30

I found this pdf referencing a study done on video gamers. Among other claims it states that "80% of those surveyed will vote in the 2016 election". It also has a pie chart claiming the following distribution of gamers by political party:

48% conservative
38% liberal
14% other

I could not find the original study. I am somewhat skeptical of the claims, the skewing for conservative is higher then I would have guessed, especially considering gamers tend, on average, to be (somewhat) younger which would usually skew towards liberal. I also find the claim of 80% participation to be surprisingly high for any American demographic that doesn't have an obvious political affiliation.

Can anyone confirm the original study? Is the study a proper peer reviewed and controlled study who's results can be trusted?

dsollen
  • 10,062
  • 12
  • 43
  • 71
  • Comments are not for extended discussion; this conversation has been [moved to chat](https://chat.stackexchange.com/rooms/85106/discussion-on-question-by-dsollen-do-video-gamers-tend-to-be-conservative-and-ha). – Sklivvz Oct 30 '18 at 21:14

2 Answers2

47

The small print on the pdf declares that it was from an online survey targeting american adults who play video games. Online surveys have pretty bad self-selection issues.

ESA looks like it's not particularly entangled with any one political side, but it is still (among other things) a lobbying group. Their own website (both hosting and linked to by the PDF) has a section Public Policy/Government Affairs with the following quote:

ESA engages policymakers at the national, state and local level on a range of legislative and public policy issues impacting the video game industry, including copyright and intellectual property laws, First Amendment protections for industry artists, privacy and trade.

In its status as a lobbying group, any evidence that suggests that gamers are a significant voting block is inherently beneficial to it.

ESA seems to be pretty serious/establishment, as such things go. (They run E3.) As such, it is probable that they are not blatantly lying, as they actually do have some credibility that they would probably prefer to retain. On the other hand, the actual claims made on the survey pdf are limited in scope. All it says is that the survey occurred, that it was a web survey, that it got certain results, and that it "provides insights".

It is likely, then, that the survey was at least somewhat biased by self-selection (and possibly through other means) While ESA appears to be at least somewhat credible, they don't claim that the survey is particularly accurate or unbiased.

Ben Barden
  • 3,657
  • 2
  • 15
  • 20
  • 30
    It might be worth pointing out that the survey took place in 2015, which was right after the gamergate harassment campaign (which in turn was tied up in the rise of the so called alt right / white supremacism). This energized people (on both sides) and might have increased the percentage of those who claimed that they are going to vote in the next election. I would assume that the numbers would have looked quite different before 2014. – tim Oct 26 '18 at 18:20
  • 2
    There is a lot of supposition in this answer about how you think it might be, and not a lot of empirical evidence that that is the way it is. – Oddthinking Oct 26 '18 at 18:41
  • 1
    @Oddthinking I've tried to tweak in a bit more. Unfortunately, I simply don't have a huge amount of empirical evidence. I admit that empirical evidence would be superior, but in its absence, I think the implications are still usefully informative. – Ben Barden Oct 26 '18 at 19:11
  • 1
    @tim that's an excellent point... but I don't know how to properly integrate it without piling on yet more supposition. – Ben Barden Oct 26 '18 at 19:12
  • 1
    You could state what the self-selection issues are and why they matter. For example, if men respond more often to online surveys – Azor Ahai -him- Oct 26 '18 at 19:15
  • 1
    @BenBarden Good point. I think providing some context is always great (even here at skeptics), but it shouldn't outweigh sourced content. Previous studies (from before 2014) would be great for comparison and to source (or disprove) my comment, but I wasn't able to find any yet. – tim Oct 26 '18 at 19:31
  • @AzorAhai I don't have that information. All I really have is "we know it's likely biased by self-selection, and probably doesn't reflect reality all that well". I could make guesses, but I have nothing that's at all solid to back them up. – Ben Barden Oct 26 '18 at 20:09
  • 3
    I don't know that "self-selection" is really the main problem, but online surveys most certainly do suffer from (frequently severe) selection bias. Namely, the only participants are the people who visit the website hosting the survey, which is very rarely an accurate sampling of any particular demographic other than the people who happen to visit that particular website. Even if they conducted an accurately-sampled poll of the people who visit that particular website, it would still only represent the views of the people who visit that particular website. – reirab Oct 27 '18 at 19:42
  • This is the same reason that a poll on msnbc.com would be skewed heavily toward liberal views, while one conducted on foxnews.com would be skewed heavily toward conservative views. – reirab Oct 27 '18 at 19:43
  • @reirab that bias is absolutely there, and that's a good point, but there's *also* a pretty strong self-selection bias (at least potentially) in that the surveys are only filled out by people who bother to fill out surveys. – Ben Barden Oct 29 '18 at 13:05
  • @BenBarden Right, that's why I said that "I don't know that it's the main problem" rather than saying "It's not a problem at all." My suspicion is that the selection bias of who visits the site in the first place outweighs the self-selection bias of who fills it out, but I could be wrong. It would be interesting to see studies on the matter. – reirab Oct 29 '18 at 13:08
20

Here is a press release that gives slightly more information about the survey.

The first-of-its-kind survey explored the political attitudes of video game players, providing a snapshot of their political leanings and positions on key issues. It finds that more than 80 percent of gamers said they anticipate voting in 2016, compared to 75 percent of non-gamers.

In other news, more than 80% of people surveyed on New Year's Eve said that THIS will be the year they finally stop eating junk food and get serious about exercise ;-)

One of the biggest challenges in political polling is that people's statements about their intentions are not very reliable, and people tend to overestimate their intention to vote. The press release states that 80% of gamers and 75% of non-gamers intended to vote in 2016 - but in fact, only about 58% of eligible voters did vote that year, so clearly Ipsos' methods are not reliable for predicting actual turnout.

This can happen even in a scientifically-conducted random sample survey with high response rates, because people are generally bad about honestly predicting their future behaviour. But in this case, the release states that the study was conducted through an online questionnaire.

This adds another major source of error: respondents to these things are generally not representative of the overall population. People who have strong opinions are more likely to participate, which would be likely to exacerbate overestimation of propensity to vote.

There are methods that can be used to mitigate this problem - e.g. you can compare the demographics of your respondents to the general population, and weight under-represented demographics - but they're hard to get right, and it's not clear from the release whether Ipsos even attempted them.

Note that the PDF has distorted the findings that were reported in the press release. It misrepresents "said they anticipate voting" as "will vote", and it removes the comparison stat for non-gamers that might have made it more obvious that this stat was a gross over-estimate. Without that information, a casual reader is left comparing "80% of gamers will vote" with "58% of eligible Americans did vote", which greatly exaggerates the gap.

Also, what is a "gamer"? The scope of the Ipsos survey is: "reported playing video games at least three to four hours per week." That's a very arbitrary cutoff, and also not clearly defined. (Over what time period are we measuring that average?) You could probably get different results by applying a different cutoff, or by being more specific in how that rate is to be measured.

GB supports the mod strike
  • 1,214
  • 1
  • 9
  • 11
  • 1
    "What is a gamer?" Exactly. I play one or two different video games on my phone every day but have never played a game on my laptop or desktop computer (not even minesweeper or solitaire.) Am I a gamer? – CramerTV Oct 26 '18 at 23:27