2012).
We then proceeded to successively re-specify the initial Retention measurement model by eliminating items with very high modification indices relative to other items. In this way, the following items were eliminated:
-
RET 2: “Within this company, my job gives me satisfaction.” RET 3: “I see a future for myself within this company.”
-
RET 5: “If I wanted to do another job or function, I would first look at the possibilities within this company.”
This resulted in a final Retention measurement model made up of three items. All of them had factor loading values greater than 0.78 (Figure 7).
As it only consists of three items and the degrees of freedom are 0, it is not possible to calculate the goodness of fit indicators. However, comparing the AIC of the initial model (AIC = 661.048) with that of the final model (AIC = 361.651) indicates a decrease, showing the better structural adequacy of the final model. It was also found that the average variance extracted was 67 percent (0.67) and the construct reliability was 0.87, which supports the convergent validity of Retention.
ESTIMATION OF THE PROPOSED MODEL OF THE CAUSES OF HUMAN RESOURCE CHURNING (ANALYSIS 3)
The purpose of this study is to estimate the causes of human resource churning, using the model initially developed in Figure 1.
Sample
The sample consisted of workers from the participating companies mentioned in the previous section, who were given a socio-demographic questionnaire to collect data on the sample (age, gender, marital status, educational qualifications, dependents), and to characterize their work (type of organization, sector of activity, size of organization, position, contractual relationship, type of work, working hours, working regime and salary).
Therefore, in this initial phase, a descriptive analysis was conducted of the 204 study participants in terms of their demographic and professional characteristics, as shown in Table 9. Confirmatory Factor Analysis (CFA) was then carried out using Lisrel statistical software.
Instrument
This study involved administering a questionnaire. This questionnaire was drawn up based on the questionnaire applied in the previous study, to which items relating to human resource retention were added and adapted (Table 10). These items were drawn from an existing, validated questionnaire of human resource retention (Kyndt et al., 2009). It should be noted that the decision to add these items was intended to select retention measurement indicators. This was then combined with the final model drawn up in a previous study (Pirrolas et al., 2023a) to create the Churning model (Level 1).
This questionnaire was also administered between October 2022 and January 2023 to the employees of the member companies presented earlier in this research. The Level 1 model consisted of 4 scales: Conditions of Professional Performance (CPP), Satisfaction with Working Conditions (SWC), Schedule (S), and Human Resource Retention (RET), the items of which are shown in Table 10.
Data processing options
This analysis used path analysis with manifest variables and structural equation modeling in a fully standardized solution. Path analysis means that “independent variables in one relationship can be considered dependent in another” (Salgueiro, 2012, p.8), as is the case with the proposed model. There is also the fact that the variables are all manifest.
Structural Equation Modeling (SEM) was used to validate the proposed model, as it enables estimation of the structural relationships among the variables. Various goodness-of-fit indicators were also used to identify the model with the best fit, as described above.
Considering that the first level used the common method (Self-Response Questionnaire) to measure the four variables that make up the model, specifically, Conditions of Professional Performance (CPP), Satisfaction with Working Conditions (SWC), Schedule (S) and Human resource retention (RET), it was decided to determine the variance bias resulting from the use of the common method (Podsakoff et al., 2003). To this end, the common variance method was used, as established by Podsakoff et al. (2003).
The Confirmatory Factor Analysis of the latent variables measured by the self-response questionnaire (which corresponds to Level 1) showed an acceptable goodness of fit (2 = 119.24, df = 48, CFI = 0.936, RMSEA = 0.086), which improved when the common factor was introduced (2 = 75.99, df = 36, CFI = 0.964, RMSEA = 0.074). It can be seen that the common factor (measured by all the items) has an average variance extracted of 32%, which, despite being higher than the reference value of 25%, is lower than the value established for measuring attitudes, as is the case with these 4 variables (conditions of professional performance, satisfaction with working conditions, schedule, and retention).
Thus, based on the above, the variance error of the common method is within acceptable limits, so it can be assumed that it will not significantly influence the estimation of the proposed model (Podsakoff et al., 2003).
