Coast & Gender — Female vs. Male & East vs. West — The Startup Investor Ecosystem Revealed

Part 2: Popularity by Gender, Bias & Loyalty

If you missed it, see Part 1: Coast & Gender — The Startup Investor Ecosystem Revealed.

How does Twitter friending stack up across the Cohort by gender? It turns out, pretty well! One can see from this chart that women and men are equally interested in tuning into the platform. Neither men nor women seem bunched up at either end of the chart, and women seem well-represented at the highest, 5,000 friend end of the chart, even though they make up only 35% of the cohort. This looks good!

Who are these very tuned-in individuals? Here are the top 25:

As you can see, even though there’s an upper limit of 5,000 friends, less than 10% of the cohort have maxed out. One can presume these are very active and engaged members of the ecosystem who spend (or at one time spent) a lot of time on Twitter. And 32% are women, in a Cohort in which 35% are women. So far so good!

Popularity by Gender

While we can learn something about an individual by how many Twitter users they have friended, we can learn something about the Cohort by looking at who within the Cohort is friends with whom: how popular they are with their peers. The following graph show how many followers from the Cohort each member of the Cohort has.

This looks less good. If you look at the entire upper quarter of the graph — the 50 most popular members of the Cohort — there are only four women. So

Within the investor community, women are not being followed nearly as much by their peers than are men.

Let’s dive in deeper to see what the uppermost tier of the popularity curve shows.

OK, Fred Wilson being at the top (again) works for me: he’s pretty famous and he’s been doing this a long time. Whichever way we’ve looked at it, according to Twitter, he’s the most popular member of the Cohort.¹ But… where are the women? Their peers do not seem to be following them at the same rate as are the men. Only Marissa Mayer and AileenLee cracked the 25 Most Popular list. Sure, women only make up 35% of the cohort, but they’re only 8% of this Top 25 list. Can we dive any deeper into this?

Absolutely. First, we can re-plot who from the cohort who is following women.

OK, that looks a little better. Though there’s a swath of zeros at the bottom of the graph, meaning there are members of the Cohort being followed by no women, women seem a bit better represented in the upper quarter of the graph. Which means women are friending women at a better rate than the Cohort as a whole.

Let’s dive in deeper. Here what things look like at the top of the previous graph:

Fred is of course still at the top. But AileenLee is now number two! And whereas there were only two women in the original Top 25 popularity contest, here there are seven. So women are friending women at a higher rate than the overall Cohort friends women. But what about the reverse? Who’s the most popular in the eyes of the men?

Fred! Still! #1! And other than: 1) Marissa Mayer and Aileen Lee switching relative places; 2) Reid Hoffman, Chris Dixon and Bill Gurley trading the various runners up spots, and; 3) some other minor jockeying among the men, this doesn’t look much different than the Cohort as a whole. But clearly, things aren’t great. Women are under-represented at the top of all the charts. Clearly…

Women are being followed by men at lower rates than men follow other men.

Wait, have I been using the word “rates”?

Absolute Numbers versus Percentages

Yep! I’ve been talking about rates. But do these graphs really tell us anything about the rates of gendered friendship? It only shows the absolute number of friends for each member of the cohort. And since the cohort skews male, it’s not surprising we see more men at the top of various graphs. If every member of the Cohort randomly liked some other members of the Cohort, each man would friend approximately twice as many men as they would women. The average female would be friended by twice as many men than women. Each woman would have on average two male friends for every one female friend. So does these graphs looks OK when taking that into account? Here’s one of the high-level summary graphs for reference:

Sort of. They show that women are being friended by both men and women in somewhat similar, absolute numbers. But there’s something missing in this analysis. Not everyone has the same number of friends! A guy at the top of a numbers-based the chart might look like a gender-parity champion due him friending 45 women in the Cohort (well over 50% of the women… what a guy!); but, what if he was also friends with all 130 men in the Cohort? That would be a female-friend-rate in the Cohort of only 26%, much lower than the 35% one would get from just random friending. So what we don’t know from these graphs is whether or not — in general — men who are friends with the most number of women also have a reasonable balance of female to male friends. We don’t know at what rates, in the Cohort, are men friending women versus women friending women, and vice versa.

Bias

So, instead of absolute numbers, let’s look at percentages. Who in the Cohort is the most popular based on the percentage of women or men who have friended them?

To clarify this chart’s meaning, and to use Leslie Feinzaig of Female Founders Alliance (at #2) as an example: she is followed by 14 members of the Cohort (that’s the number is parenthesis) and approximately 80% of those followers (y-axis) are women.

