For a while now I had a theory that Software Engineers are more likely to have more boys than girls. I know it sounds like a stupid idea, but I had some strong anecdotal evidence to suggest that I might be onto something.
Anecdotal Evidence
Here is the list of past 3 companies where I worked at and break down for my co-workers who are Software Engineers and have kids. I would like to note that my co-workers are a decently diverse group by Software Engineering standards.
- Citrix
- 5 co-workers with kids
- 7 boys
- 1 girl
- Support.com
- 2 co-workers with kids
- 4 boys
- 0 girls
- Apixio
- 6 co-workers with kids
- 8 boys
- 5 girls
- (Plus a designer with 3 more boys that I did not include into this count)
In additional I have 2 boys myself, so I moved the bias even further to the boy side. Even with my sons of the picture, however, the anecdotal evidence was strong enough to at least make me curious to investigate this further.
Poll
I’ve joked with our designer (the one with 3 boys) about getting a poll together to test my suspicion for some time and few days ago I decided to go for it.
Without putting much thought into it, I’ve set up 1 question poll on Twitter, that looked as follows:
Software Developers with Kids that you know, are more likely to have:
– Boys than girls
– Girls than boys
– About the same amount
In hindsight, I should have put more thought into it, but it was meant just as a quick and fun experiment.
Results
I’ve asked my connections on Slack, Twitter, Facebook and LinkedIn to vote on the poll. In addition I’ve submitted the poll to Reddit and Hacker news.
After 5 hour or so, I was able to gather 27 votes.
27 Votes
The 27 votes looked as followed. Unfortunately this is the best screenshot that I have.
After that time all my connections that cared have voted, my Reddit link has been downvoted to hell, and my Hacker News link flagged and removed.
67 Votes
I feared that I had too small of a sample and that it was being skewed by my connections knowing same group of people that I did.
For that reason I decided to pay 40 more people to take the poll via the Mechanical Turk, bringing the total vote count to 67.
Here is my original poll for your reference.
Software Developers with Kids that you know, are more likely to have:
— Alex Kras (@akras14) January 17, 2017
Analysis
As I’ve said earlier, I did not put much thought into setting up the poll, so I wasn’t really sure how to interpret the data.
I’ve asked some feedback from the host of the Data Skeptic podcast Kyle, as well as some of his listeners. Kyle suggested, that assuming that people are equally likely to have kids of either gender, we would expect the data to be uniformly distributed.
In other words the third option “About the same amount” doesn’t really matter, but we would expect to see “Boys than girls” groups to roughtly equal “Girls than boys” groups.
So break down of 2%, 2%, 96% would be in line with the expectation as would a break down of 30%, 30%, 40%.
As you can tell from the data, the final break down is 34%, 18%, 48%, so it does show some bias towards boys.
Confirmation Bias
One of the biggest leasons from this expriment for me was realising just how strong Confirmation Bias can be.
As I’ve mentioned before, I have 2 boys, so if you add those numbers into every work place that I’ve worked at, I couldn’t help but be convinced that I was onto something.
Throughout this experiment I had to fight really hard my desire to bend the outcome to match my expectations.
Another interesting note is that when looking at my groups of friends who are Software Engineers and have kids, the break down is:
- 3 people
- 3 boys
- 3 girls
I failed to “noticed” this fact until I sat down to put together numbers for this post.
Conclusion
So do I still believe that Software Engineers are more likely to have more boys than girls?
On one side, the data that I collected seems to weakly agree with my initial notion. On the other side, I think the data is close enough to the equal breakdown that it does not warrant much excitement.
I would love to get my hands on a large dataset that links occupation to number of children by gender. If you are aware of such dataset, please let me know.
For now, my two biggest takeaways are:
- If you are going to run a poll, make sure to take the time to set it up properly
- Watch out for Confirmation and other biases
I do still think that it is possible that working in Software Engineering roles and other similar occupations COULD be linked to higher probability of having male children. That being said, correlation does not imply causation and I could think of a lot of other factors that could be at play here (geographic, lifestyle etc).
What do you think?