Estimating the proposed model and testing the hypotheses
All four latent variables were transformed into manifest variables by averaging the responses (10-point Likert scale) obtained by the items (manifest variables) that measure them. This resulted in a set of descriptive statistics for the variables in the model (Table 11).
It should be noted that churning is not comparable with the other variables because it is determined by management indicators, as developed above. The variable with the highest value is Schedule (S), followed by Conditions of professional performance (CPP), while Satisfaction with Working Conditions (SWC) and Retention (RET) are at the mid-point of the scale. On the other hand, churning is significantly correlated only with Satisfaction with Working Conditions (SWC). However, this correlation is not supported by the theoretical framework, as it is positive.
Churning is also negatively correlated with the correlation, although this is not statistically significant. Finally, there is no correlation between churning (CHURN) and Conditions of professional performance (CPP) and Schedule (S). Having done this preliminary analysis, we proceeded to estimate the proposed human resources churning model.
The initial model has a 2 of 0.000, with 3 degrees of freedom (df), a P-Value of 1.000, and an RMSEA of 0.000. The indication in the Lisrel output that this is a perfect fit (“The fit is perfect!” extracted from the Lisrel output) can therefore be considered to be supported. However, the causal influence ratios for churning are very low, specifically close to zero (Figure 8).
Given the values of the structural relationships of influence between variables, the following hypotheses were rejected (Table 12).
The proposed model was therefore re-specified by eliminating these structural relationships. The resulting model, like the initial model, also has a very good fit (Table 13).
The values of the structural relationships of influence between the variables are statistically significant. On the other hand, this model explains only 5.9 percent of the variance in churning (R2 = 0.059) and 16.2 percent of the variance in retention (R2 = 0.162), which is clearly insufficient as an explanatory model of churning (Figure 9).
Analyzing the values obtained leads to the non-rejection of three hypotheses and the rejection of one. This rejected hypothesis is not about the significance of the influence relationship, but rather about its nature, since it assumes a negative influence and a positive influence was estimated. Specifically, it is hypothesis H3b (Table 14).
CONCLUSIONS, LIMITATIONS AND CONTRIBUTIONS
This research brings together three studies (three analyses). As this is a multi-level model, level 1 corresponds to employees and level 2 corresponds to organizations. This study complied with the requirement that the employees (level 1) belong to the same organizations as the human resources managers (level 2) who participated in this study.
The first study aimed to collect indicators for measuring human resource churning and to analyze its causes. The second study validated the retention measurement model, and the third aimed to estimate the proposed model of the causes of churning.
In study 1, an existing, validated questionnaire on the causes of churning was used. For study 2, questions on human resource retention were added to this questionnaire. These questions were taken from an existing questionnaire (Kyndt et al., 2009) on retention, which was translated and adapted. Their purpose was to validate the Retention measurement model. Finally, to complete this study, the third study involved estimating the proposed model of churning causes.
In the first study, after collecting the data, two estimated models were developed, and based on the factor loadings, we opted for the second model, which had higher factor loadings.
In the second study, to validate the Retention model and confirm its inclusion in the final proposed model, Confirmatory Factor Analysis (CFA) was conducted, which led to the elimination of items with very high modification indices. In this sense, of the 6 items that made up the model, only 3 were considered, all with factor loadings greater than 0.78.
In study 3, when the proposed model was estimated, the initial model was found to be a perfect fit, but the estimated causal influence relationships were very small, close to zero.
Regarding the structural relationships of influence among the variables, 4 of the 7 proposed hypotheses were rejected because they had negative standardized estimates and P-values. However, the rejected hypotheses do not corroborate the theoretical framework.
As a result of the values obtained, the model was re-specified, and these structural relationships were eliminated. The resulting model after these re- -specifications, like the initial model, has a very high goodness of fit, and the values of the structural influence relationships are statistically significant. However, the new model explains only 5.9 percent of the variance in churning, which is insufficient as an explanatory model.
Having analyzed this model, three of the seven proposed hypotheses were found to be true, namely:
-
H1: Employee Retention (RET) negatively influences Churning (CHURN), with a standardized estimate of -0.18 and a T-value of -2.49.
