Business Statistics and Data Analytics Solution

Title

ANALYSIS OF IBM EMPLOYEE HR ATTRITION

Abstract

Human Capital is regarded as a significant part of organization and employee voluntary turnover is identified as a main issue. This study investigates the prediction of individual-level voluntary employee turnover. These predictions help firms to act for retention or succession planning of employees. The relation between the most significant variables that impact attrition are identified and the strength of the association is analysed. These variables considered in this study are: Age, Gender, Education, Tenure, Earning, Job performance, Job Satisfaction and Job involvement. The study is done on the publicly available data by IBM.

Introduction

Armed with the forces of a globalised workplace and internet technology, there has never been an easier time for the employees to find new jobs. With an increased transparency of corporate workplaces over the internet, employees can quickly search for jobs that match their skills. Potential employees are now able to find open job positions in any corner of the world and have true information of the real workplace culture before they even join. A recent survey calculated that voluntary turnover rates are at an all-time high with around half of American employees intending to leave their current job (Economist, 2018). It is not uncommon for employees to be even passively job hunting over the internet, despite being happily employed in their current workplace. This growing level of voluntary employee turnover is a source of worry for organisations around the world (Frith, 2017). Employee turnover can be a source of significant loss to any company. The loss from employee turnover accrues to the firm in form of direct costs of replacing the employee who left such as sourcing and recruitment cost of new hires as well as the training cost for these new employees. Deloitte pegs this direct cost of sourcing and recruitment to be an average $3976 per hire while the cost of providing training and orientation to the new hire can be upwards of $3000 per hire (Erickson, 2016). In addition to the hard dollar costs, organisations also have to face indirect costs of voluntary turnover such as loss of social capital asset, increased pressure on remainder staff, decline in service quality and employee morale as well as loss of organisational productivity (Dess & Shaw, 2001). Thus it is important for the company to be able to predict employee attrition, especially the voluntary component of it. By doing so, the company may be able to design and administer policies for workforce retention. Such predictive analytics is especially advantageous if the company managers are able to predict if a particularly valuable employee is likely to voluntary leave the organisation and when.

Research Question

In this study the research question primarily focuses on how to predict employee attrition rates based on a combination of a variables so as to be able to choose employee attributes which are closely associated with voluntary turnover. Following research questions will be addressed in this analysis:

• Is age is related to voluntary turnover?
• Is there an association between employee’s education and their likelihood of voluntary turnover?
• Is attrition correlated with Gender of the Employee?
• Does Employment tenure impact attrition?
• Is Attrition affected by Employee’s Pay?
• Is Job satisfaction significantly associated with attrition?

Literature Review

There have been considerable studies on factors responsible for Attrition. A meta-analysis done by John Cotton and Jeffry Tuttle (1986) found 26 variables which were responsible for the attrition. These include work relate factors such as employee pay job satisfaction, ob performance, relationship with co-workers, promotions; personal factors such as age tenure, gender, education, marital status, intelligence and external factors such as unions. Another important factor that plays an important role as a turnover predictor is the job involvement. Job Involvement is the factor that indicates how much an employee is involved in a particular job and carry out the job from start to end with various responsible role. This gives an employee feeling of ownership and motivate him to do that job effectively (Ongori, 2007). Pritchard (2007) and Messmer(2000) have found that on job training and development conducted by firms is an important part of the retention programmes in organisations. In a study relating age with intention for turnover, workers aged 25 to 34 were shown to have the greatest intent to voluntary exit, reporting the shortest average job tenure of three years (NG & FELDMAN, 2010). Developing a predictive analytics modelling framework of employee retention, the strongest predictors for voluntary turnover were tenure, overall job satisfaction, job performance, individual demographic characteristics (age/experience, gender, ethnicity, education, marital status), geographical factors, salary, working conditions, job satisfaction, recognition, growth potential (Ribes et al., 2017). Other factors include employee mismatch with job requirement, unbalanced interpersonal relationship with supervisor and peers, work stress, unsatisfactory work-life balance which results into employee unhappiness and influences an employee to resign from an organization (Jessica Sze-Yin Ho, 2010). Another research done on the Indian IT firms reveals the factors responsible for the attrition are Below expectation salary, Low incentives, Relationship with superior, Relationship with subordinates, Job knowledge, Skills utilization, Skills recognition, Acknowledgement of work by superior, Lack of appreciation, Unsatisfied work culture (Banerjee et al., 2017). The managerial rewards also play an important role in retaining the employee. Thus, manager’s interpersonal skills play a very important role in holding on the evaluable employee (Hoffman & Tadelis, 2018).

