Why do employees want to leave their jobs?
In the new era of ‘job hopping,’ employers have started to focus more and more on job retention. Fortunately for employees, this means that many companies are attempting to provide more perks, improve job satisfaction, and overall improve the culture and diversity of their workplaces.
But what are the factors that correlate with an employee wanting to leave in the first place? That’s what I was curious to look into for my data science bootcamp data exploration project.
Fortunately I found a giant, publicly available dataset to help me look into that, through the lens of federal government employees. The United States Office of Personnel Management distributes the Federal Employee Viewpoint Survey (FEVS) every year, and publishes their results.
The survey asks for employees’ opinions across several sections entitled ‘Work Experience’, ‘Work Unit’, ‘Agency’, and ‘Satisfaction’, and collects basic demographics, including whether or not someone is thinking of leaving their organization in the next year.
For this article, I looked into the most recently available data from 2018. Just over 598,000 of the 1.4 million Federal employees who received the survey completed it, between April and June 2018, representing 82 agencies.
For the purposes of my analysis, I narrowed the surveys down to those which were fully complete and in which the responses indicated either ‘Yes’ or ‘No’ to whether an employee was considering leaving within the next year (either to another Federal job, or outside the Federal Government). This left me with just over 240,000 surveys.
So what did I find? Overall, the group of employees who were thinking about leaving in the next year answered, on average, with less satisfaction or agreement on every survey question. The chart below contains a lot of data points, but I wanted to show the clarity of the difference. Overall it’s pretty apparent how divided these two groups of data are, though they follow the same trends. In fact the only two questions that you see that come closest to overlapping, Question 7 and Question 8, are the two questions about personal performance and work ethic. See this PDF report for a list of all questions.
So what factors can we pinpoint, if satisfaction is lower across the board? I looked at the correlations of each question with the response on the ‘Leaving’ question. See below for a visualization to see how strongly the negative correlation was with each question. Remember a negative correlation means that if someone answered with less satisfaction on a question, they were more likely to have indicated they wanted to leave their job.
I found the five questions with the highest negative correlation to the ‘Leaving’ question, and looked at them in a little more depth.
Here are the questions:
Q69: Considering everything, how satisfied are you with your job?
Q71: Considering everything, how satisfied are you with your organization?
Q40: I recommend my organization as a good place to work.
Q67: How satisfied are you with your opportunity to get a better job in your organization?
Q11: My talents are used well in the workplace.
In this box plot, you can see that, for those employees planning to stay in their job, most of them answer of these questions in the satisfied to very satisfied range, with only outliers answering as dissatisfied or very dissatisfied. For employees planning to leave, however, there is a wider spread of answers, and they tend to be more neutral or negative.
Of course, it makes sense that overall job satisfaction and organization satisfaction would be highly correlated with someone wanting to leave. Notably absent from this top five list of correlated questions, however, is overall satisfaction with pay (Q70). Based on these outcomes, it seems that it is even more critical for employees to feel their organization is a positive place to work, that there are opportunities for upward advancement, and that their talents are used well.
There is so much more information that could be gleaned from this dataset. Of course, it may not be generalizable to the population, since it is a specific subset of Federal Government employees, and I should also note that I did not weight the data to be reflective of the Federal employee population either (see here for more information on their weighting system).
However, I think it’s important nonetheless to continue to learn as much as we can about how to create work environments and organizations that people just don’t want to leave.
This analysis was completed as part of a student project in data exploration for the first unit of the Data Science track at Lambda School. See my full Python notebook in GitHub.