-
H2a: Conditions of professional performance (CPP) have a positive influence on Retention (RET), with a standardized estimate of 0.22 and a T-value of 2.36.
-
H4a: Schedule (S) positively influences retention (RET), with a standardized estimate of 0.22 and a T-value of 2.42.
However, although H3b shows positive values, the hypothesis was rejected given its nature: a negative influence is assumed, whereas a positive influence was estimated. Specifically: H3b: Satisfaction with working conditions (SWC) negatively influences Churning (CHURN), showed a standardized estimate of 0.21 and a T-value of 3.02.
It is assumed that organizations should act on professional performance conditions and working hours as strategic measures for retaining human resources. Considering the results, the positive values for professional performance conditions and working hours contribute to the retention of human resources, which in turn negatively influences the occurrence of churning.
Theoretical and practical contributions of the study
The questionnaire, which was applied and validated (Pirrolas et al., 2023a), added items related to the retention construct, drawn from an existing and validated questionnaire (Kyndt et al., 2009). This allowed the theoretical framework to be expanded by analyzing the final model presented above, which we assume is the first time it has been designed and analyzed in Portugal.
When the proposed model was validated, the results did not replicate the proposed factor structure, and the initial factors were removed due to low factor loadings and negative T-values. After their respecification, it was possible to proceed with their validation, which, through the factor loadings and T-values, allowed the model to be considered for analysis, with the aim of it being operationalized by organizations, given the insight it offers into which factors should intervene in order to minimize the occurrence of future exits.
Given the scarcity of literature and empirical studies on human resource churning, this could be considered the first empirical study to use a multilevel analysis to address this issue in Portugal. This research aims to contribute to the increase in literature and empirical studies, with the main focus on demystifying the subject of churning, as well as making it possible to implement it in organizations, given that it can be a preventative model for possible voluntary departures of human resources and minimize replacement costs (churning).
Limitations of the study
One limitation concerns the size of the sample. It is suggested that future studies increase the sample size and review the variables that make up each indicator to obtain higher factor loadings that reflect the weight they exert on the churning variable.
Another limitation was the difficulty in finding organizations that would allow two questionnaires to be administered in the same organization (as mentioned above, one questionnaire was aimed at human resources directors and the other at employees of the same organization). When the questionnaires were collected, many were excluded because they did not meet the criteria mentioned above, due to human resources directors not being allowed by the organization to apply the questionnaires to the organization’s workers, or vice versa, the workers filled in the questionnaires but the organization did not allow the human resources directors to take part.
Open Data:
The contents underlying the research text are included in the manuscript.
REFERENCES
- Ajit, P. (2016). Prediction of employee turnover in organizations using machine learning algorithms. Algorithms, 4(5).
-
Alkhawaja, A. (2017). Leadership style and employee turnover a mythical relationship or reality? MA Leadership studies: Capstone project papers, 16. https://digital.sandiego.edu/solesmalscap/16/
» https://digital.sandiego.edu/solesmalscap/16/ -
Anderson, J. C., & Gerbing, D. W. (1988). Structural equation modeling in practice: A review and recommended two-step approach. Psychological Bulletin, 103(3), -423. https://doi.org/10.1037/0033-2909.103.3.411
» https://doi.org/10.1037/0033-2909.103.3.411 -
Berengueres, J., Duran, G., & Castro, D. (2017). Happiness, an inside job? Turnover prediction using employee likeability, engagement and relative happiness ASONAM ‘17: Proceedings of the 2017 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining 2017, p. 509-516. https://doi.org/10.1145/3110025.3110132
» https://doi.org/10.1145/3110025.3110132 - Brill, M., Weidemann, S., Olsen, J., Keable, E., & Bosti. (2001). Disproving widespread myths about workplace design Research report.
- Broad, M., & Newstrom, J. (1992). Transfer of training: Action-packed strategies to ensure high payoff from training investments Addison-Wesley.
- Budhiraja, S., & Malhotra, M. (2013). Leadership style and organizational effectiveness in Indian IT and banking industry. Indian Journal of Industrial Relations, 49(2), 270-285.
