Everybody Writes, Part Two
December 14th, 2007(This post is the second part of a two-part series on “Everybody Votes”, a channel on the Nintendo Wii dedicated to voting on everyday topics. The first part is here. If you read any of those links they’ll bring you up to speed. You might be able to make sense of this post without reading any of them, but it’d probably be a lot of hassle. So, to prevent a lot of repetition, I won’t be going over what “Everybody Votes” is any further than that, or how I got the data.)
Last time I went over general trends in “Everybody Votes” voting data. This week, I’ll be going over differences in voting by gender. (Let the gender wars begin!)
But before I do that, let me tell you the answers to the three “Everybody Votes” questions I posed last week (if you haven’t looked them up already). People think funny costumes are better than scary ones (57%-43%) and that a broken mirror is more unlucky than Friday the 13th (55%-45%). But the biggest surprise was that the majority of people are not afraid of speaking in front of groups (52%-48%). While prediction accuracy on the first two was around 40% (very low since 67% is average), on the last question it was 24%, much lower than any other. (Assuming people told the truth, which is always an issue on self-assessment questions.) Kudos to you if you got all three right! (And hey, if you didn’t, at least you’ve got a lot of company.)
But I digress. Prediction accuracy and general trends are only one part of “Everybody Votes” data. The other major component breaks down how men and women vote differently: that is, by gender. Last week we saw some notable trends in people’s voting in general, so, do men and women have distinct trends too?
At first glance, no. The average voting spread for men (between the two choices) was 33.06%. For women, it was 31.00%. Those spreads are usually in the same direction, as well. That is, there’s rarely a case where men vote exactly opposite women with the same spread. (So instead of men being from Mars and women being from Venus, you might say men are from Mars and women are from Phobos - one of Mars’ moons.) Here’s the same point graphically, with 1st Choice votes on top and 2nd Choice votes on bottom:
This confirms a minor trend I noticed. Namely, women vote less extremely than men (i.e. they have smaller voting spreads). Though it’s not obvious from my chart, you can clearly see this effect in other places, such as the Internet Movie Database (IMDB). If you look at male vs. female rankings of movies, women rate movies less highly than men but also don’t rate them as low either. In particular, you can tell that women are kinder to the worst movies.
(I’ve only seen one of the movies women rate extremely poorly: “House of the Dead”. It’s clear they think it’s the worst, though. I can tell you from experience it’s laughably bad; it’s tough to imagine any woman identifying with role models like Liberty, who wears a red, white, and blue leotard through much of the movie.)
Ironically, my favorite movie, “Lord of the Rings: Return of the King”, is ranked much higher by women than men. (Maybe I played with dolls as a kid? You tell me.)
At any rate, men and women also differ a bit in how they vote - the average difference in voting spreads for men and women is 9.51%. Here’s a graphical comparison for all 45 polls:
As the title of the chart makes clear, men and women usually differ in their voting spreads by 10%, but occasionally there are much wider spreads. For example, men and women differ most on stuff relating to gender roles, as you might expect. Questions like: “How do you react to thunder and lightning? Fascinates me or Scares me?”, “When lost, which are you more likely to do? Ask for directions or Find it without help?”, and “Do you believe in fortune-telling?” Of course more men say thunder and lightning fascinates them, that they don’t ask for directions, and that they don’t believe in fortune-telling. Women often feel the opposite. This is standard fare (scroll to the bottom) for stand-up comedians, but it’s nice to see it confirmed by data. (”Did you hear the one about the guy who got hit by lightning after getting lost because he ignored the fortune-teller?” Ok, maybe my routine needs a little work…)
Interestingly, men and women also feel most similar about questions that have to do with gender roles - at least in part. Questions like: “You have plans when it starts raining. What do you do? Keep my plans or Change my plans?”, “Are you afraid of speaking in front of groups?”, and “Which would you rather have more of? Intelligence or Confidence?” The difference in spreads for all three of these is about half a percent, and men and women polled within 2% of each other. I would think that men, being macho types, would say they care less about rain, speaking in front of groups, or wanting to be smarter, but the difference is not that large. Shows what I know! (I don’t often hear stand-up comics joking about this kind of stuff, though, so maybe I should have known. Psychologists out there, take note - watching stand-up counts as research.)
Still, there’s even better data you can get from the breakdown of male and female votes. With a little bit of finagling, you can guess how many men vs. women vote on each poll. All you have to do is look at the overall spread (which I covered in the previous post) and compare it to the male and female voting spreads. That tells you what component of the overall spread is men, and what part is women.
(Warning: as in the previous post, this will take a bit of confusing math. I’ll try and keep things as simple and intuitive as possible, though.)
