Have you ever pondered how long an employee will work for your company or how long you will remain employed? Continue reading to learn more.
As I was strolling around the streets of Netherlands, I counted how many "Help Wanted" signs there were. I found it incomprehensible that almost every restaurant and shop I passed needed assistance. Then I got to Schiphol Airport and saw few employees but a ton of suitcases—suitcases literally everywhere. This is due to the staff shortages at the airport, specifically related to baggage claims and security. Upon conducting a short, but insightful, Google search I came across some fascinating statistics regarding the Dutch labour market- at the moment, Netherlands has a negative unemployment rate sitting at –9%. This is attributed to labour market expansion, which is reported to be ongoing without an apparent end in sight.
As an actuarial graduate, I decided to test my modelling prowess on something slightly different. This may be the start to applying actuarial models not only to people's lives or in the financial industry, but instead to their 'survival' in a particular role. This is referred to as 'Survival Analysis', a modelling method used to represent the time leading up to an event and it is used to estimate how long an event will take to occur.
To do this, I developed a Python script for forecasting the likelihood of remaining in a specific role as well as the impact of certain factors impacting this probability using statistical models, Kaplan Meier & Cox-Regression Models, which I will spare you the details of. Using these complex (not so complex, with the help of Python) statistical models I decided to take this current issue in the Netherlands and figure out how to apply my actuarial background to it. As easy as pie(thon).
As a starting point, we may consider the number of years someone has worked in a job and the likelihood that they will continue in that capacity.
Above, it is evident that the highest probability exists of those leaving a role is in the first year of work. If we want to take this a step further, we might argue that there is a 98.57% chance that an employee will leave their position during the first year of employment. Interesting but also anticipated.
Now let us analyse more specific aspects including:
Work life balance
Salary
Job status
Work environment
Salary Increases
A score was assigned to each of the aforementioned characteristics between 1 (poor) - 5 (excellent). What shocked me about the results of the model is how highly 'Job status' scored and how poorly ‘Salary Increases’ scored.
It demonstrates that people are more concerned about the title of their LinkedIn profile rather than what is in their bank account at the end of each month!
One factor that also piqued my interest was ‘Work Life Balance’ – this one is for us workaholics. Specifically examining a factor such as ‘Work Life Balance’ we find that the likelihood of leaving a company due to ‘Work Life Balance’ is greatest in the first two years of employment. Consequently, individuals with higher ‘Work Life Balance’ tend to remain longer at a company compared to those with lower ‘Work Life Balance’.
The model may further anticipate how long each employee is likely to remain with the organization. Alongside, the numbers 1-6 represent each unique employee (for example: 1 along the x-axis may represent "Bob") and 0.0-8.0 indicate the number of years worked at the company. The values filled in the table show the probability of leaving after ‘x’ number of years worked. For instance, employee 4’s probability of still working at the company after 8 years is 77.18% whereas employee 6’s probability of still working at the company after 8 years is 93.74%. We can make this even more personal and bring employee number 1, Bob, back into the picture. Bob is 88.34% likely to still be working at the company after 7 years have passed.
This approach could be used to tackle the Dutch labour crisis to predict when and which workers are most likely to leave. More so, since each company has its own unique characteristics which influence an employee's likeliness to continue their employment, this model can essentially help a firm identify their weaknesses and play on their strengths to truly take advantage of this model. Firms may concentrate more on reducing employee loss and ultimately construct a more engaged, productive, and satisfied workforce. As Dough Conant said, "To win in the marketplace, you must first win in the workplace".
About the author: Nathalia is a Greek South African with a Bachelor of Science degree in Actuarial and Financial Mathematics from the University of Pretoria in South Africa. In her free time, Nathalia enjoys watching horror movies, making cosmetic products, and Latin dancing. Nathalia’s studies focused on risk, as well as the analysis of consequences of that risk from a statistical, mathematical and coding perspective. Despite enjoying the course work, Nathalia realised she wanted to work in the field of data science and explore her own capabilities within this field. In addition to her passion for data, Nathalia has always been passionate about the environment, sustainable alternatives and human rights issues.