-
Burgess, S., Lane, J., & Stevens, D. (2001). Churning Dynamics: An Analysis of Hires and Separations at the Employer Level. Labour Economics, 8, 1-14. https://doi.org/10.1016/S0927-5371(00)00027-0
» https://doi.org/10.1016/S0927-5371(00)00027-0 -
Burgess, S., Lane, J., & Stevens, D. (2000). Job Flows, Worker Flows, and Churning. Journal of Labor Economics, 18, 473-502. https://doi.org/10.1086/209967
» https://doi.org/10.1086/209967 - Dale-Olsen, H. (2006). Wages, fringe benefits and worker turnover. Labour Economics, 13, 87-105.
- David, S., Kaushik, S., Verma, H., & Sharma, S. (2015). Attrition in “IT” Sector. International Journal of Core Engineering and Management (IJCEM), 2(1), 74-92.
- Deery, M. (2008). Talent management, Work-life Balance and Retention Strategies. International Journal of Contemporary Hospitality Management, 20(7), 792-806.
-
Dhanpat, N., Modau, D., Lugisani, P., Mabojane, R., & Phiri, M. (2018). Exploring employee retention and intention to leave within a call centre. SA Journal of Human Resource Management, 16, 1-13. https://doi.org/10.4102/sajhrm.v16i0.905
» https://doi.org/10.4102/sajhrm.v16i0.905 - Dickter, N., Roznowski, M., & Harrison, D. (1996). Temporal tempering: An event history analysis of the process of voluntary turnover. Journal of Applied Psychology, 81(6).
- Flórez, L., Morales, L., Medina, D., & Lobo, J. (2017). Labour flows across firm’s size, economic sectors and wages in Colombia: Evidence from employer-employee linked panel. Borradores de Economía, 1013, Banco de la Republica de Colombia.
- Ghosh, K., & Sahney, S. (2011). Impact of organizational sociotechnical system on managerial retention: a general linear modeling approach. Journal of Modelling in Management, 6(1), 33-59.
- Guchait, P., & Cho, S. (2010). The impact of human resource management practices on intention to leave of employees in the service industry in India: the mediating role of organizational commitment. The International Journal of Human Resource Management, 21(8), 1228-1247.
- Gwavuya, F. (2011). Leadership Influences on Turnover Intentions of Academic Staff in Tertiary Institutions in Zimbabwe. Journal of Academic Leadership, 9(1).
- Hair, J., Black, W., Babin, B., & Anderson, R. (2014). Multivariate data analysis (7th ed.). Pearson New International Edition.
-
Hatane, S. (2015). Employee satisfaction and performance as intervening variables of financial performance. Procedia - Social and Behavioral Sciences, 211, 619-628. https://doi.org/10.1016/j.sbspro.2015.11.081
» https://doi.org/10.1016/j.sbspro.2015.11.081 - Hom, P., Lee, T., Shaw, J., & Hausknecht, J. (2017). One hundred years of employee turnover theory and research. Journal of Applied Psychology, 102(3), 530.
-
Horwitz, M., Heng, T., & Quazi, A. (2003). Finders, keepers? Attracting, motivating and retaining knowledge workers. Human Resource Management Journal, 13(4), 23-44. https://doi.org/10.1111/j.1748-8583.2003.tb00103.x
» https://doi.org/10.1111/j.1748-8583.2003.tb00103.x -
Hudson, C. (2015). Churning in the human services: Nefarious practice or policy of “creative destruction”? New England Journal of Public Policy, 27(1), 1-11. https://scholarworks.umb.edu/cgi/viewcontent.cgi?article=1670&context=nejpp
» https://scholarworks.umb.edu/cgi/viewcontent.cgi?article=1670&context=nejpp - Hwang, H., Mortensen, D.T., & Reed, W.R. (1998). Hedonic wages and labor market search. Journal of Labor Economics, 16, 815-847.
- Hyman, J., & Summers, J. (2004). Lacking balance? Work-life employment practices in the modern economy. Personnel Review, 33, 418-429.
