Search

What Drives Employees to Leave? Using Regression to Reveal the Truth

Michael Lee, MBA
12 hours ago
3 min read

The Power of Regression Analysis in HR

Ever feel like some employees are already halfway out the door—and you can’t quite explain why? Let’s use data to find out.

Welcome back to our Inferential Analysis Series. So far, we’ve explored proportion tests, Chi-square analysis, correlation, and t-tests. Now, we turn our focus to the statistical workhorse that helps you predict outcomes: regression analysis.

If you’ve ever wondered what really drives employees to resign, this one’s for you.

What is Regression Analysis?

Regression analysis examines the relationship between one dependent variable (what you want to predict) and one or more independent variables (the predictors).

In HR, that might look like:

Predicting attrition risk based on job satisfaction, years at company, or number of projects.
Estimating training effectiveness based on feedback scores and participation.
Forecasting promotion chances from performance scores and tenure.

✅ Use regression when your dependent variable is numerical and your goal is to understand or predict based on contributing factors.

Today’s Case Study: Who’s Likely to Leave?

We’ve simulated data for 30 employees, with the following indicators:

Years at Company
Number of Projects
Work-Life Balance Score (1 to 5)
Job Satisfaction Score (1 to 5)

Our target variable is their Attrition Risk Score on a scale from 0 (low) to 100 (high).

Years	Projects	Work-Life Balance	Job Satisfaction	Attrition Risk
1	5	2	2	82
2	6	3	2	78
3	4	3	3	72
1	7	1	1	90
4	5	4	4	62
...	...	...	...	...
9	2	5	4	32

Building the Regression Model

We fit a multiple linear regression model using the four predictors.

Hypotheses:

H₀ (null): None of the variables significantly predict attrition risk.
H₁ (alternative): At least one variable significantly affects attrition.

Regression Equation: Attrition Risk = β₀ + β₁(Years) + β₂(Projects) + β₃(WLB) + β₄(Satisfaction) + ε

Results: What Did We Find?

Here are the updated coefficients from the actual dataset:

Intercept: 96.93
Years at Company: -4.80 (p = 0.000)
Number of Projects: -0.06 (p = 0.967)
Work-Life Balance: -1.15 (p = 0.476)
Job Satisfaction: -2.86 (p = 0.106)

Interpretation:

Years at Company: The only statistically significant predictor. Each additional year reduces attrition risk by 4.80 points.
Number of Projects, Work-Life Balance, and Job Satisfaction showed negative relationships but were not statistically significant (p > 0.05).

While “Number of Projects,” “Work-Life Balance,” and “Job Satisfaction” showed negative associations with attrition risk, they were not statistically significant predictors in this model.

The one clear takeaway? Tenure matters. Employees who’ve been with the company longer are significantly less likely to leave.

That said, don’t ignore the other factors—low significance may result from small sample size or measurement limitations. Future models with more data may uncover stronger signals worth acting on.

Is This a Good Model?

Let’s look at the goodness-of-fit metrics from Excel:

R-squared = 0.940
Adjusted R-squared = 0.930

What does this mean?

Our predictors collectively explain 94% of the variation in attrition risk.
Adjusted R² gives us a better picture for multiple regression: it tells us that even after adjusting for the number of variables, our model still explains over 93% of the outcome.

✅ In short: this is a very strong model, but not all individual predictors contribute equally.

Why This Matters to HR

Regression analysis allows HR to:

Defend policies with data, showing how factors like tenure affect retention.
Build predictive dashboards that flag high-risk individuals early.
Focus retention strategies on what matters most—employee tenure.

This isn’t just analysis—it’s action-ready insight. For example, your HR team might consider initiatives that reward loyalty or tailor onboarding to improve early employee retention. For instance, your HR team might decide to limit high-performer project overload or invest in better WLB policies.

Final Thoughts

Regression analysis doesn’t just tell you what happened. It predicts what might. That’s powerful—especially in talent management, where anticipating churn can save costs and retain talent.

If you’d like to learn more through hands-on experience, join our 2-day course Problem Solving Using Data Analytics, or for those curious about how Generative AI can boost your analytics capabilities, check out our Data Analytics in the Age of AI program.