OBJECTIVE

This research essay aims at finding the attributes responsible for employee attrition so as to be able to determine the retention factors and display the most effective retention factors for each employee to improve the employee retention, thereby saving the HRM cost.

DATA COLLECTION

The dataset used in the analysis comes from a publicly available dataset which was released by IBM in 2015. This dataset represents the Human Resources Employee Attrition data of IBM Employees (IBM, 2015). IBM had created this data set in order to undertake advance HR Analytics and Predictive Modelling of Employee Attrition and performance behaviour using techniques of Machine Learning with Watson Analytics. This IBM employee attrition and performance dataset comprises of real employee data of IBM. The dataset consists of 34 variables with record of 1470 employees. The employee attrition status has been recorded using the variable Attrition that has been recorded as Yes for those who have voluntarily left the organisation and No for those who have not left. Other variables that have been recorded consist of a variety of employee’s personal and demographic parameters such as age, gender, marital status, education field and education level; job-role based variables such as department, job level, job role; as well as variables of employee’s earnings such as hourly rate monthly rate, percentage of income hike, daily rate, monthly earning and more. There are also work-related variables such as on satisfaction and job performance, environment satisfaction, work life balance.

Statistical Techniques

The dataset was at first subjected to pre-processing in order to spot any missing or invalid data. The data set did not contain any missing values. In an attempt to check data validity at the beginning, a diagnostic test on some interrelated set of variables were performed. Throughout the dataset, these diagnostics were performed to be check whether the numerical value of Total Working Years at the company was greater than the numerical value of the years at the company. Secondly, the numerical value of years at the company must be greater than the numerical value of years in current role as well as years since last promotion and years with current manager for each employee. Similarly, the numerical value of the variable Monthly rate should always be greater than that of the respective value for daily rate, while the latter should be greater than the value of the hourly rate. The data was therefore found to be valid and consistent in these pre-processing steps. In this study, IBM SPSS is used for analysis. The data was the subjected to exploratory data analysis to identify the important variables in the dataset and to understand the underlying distribution of all variables. The categorica variabls such as Attrition, Gender, Job Satisfaction Employee rating, Job Level, Department etc were all transformed and recoded.  The variables with zero variances (i.e. containing only 1 unique value were dropped. Thus, Over18, EmployeeCount, and StandardHours, and got removed. The variable of Attrition was also encoded into a new dummy variable known as “Yes Attrition” which holds the value of 1 for employees who have reported Yes for Attrition and value 0 for those who responded No to Attrition. The variable “Employee Number” was also rejected as it had no implication for attrition. The relationship and association between variables was analysed through cross tabulations, bivariate frequency distribution, Chi square test for independence and independent sample t test.

Results

In the present dataset, 16% (237 out of 1470) were labelled as Yes for attrition that is they had left the organisation.

The age of the employees was found to be between 18 and 60 years of age, with a mean age of 36.92 years. 60% of the employees were males while 40% were female. The dataset consisted of employees ranging from freshers having 0 total working years to 40 years of working years with an average work experience of over 11 years. For the employees, the monthly income ranged from $1009 to $19999 with a mean monthly earning reported as $6503. The employees included in the dataset were only good performing employees as the job performance score of only 3 and 4 were present. The dataset spanned across three departments of Human Resources, Research and Development, and Sales as well as six job roles, namely, Sales Executive, Research Scientist, Laboratory Technician, Manufacturing Director, Healthcare Representative, and Manager, as well as ‘Other’.