-
Iqbal, S., Guohao, L., & Akhtar, S. (2017) Effects of Job Organizational Culture, Benefits, Salary on Job Satisfaction Ultimately Affecting Employee Retention. Review Pub Administration Manag, 5(3), 1-8. https://doi.org/10.4172/2315-7844.1000229
» https://doi.org/10.4172/2315-7844.1000229 -
Irshad, M. (2011). Factors affecting employee retention: Evidence from literature review. Abasyn Journal of Social Sciences, 4(2), 307-339. https://ajss.abasyn.edu.pk/article?paperID=45
» https://ajss.abasyn.edu.pk/article?paperID=45 - Kamalaveni, M., Ramesh, S., & Vetrivel, T. (2019) A review of literature on employee retention. International Journal of Innovative Research in Management Studies (IJIRMS), 4(4), 1-10.
- Katz, D., & Kahn, R. (1966). The social psychology of organizations Wiley.
-
Kaye, B., & Jordan-Evans, S., (2001). Retaining key employees. Public Management, 1, 6-10. https://icma.org/sites/default/files/6717_500363.pdf
» https://icma.org/sites/default/files/6717_500363.pdf - Kaye, B., & Jordan-Evans, S., (2002). Retention in tough times: Here’s what 25 global talent leaders say about keeping good people-especially now. Talent Development, 56, 32-37.
- Kyndt, E., Dochy, F., Michielsen, M., & Moeyaert, B. (2009). Employee retention: Organizational and personal perspectives. Vocations and Learning, 2(3), 195-215.
- Ma, X., Zhai, S., Fu, Y., Lee, L., & Shen, J. (2019). Predicting the Occurrence and Causes of Employee Turnover with Machine Learning. Computer Engineering and Applications, 8(3), 217-227.
-
McNeese-Smith, D. (1995). Job satisfaction, productivity, and organizational commitment. Journal of Nurse Association, 25, 17-26. http://dx.doi.org/10.1097/00005110-199509000-00006
» http://dx.doi.org/10.1097/00005110-199509000-00006 - Milkovich, G., Newman, J., & Gerhart, B. (2013). Compensation (11th ed.). McGraw-Hill Education.
-
Miranda, R., Avelino, B., & Takamatsu, R. (2016). Relação entre satisfação no trabalho e desempenho financeiro das empresas. Contabilidade, Gestão e Governança, 19(3), 336-355. https://doi.org/10.21714/1984-3925_2016v19n3a1
» https://doi.org/10.21714/1984-3925_2016v19n3a1 - Nagaraju, Dr. B., & Pooja, J. (2017). Impact of salary on employee performance empirical evidence from public and private sector banks of Karnataka. International Journal of Marketing and Human Resource Management, 8(4), 43-51.
- Nasir, S., & Mahmood, N. (2016). HRM practices for employee retention: An analysis of Pakistani companies. European Journal of Business and Management, 8(30), 96-104.
- Nor, A. (2018). How to keep top talented employees on board: A review of literature. Global Journal of Human Resource Management, 6(2), 52-59.
- Perryer, C., Jordan, C., Firns, I., & Travaglione, A. (2010). Predicting turnover intentions: The interactive effects of organizational commitment and perceived organizational support. Management Research Review, 33(9), 911-923.
-
Pirrolas, O., & Correia, P. (2020). Profissão, família e educação - conciliação da tripla jornada: uma questão de políticas e práticas organizacionais ou uma questão de sexo? Revista da FAE, 23(1), 7-22. https://revistafae.fae.edu/revistafae/article/view/642
» https://revistafae.fae.edu/revistafae/article/view/642 -
Pirrolas, O., Correia, P., & Nascimento, J. (2023a). The scale of causes of churning: Elaboration and validation for Portuguese human resources. Social Sciences, 12(1), 1-18. https://doi.org/10.3390/socsci12010001
» https://doi.org/10.3390/socsci12010001 -
Pirrolas, O., & Correia, P. (2023b). Churning de recursos humanos: Proposta de um modelo teórico-metodológico. Revista de Administração Mackenzie, 24(5), 1-29. https://doi.org/10.1590/1678-6971/eRAMG230256.pt
» https://doi.org/10.1590/1678-6971/eRAMG230256.pt - Pitts, D., Marvel, J., & Fernandez, S. (2011). So hard to say goodbye? Turnover intention among US federal employees. Public Administration Review, 71, 751-760.