Let’s also look at the reverse:

Yes! Super interesting! We’re about to dive in deeper. Stuff needs to be said. But before we do, there’s something we need to fix. Do you see it? Because we’re working with percentages, people who have only been friended by one person in the Cohort are at 100% on one of the two graphs. They’re at the top! That’s no good. We can draw conclusions from someone with only one Cohort friend of either gender. Two also sounds like too little. For us to make fair conclusions about gender bias, we need to eliminate the outliers. We need a cutoff. I chose ten.

Let’s redraw those graphs but only include members of the Cohort who’ve been friended by ten or more members of the Cohort. And since the friend/follower terminology can be confusing, just a reminder: folks who friend other folks are also those other folks’ followers.

Gender bias for one’s own gender in the Cohort is unequivocal.

Women consistently friend women at a much higher rate than they do men, and men consistently friend men at a much higher rate than they do women. Full stop.

But one more thing ought to be mentioned. Though I’m slinging around the word bias, please remember that it does not apply to individual members of the Cohort. The data has meaning at a Cohort level, not at an individual level. We have demonstrated a trend within the Cohort: nothing else. While the data calls out certain individuals for having friends in the Cohort that are predominantly of their own gender, it doesn’t know anything about the gender of their friends that are outside the Cohort.

Loyalty

Since I’m referring to “being friended by one’s peers” as popularity, I’m going to refer to “friending one’s peers” as loyalty. What does this mean gender-wise? The female chart first, then the male:

Jenny Fielding wins this one hands down! She’ the most loyal member of the Cohort for both men and women. (It’s nice to finally see a chart without Fred Wilson at the top!) Where’d Fred go? While Fred is himself super popular, and has over has friended over 1,000 individuals, he’s only friended 38 members of the Cohort. No competition for Jenny. And I’m happily surprised at my sudden and unexpected appearance on a chart (#2 in the “friending of female peers” category). General conclusion: women and men are loyal to their gender.

But hold on. Before we jump to any gender-self-selection-bias conclusions, which the data seems to support based on casual inspection, don’t we need to do the same thing we did earlier? Don’t we need to look at the data in terms of rates/percentages? Yes we do! If we care about gender imbalance in this ecosystem, then we care about the percentages. I’ll skip the step where I show the data with outliers. The following charts show how things look percentage-wise with outliers removed (female chart first, then male):

There’s a strong gender self-selection bias in the Cohort.

Whoa! 80% of the Top 25 Cohort member’s rates of female friendship are held by women, while 100% of the Top 25 Cohort member’s rates of male friendship are by men. I guess that’s no surprise given what we’ve seen.

What are those horizontal lines on the chart? They show the percent of female and male Cohort members. Remember the idea mentioned earlier of would happen if Cohort members randomly friended other Cohort members? 2/3 would be men and 1/3 would be women? The lines represent a zero bias. So what happened to Jenny Fielding? She went from the #1 loyal member of the Cohort to… off the chart? No worries. She’s still a super-connected and well-respected member of the ecosystem. She just happens to be just off the left-hand chart at 35%… right on the zero bias line!

The Final Story on Bias

So. What does the big picture look like for bias? I guess we’d want compare the rates at which the different genders favor themselves. If women and men are favoring their own genders at the same rates, then while that would be evidence of gender self-bias, the self-bias would be equal for each gender. That seems like it would be OK. Or at least it would be more OK than if one gender was substantially more self-biased than the other. My prediction? Equal self-bias.

OK, let’s look. After removing the outliers, here’s what we see:

Male investors have an extremely strong bias for following other men!

Whoa! This is surprising in the clarity of the dividing lines. Men in the Cohort have a very strong gender bias for men in the Cohort! Nearly all the men in the Cohort friended a greater percentage of men than the “unbiased rate” of 65%.

Women do not show as strong a bias for their own gender as do men.

The converse for women in the Cohort is not nearly as black and white². Women, while somewhat less biased in gender selectivity, have done the opposite by not overwhelmingly favoring their own gender. Though a little over one third of women friended a greater percentage of woman than the “unbiased rate” of 35%, almost two-thirds of women friended men a higher rate than the unbiased rate!

Up Next

Stay tuned for Part 3: The Network; and the concluding Part 4: What’s in the Tweets?

And if you missed it, see Part 1: Coast & Gender — The Startup Investor Ecosystem Revealed. It has the study’s methodology as well as high-level insights into the Cohort’s Gender and Coast preferences.

Footnotes

[1] He may have invested in Twitter, so the fix might be in!

[2] Salmon and teal.