A correlation matrix of all numeric variables was constructed in order to check for correlation of the variables with each other. A high degree of correlation was found to be in the following pairs of variables:

• Years at Company and Years in Current Role (r =0.759, p < 0.001)
• Years at Company and Years with Current Manager (r =0.769, p < 0.001)
• Years at Company and Years since Last Promotion (r =0.618, p < 0.001)
• Years in Current Role and Years with Current Manager (r =0.714, p < 0.001)
• Years in Current Role and Years since Last Promotion (r =0.548, p < 0.001)
• Years with Current Manager & Years Since Last Promotion (r =0.510, p < 0.001)
• Total Working Years and Monthly Income (r =0.773, p < 0.001)

Hypothesis 1: Is age is related to voluntary turnover?

The attrition rate was found to be highest for employees in the age group ranging from mid 20s to early 40s. An independent samples t-test results indicate that there is a statistically significant difference between the mean age for those who left IBM versus those who did not (t =6.179, p < .001).  In other words, younger employees are more likely to attrite while older workers are less likely to quit.

Hypothesis 2: Is there an association between employee’s education and their likelihood of voluntary turnover?

It is also seen that, employees having degree in Life Sciences have the highest rate of attrition followed by medical. A chi-square test of independence confirmed that there was significant association between attrition and education field. (X² (5, N = 1470) = 16.025, p<0.05)

Hypothesis 3: Is attrition correlated with Gender of the Employee?

The Attrition rate of Male Employees in the dataset was found to be greater than that for the female employees sampled. While only 12% of total female employees left the company, the total male employees in dataset who left were 20%.  A chi-square test of independence was performed to examine the relation between attrition and gender. The relation between these variables was not significant (X² (1, N = 1470) = 1.275, p =0.259)

Hypothesis 4: Does Employment tenure impact attrition?

Various variables captured in the dataset that represent the tenure of the employees are Years at company, years in current role, years since last promotion and years with current manager. In an independent sample t test there was a significant difference between the mean years spent at company of employees who have left the company versus those who had not(t=5.96, p<0.001); former(5.13 years) being lesser than the latter(7.37 years).

There is a higher likelihood for employees to quit in early years such as after 1 year and until 10 years. After spending 10 years, the employees are highly unlikely to leave. Similarly new joiners who have spent 0 years in the current role had quit the most.  

Also, it can be seen from the total working years graph, that employees with 1 to 10 years have larger proportion of attrition than the senior employees.

Hypothesis 5: Is Employee’s pay associated with the likelihood of voluntary turnover.

Other aspects of attrition pattern visible from the data were found that there was a higher rate of attrition in case of employees earning towards the bottom quantiles of Monthly Income. The average monthly income of employees who had left the company ($6832.74) was significantly lower than that for those who did not leave ($4787.09) (MD = $2045.54, t = 6.204, p<0.001)

Hypothesis 6: Is Job satisfaction significantly associated with attrition?

There was a statistically significant association between Job Satisfaction and Attrition (X² (3, N = 1470) = 17.505, p=0.001).

FEW OTHER INSIGHTS

Marital Status was also found to be significantly associated with attrition (X² (2, N = 1470) = 46.164, p<0.001). Employees with marital status of “Single” had a greater likelihood of voluntarily leaving compared to married and divorced.

Attrition rate was also found to be lowest for employees who do not undertake business travel compared to employees who rarely or frequently travel for business. A chi-square test of independence was performed to examine the relation between attrition and business travel. The relation between these variables was found to be significant (X² (2, N = 1470) = 24.18, p <0.001). Thus business travel is a significant determinant of attrition.

