I have collected employee turnover rates over the year. The attrition rates are as such:
| Month | Attrition Rate |
|---|---|
| Jan | 2.23% |
| Feb | 1.40% |
| Mar | 2.99% |
| Apr | 1.43% |
| May | 1.72% |
| Jun | 1.15% |
| Jul | 1.64% |
| Aug | 1.60% |
| Sep | 2.00% |
| Oct | 2.15% |
| Nov | 1.08% |
| Dec | 1.30% |
I'm comparing these attrition rates to the industry benchmark rate of 1.44% by using a One-Sample t-test.
Is this correct, and if not, what is the best way to confirm if the Mean Attrition Rate in this year is significantly higher or lower than the industry benchmark?
I need to be able to make a statement and say that: "Year XXX has an employee attrition rate of X.X%, and this is (not) significantly higher/lower that the industry benchmark of X.X%"
In addition, if I were to compare Mean attrition rate of this year versus the previous year, is it correct for me to do a paired samples t-test, or should I do an Independent 2 Sample T-Test?
EDIT
After giving this more thought, I'm thinking that if I were to test the means of percentages using the the t-test, what I am comparing is effectively the proportion, between 0 to 1, as a whole whether the proportions are significantly different…
However if I tested just the actual turnover headcount figures, I would be comparing the mean turnover headcount of a group of months vs (let's say for example) the mean turnover headcount vs a previous month… This would be a more direct measure of turnover, but because it's not a proportion, it's less meaningful because it's not based on any thing. I wanted to compare turnover rates because turnover rates = Leavers / Total Employees. The denominator here plays a significant role in moderating the turnover rate.
Any advice/thoughts/comments?
Best Answer
It makes sense to use the attrition rates so you can compare them with those of other companies, which usually have a different amount of employees. The industry benchmark also makes only sense as a rate.
If the attrition rates within the year are approximately normal iid, it makes sense to use the one-sample t-test to compare to the industry benchmark.
And, finally, if you want to compare the attrition to the previous year, and you presume the attrition to be dependent on the month (e.g. there is more attrition in the winter than in the summer), it makes sense to use the paired t-test.