This is best illustrated by example. Take the poll, “Do you believe in fortune-telling?” (It was a worldwide poll, by the way.) 30.10% voted “Yes”, and 69.90% voted “No”. So, the spread is 69.90% - 30.10% = 39.80% (fairly extreme).
Men voted 24.60% “Yes” and 75.40% “No”, for a spread of 50.80%. Women voted 41.90% “Yes” and 58.10% “No”, for a spread of 16.20%. It’s fairly obvious that 30.10% for “Yes” overall is much closer to the 24.60% for men than the 41.90% for women. In fact, it’s 11.8/5.5 = 2.15 times closer, roughly. In other words, women’s results are about 2.15 times further away from the total result than men. Thus, you’d expect that 2.15 men voted for every woman, since they were able to close the gap between themselves and the poll by 2.15 times as much as women and had 2.15 times the influence on the result. (Sorry hermaphrodites and eunuchs, I don’t think there’s a separate category for you on there.)
With that in mind, here’s a graph of estimated men vs. women voters (measured by the number of men voting for every one woman):
As you can see from my example and the graph, the average estimated number of men to women voting is 2.62, or around 2.5. This confirms what I’ve seen on “Everybody Votes” personally: men usually matter much more in the votes. Also, this is not a weighted average, of course, since I don’t have the total number of voters. As I mentioned before, this probably “overweights” early votes and polls without a lot of voters, and “underweights” later votes and polls with lots of voters. I don’t think this is a big deal, though, since the ratio of men/women seems fairly constant.
Men vote more on questions such as “Where do you like to sit on a bus? Near the front or Near the back?” and “How many people do you live with? Four or more or Three or less?” (Represented by the two arrows on the chart.) Funnily enough, women only voted a lot more on one question: “Which do you spend more time doing? Surfing the internet or Watching TV?” (Represented by the other arrow on the chart.) None of these questions had high vote spreads anyway, so most of these gender disparities are probably due to the mathematics of close votes (which inflate ratios). On that family question, the ratio was so extreme women had no influence on the vote whatsoever. (Maybe a lot of women are too busy raising said families to vote. I can just imagine them yelling their votes to their husbands over the wails of screaming children.) Since the ratio of men to women would be infinite in that case (dividing by zero, yuck!), I just assumed a more likely ratio of 10 to 1. I don’t see any striking patterns here, other than the fact that a lot more men vote on “Everybody Votes” than women.
What about prediction accuracy, then? You might wonder if there any relationship between the male/female voting spread, the male/female voting ratio, and prediction accuracy. Unfortunately, there’s probably no relation. Here’s a chart so you can see for yourself (keep in mind the spread and prediction ratio are on the left axis, and the voting ratio is on the right axis):
That graph can be a little confusing, but you can see by looking at it that none of the three lines seem to be correlated (either positively or negatively). They mostly seem independent of each other. That is, how men and women vote and the ratio of men to women voting does not seem to affect how well they predict as a whole, or vice versa. This somewhat contradicts my theory in the last post that people predict other people voted like they do. If they did, you would think the prediction accuracy would strongly track the number of men voting (since they should predict the same way they voted in large numbers). But it doesn’t. Since I can’t break down prediction accuracy by gender, though, I can’t be sure one way or the other anyway. My theory lives to fight another day.
And that’s all I have to say about “Everybody Votes”. Here are some conclusions:
Part 1:
- The closer the vote is, the worse people predict
- People somewhat predict that others vote like they do
- A small majority of people predict correctly most of the time, but are occasionally very wrong
Part 2:
- The ratio of male to female voters is about 2.5 to 1
- Women consider both poll choices more and vote “less extremely” (truthfully?) than men, but not by that much
- Men and women’s voting differs by about 10% most of the time, but occasionally they vote very differently from each other
Thanks again for your data, Tom. Hope you all enjoyed! (And thanks for the positive coverage from Wii Live.)
(As promised, here’s the workbook I used to prepare the post. You can see as you scroll across the spreadsheets how I transformed Tom’s blog posts into data and charts. You can also see how disorganized and haphazard the workbook is compared to this post. Just like a duck - calm and smooth on the surface, paddling like hell underneath, heh heh.)
| | | del.icio.us |
December 15th, 2007 at 10:57 pm
This was quite an interesting analysis of the everybody votes channel. This was quite an interesting two part read, thanks!
Tristan, I appreciate the positive feedback as always. Glad you enjoyed!
- Dave
December 21st, 2007 at 3:31 pm
Kristina and I would vote on almost every topic if my friend Stephen ever finishes playing through Metroid Prime 3 and gives me my Wii back. My own voting history is also slightly more sporadic since my Wii resides in a different city than I do during the week. I’m going to guess that I’m an extreme case here, though.