In an independent sample t-test it was found that employees living closer to the office were less likely to voluntarily leave the company compared to those who were living further. The mean distance from home was greater for employees who had left the firm compared to those who did not (t = -2.995, p = 0.003)

While the sample indicated that attrition rate is highest in the Department of research and development and least for the human resources department; a chi-square test of independence was performed to examine the relation between attrition and department. The relation between these variables was found to be significant (X² (2, N = 1470) = 10.796, p <0.05). Thus mean attrition rates were significantly different for three different departments. 

Similarly, a chi-square test of independence confirmed that there was significant association between attrition and job role. (X² (8, N = 1470) = 86.190, p<0.001). The highest rate of attrition was reported for the role of laboratory technician, followed by sales executive and research scientist.

An independent sample t test revealed that employees who left the company had lesser number of trainings in the previous year compared to those who did not leave (t = 2.283, p =0.02) and are more likely to be staying further away from the office compared to those who did not leave(t= -2.995, p= 0.003). This is indicative that employees who are provided more traingings are lesser likely to voluntarily leave.

Conclusion

An analysis of employee attrition was conducted. Attrition was found to be significantly associated with several important variables such as age, Gender, Education, Tenure, Earning, Job performance, Job Satisfaction and Job involvement. Also, additional insights were discovered such as there are several job roles that displayed greater attrition. It may be beneficial for the organisation to conduct localised and qualitative investigations to discover the issues that might be contributing to higher attrition rates among job roes such as Sales Executives and Lab technician and among departments such as Research and Development.

References

Cotton, J., & Tuttle, J. (1986). Employee Turnover: A Meta-Analysis and Review with Implications for Research. The Academy Of Management Review, 11(1), 55. doi: 10.2307/258331

Dess, G., & Shaw, J. (2001). Voluntary Turnover, Social Capital, and Organizational Performance. Academy Of Management Review, 26(3), 446-456. doi: 10.5465/amr.2001.4845830

Economist, T. (2018). The high costs of staff turnover. Retrieved from https://www.economist.com/business/2018/09/20/the-high-costs-of-staff-turnover

Erickson, R. (2016). Talent & Workforce Employee Engagement. Retrieved from https://login.bersin.com/uploadedFiles/021517-calculating-true-cost-voluntary-turnover.pdf

Frith, B. (2017). Turnover rate reaches new high. Retrieved from https://www.hrmagazine.co.uk/article-details/turnover-rate-reaches-new-high

Hoffman, M., & Tadelis, S. (2018). People Management Skills, Employee Attrition, and Manager Rewards: An Empirical Analysis. NBER Working Paper, 24360. doi: 10.3386/w24360

Pritchard CW (2007). 101 Strategies for recruiting success: where, when, and how to find the right people every time. New York: AMACOM

Messmer, M. (2000). Orientations programs can be key to employee retention. In Strategic Finance. 81 (8):12-15.

Edouard Ribes & Karim Touahri & Benoît Perthame, 2017. "Employee turnover prediction and retention policies design: a case study," Working Papers hal-01556746, HAL.

Ho, Jessica Sze Yin and Downe, Alan G. and Loke, Siew-Phaik, Employee Attrition in the Malaysian Service Industry: Push and Pull Factors (March 30, 2010). The IUP Journal of Organizational Behavior, Vol. 9, Nos. 1 & 2, pp. 16-31, January & April 2010.

Archita Banerjee, Rahul Kumar Ghosh, Meghdoot Ghosh, 2017. A Study on the Factors Influencing the Rate of Attrition in IT Sector: Based on Indian Scenario. Pacific Business Review International, 9(7), p. 01

Ongori, H. (2007). A review of the literature an employee turnover. Afri. J Bus. Management, 1(3), 49-54.

Hoffman, M., & Tadelis, S. (2018). People Management Skills, Employee Attrition, and Manager Rewards: An Empirical Analysis. NBER Working Paper, 24360. doi: 10.3386/w24360