Girls are not more likely to drop out of CS and Engineering

It is often claimed that women are more likely to drop out of computer science and engineering than men. Is this claim supported by evidence at Waterloo? Let’s take a look at the enrolment statistics to find out!

The dataset contains about 88000 rows, each row indicating the number of students with a particular set of attributes: for example, number of male, first year, international, CS students not in co-op, etc.

To determine the drop-out rate of various programs, we track groups of students into “cohorts”. A cohort is a group of students in the same program that entered in the same year, so for example, we can track the number of students in first year CS in 2009, then second year CS in 2010, third year CS in 2011, fourth year CS in 2012, and so on.

For each program, we look at 5 cohorts for classes with starting years from 2009 to 2013, and plot the total number of students and proportion of female students in each cohort from first year through fourth year.

Here are the plots for a few engineering programs:

In the top graph, the cohort decreases in size, which is expected as students drop out. There does not appear to be any significant change in gender ratio within the cohorts from first year until graduation, implying that men and women are dropping out at an equal rate.

To be sure, I fitted a linear regression model to the bottom plot. The t-test measures how strong of a relationship exists between the variables: it agrees that there is no relationship between gender ratio and student year (p >> 0.05 in all 4 cases).

In computer science, surprisingly the proportion of girls actually seems to increase as we get into upper year. (p = 0.03)

Before we draw too many conclusions, I should note some disclaimers.

I’m tracking cohorts as if everybody followed a traditional 4-year program. In UW, many students have co-op, taking 5 years instead of 4. It’s difficult to properly account for various co-op streams with this dataset, so this methodology is only a rough approximation. Assuming that gender ratios stay relatively constant from year to year, this error shouldn’t bias the result.

Also, the dataset doesn’t show if a student transfers to a different program, so the analysis would be less accurate in faculties where switching programs is common. In the math faculty, many students start in general “Math” and eventually specialize in second or third year, so it’s hard to deduce drop-out rates.

That’s it for now, I don’t see any evidence that girls are more likely to drop out than guys in CS or engineering, at least once they’ve enrolled into a university program. Are they more likely to leave the industry after graduating? That’s a statistical analysis for another day.

Data from UW Student Headcounts