- Podsakoff, P. M., MacKenzie, S. B., Lee, J. Y., & Podsakoff, N. P. (2003). Common method biases in behavioral research: A critical review of the literature and recommended remedies. Journal of Applied Psychology, 88(5), 879-903.
-
Punnoose, R., & Ajit, P. (2016). Prediction of Employee Turnover in Organizations using Machine Learning Algorithms. International Journal of Advanced Research in Artificial Intelligence, 5(9). https://doi.org/10.14569/IJARAI.2016.050904
» https://doi.org/10.14569/IJARAI.2016.050904 - Rahimic, Z., Resic, E., & Kozo, A. (2012). Determining the level of management competences in the process of employee motivation. Procedia - Social and Behavioral Sciences, 41, 535-543.
-
Rossberg, I., Eiring, O., & Friss, S. (2004). Work environment and job satisfaction: A psychometric evaluation of the Working Environment Scale-10. Social Psychiatry and Psychiatric Epidemiology, 39(7), 576-580. https://doi.org/10.1007/s00127-004-0791-z
» https://doi.org/10.1007/s00127-004-0791-z - Salgueiro, M. D. F. (2012). Modelos com equações estruturais Edições SP; Sociedade Portuguesa de Estatística.
-
Saradhi, V., & Palshikar, G. (2011). Employee churn prediction. Expert Systems with Applications, 38(3), 1999-2006. https://doi.org/10.1016/j.eswa.2010.07.134
» https://doi.org/10.1016/j.eswa.2010.07.134 -
Sisodia, D., Vishwakarma, S., & Pujahari, A. (2017). Evaluation of Machine Learning Models for Employee Churn Prediction 2017 International Conference on Inventive Computing and Informatics (ICICI), p. 1016-1020. https://doi.org/10.1109/ICICI.2017.8365293
» https://doi.org/10.1109/ICICI.2017.8365293 -
Steil, A., Penha, M., & Bonilla, M. (2016). Antecedentes da retenção de pessoas em organizações: Uma revisão de literatura. Revista Psicologia: Organizações e Trabalho, 16(1), 88-102. https://doi.org/10.17652/rpot/2016.1.542
» https://doi.org/10.17652/rpot/2016.1.542 -
Torre, E., Zatzick, C., David Sikora, D., & Solari, L. (2017). Workforce churning, human capital disruption, and organizational performance in different technological contexts. Human Resources Management Journal, 28(1), 112-127. https://doi.org/10.1111/1748-8583.12167
» https://doi.org/10.1111/1748-8583.12167 - Trevor, C. (2001). Interactions among actual ease-of-movement determinants and job satisfaction in the prediction of voluntary turnover. Academy of management journal, 44(4), 621-638.
-
Yigit, I., & Shourabizadeh, H. (2017). An approach for predicting employee churn by using data mining 2017 International Artificial Intelligence and Data Processing Symposium (IDAP),1-4. https://doi.org/10.1109/IDAP.2017.8090324
» https://doi.org/10.1109/IDAP.2017.8090324
Notes. Conditions of Professional Performance (CPP); Satisfaction with Working Conditions (SWC); Schedule (S); Retention (RET); Churning (CHURN).
Note. DORG: Size of the organization; S %: Percentage of voluntary departures; E %: Percentage of entries by replacement; Rot %: Percentage of voluntary turnover; CSUB: Replacement costs; CHURN: Churning.
Note. Chi - Square = 10.71, df = 5, P-value = 0.05741, RMSEA = 0.322
Note. Chi - Square = 1.38, df = 2, P-value = 0.50253, RMSEA = 0.000
Note. Chi - Square = 88.72, df = 9, P-value = 0.00000, RMSEA = 0.209
Note. Chi - Square = 0.00, df = 0, P-value = 1.00000, RMSEA = 0.000
Notes. CHURN: Churning; CPP: Conditions of professional performance; SWC: Satisfaction with working conditions; S: Schedule; RET: Human resource retention.
Notes. CHURN: Churning; CPP: Conditions of professional performance; SWC: Satisfaction with working conditions; S: Schedule; RET: Human resource retention.