Visualising Gender Dynamics in my Twitter Network

Males following males and females following females

For some time now, I’ve been thinking about whether or not who we listen to within our social networks has an influence on our perception and worldview. Do the people that we follow on Twitter subconsciously strengthen certain biases we may hold? Since the issues of gender have been quite pervasive in my thoughts and among both my professional and social circles, and having already played around with the problem of visualizing this issue, I decided to dissect my own Twitter network in search of these hidden influences.

The purpose of this project is purely to visualise how connected the people I follow are to each other, not as a statistical analysis of whether or not people are likely to follow one gender or another, although it may be possible to infer some kind of effect based on these numbers (not that the sample is nearly big enough). What I would like to focus on is the overall network effects that occur as a byproduct of having an imbalance in the diversity of who you follow.

Here are some visualisations I created to explore this topic. There is also an interactive version of these graphs.

Who I follow on Twitter

Who I follow

The above graph shows all the accounts that I follow on Twitter, excluding non-people (eg. businesses) and protected profiles, with each person represented as a single dot. With the socially normative pink for women and blue for men, you can see that I follow around one third women and two thirds men. The proximity of the node to the center signifies how connected this individual is within the people I follow (ie. the closer they are to the center, the more connections they have to other dots).

I am someone who prides herself in making an effort to follow as many inspiring women in my areas of interest as possible, but unfortunately since my area of interests predominantly consist of STEM fields, this limits my ability to follow an equal percentage of women and men. I can only imagine this graph looking even more skewed for those in the field who are not conscious or concerned about diversity in who they follow.

Women following womenMen following Men

Connections within gender

The above graphs show how connected women are with each other and how men are connected with each other. The density of the lines reveal that there are a lot of connections between men compared to those between women, but that is to be expected since I follow twice the number of men compared to women.

When I last updated the data on 2014-07-30, there were 585links between women compared to 2280 links between men.

Who follows women

Connections within gender with number of follows

The above graph has the same data as the previous section, but combined into one graph. In this version of the visualisation, the size of the outer translucent circles indicates the number of followers that individual has within my network. It is interesting to note that although women and men located on the same circular band have a similar number of followers, the women’s outer circles are much smaller than the men’s. This implies that in this network, the majority of women’s followers are men.

Women following men Men Following Women

Women following men vs Men following women

The graph on the left shows how women are following men and the graph on the right shows how men are following women. Interestingly, there seems to be fewer follows to men on the outer reaches of the graph (since they have lower number of total follows within the network), but the men who follow women seem to be fairly evenly distributed.

The number of connections are: 907 women to men and 750men to women.

Who follows womenWho follows men

Follows to women vs Follows to men

The graph on the left shows all the follows to women (from men and women) and the one on the right shows all the follows to men. The line colour indicates the gender of the originator of the follow (pink line means that a woman is following, blue: man). You can see that the follows to women are sparser compared to those to men, from both women and men. However, despite there being fewer women, the links of women following men are still quite dense.

There are 1335 follows to women and 3351 follows to men.

Women following men Men Following Women

Women following men vs Men following women with number of follows

Here is the same graphs as above, but with total number of followers displayed. From these, you can see that men are doing most of the following to both men and women.

The number of connections are: 907 women to men vs 497women to women.

Some Stats

On average, using the mean calculation, each man follows 4.4%of all women (excluding myself) in my network, while each woman follows 6.3% of all men.

If you use a median calculation however, each man follows3.5% of all women, while each woman follows 4.3% of all men.

Within the genders, each woman follows 5.5% of all women, while each man follows 8.5% of all men (median 4.5% and7.9%).

The variability between the mean and median indicates that there may be a few individuals that are skewing the average one way or the other, as implied by some of the visualisations shown already. If I were to postulate, I would say that there are a few notable females that are being followed by a lot of men, but many more are falling through the cracks. However, what is clear is that the men in my network are more likely to be followed than the women in my network.


This viz cannot be used as an objective or statistically accurate representation of biases in my network since it only includes the people I follow and not a complete set my followers and who the people that I follow follow. However, you can make some observations around following choices when it comes to gender and also infer some trickle-on effects that following less women may produce.

There are also many other factors such as race, socioeconomic background and industry that can be used to categorise the network. I would love to see visualisations using these approaches, however it would be quite time-intensive to manually categorise each individual as Twitter does not provide this metadata.

If I were to TLDR this, here is what I observed:

  • The men in my network are highly connected.
  • The women in my network are loosely connected (compared to the men).
  • In this network where there are more men than women, the men do most of the following.
  • Men are more likely than women to be followed by either genders.

I’m not of the appropriate academic background to extrapolate these findings to make concrete assumptions about socialogical broader effects, but here are some guesses of what some effects may occur in a network like this:

  • Women are less likely to have interactions with each other, and therefore back each other up in debates etc.
  • Women are more likely to receive replies from men rather than other women.
  • Both women and men are exposed to fewer female opinions.
  • There are fewer options for women to find female mentors and build social connections with other women in their industry.

I can say that I have felt these things based on my own experiences on Twitter. There is also one study (of a few) that does back up my findings. This study found that despite women being the majority gender on Twitter as a whole, both men and women are more likely to follow a man than a woman, with men being 50% more likely and women 25%more likely.

More recently, Mikki Kendall, a black woman, conducted an interesting social experiment where she replaced her profile picture with that of a white male’s. Despite tweeting the same type of things she had in the past, her interactions changed dramatically. People were more respectful, there was less trolling and insults, more people paid attention to what she had to say.

It is quite unfortunate to think our experience on social networks can be so dramatically affected by qualities like race and gender, which are things we have no control over. This happens often in the real world, but the internet does represent this ideal egalitarian place where anyone with access to it can create something meaningful and connect with others in a positive way. This is sadly an overly optimistic perception and one that is more likely to be held by those who do not greet these biases day-to-day.


After completing We Can Do Better, which was a very limited and limiting dataset (just company names and employee numbers), I wanted to tackle something a little deeper and complex. One thing I realised while exploring this data, was how much more rewarding a richer, more connected and more contextual dataset could be to work with. Despite the higher investment required to traverse this data, I was much more motivated by the aesthetic potential of this data.

As much as I desire a more in depth analysis on my Twitter network, with a complete picture of who the people I follow follow, the current version of the Twitter API is very limiting for collecting this type of data. Because of their rate-limiting constraints (15 requests per 15 minute window), it takes me approximately 4 hours to query connections for the ~300 people I follow. (If someone who works at Twitter would like to whitelist me, please email me).


The interactive version of this viz was built in D3 using the force layout and the SVG image files pulled from that. Feel free to try this on your own network! However, there is some work you have to do beforehand to get gender and connections. You can view the code on github.


For those that are curious, here are some aesthetically interesting shots collected during the process of this project:


Written by Ri Liu on 4 Aug, 2014


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s