A Calamity, A Catastrophe, An Apocalypse!

Patrick Cahalan in reply to Chris says:

In fact, the data doesn’t tell us much of anything.

Well, it tells us that there’s a difference between how liberals (adopt and use social media) vs. how conservatives (adopt and use social media).

But it doesn’t tell us anything about how liberals adopt social media vs. conservatives adopt social media, or how liberals might use social media vs. how conservatives might use social media.

First you’d need to find out why there’s such a huge difference between the adoption rate in social media. That, in and of itself, will probably point you in a direction as to why the usage is different.Report

Patrick Cahalan in reply to wardsmith says:

Ward, you’re doing that thing you do where you start talking to me like I’m an idiot and it’s going to make me angry again.

I will respond to this in length, right now, and then I’m probably not going to bother to engage you on any social science topic again, because it seems likely that – while you do understand statistics – you don’t know the first thing about *social* science methodology (you’ve displayed “more than healthy skepticism” in the past, and this comment reflects that. And your apparent inability to understand a pretty basic concept in social science makes you difficult to engage with constructively when you hop on your “I understand statistics” high horse and start talking down to me from up there.

I admit, this might be me projecting, okay? That might not be what you’re trying to do… but it sure comes across that way. Allow me to explain:

> I NEVER said that the report was representative
> of society at large, nor did Pew.

Tom introduced this, and without drawing any conclusions out of the material other than what was quoted at The Daily Caller. I went and read the study, and explained my problem with it. You then interjected yourself into the comment thread here, saying that you thought I’d missed something based upon the comments so far, and then wrote a comment that actually wasn’t anything about the comments I’d written so far. Which indicated to me that you had missed my point.

I well understand that the denominator used to establish the relative percentages of liberals who defriend people vs. conservatives. The problem is that the denominator is not independent.

> Your preselection THEORY is just that unsupported
> by anything other than your imagination.

No, Ward, my “preselection theory” is based upon a 27% difference in adoption rate. That is, in fact, pretty crazy-solid there. There is a huge statistical difference between “people who self-identify as liberals who use social media” and “people who self-identify as conservatives who use social media”, by these numbers.

I don’t need to go any farther to defend a preselection theory: it is evident that there are behaviors (as yet not described) that lead liberals to adopt social media at a much higher rate than conservatives. No imagination needed.

Now, I speculate on what might be the effect of this preselection bias, but it’s not arguable that the bias is there, Ward. It’s there, sure as the gravity of a black hole we can’t see.

> Yes you try to can excuse it away, because it
> pinches a bit by pointing to number differences
> but the implication is still there, and would
> be with larger samples.

I think you’re misunderstanding an attempt to describe a phenomena with a desire to correct it to an unheld bias. See, Ward, this study doesn’t tell me anything pejorative about liberals -> it apparently tells you something pejorative about liberals because you have stated (elsewhere) that you believe that liberals are more intolerant than conservatives, and this can be interpreted as a piece of evidence to confirm that belief. But, see, I don’t have any idea who is more intolerant, this isn’t my shibboleth.

For all I know, liberals are way more intolerant than conservatives.

(I actually expect they are equally intolerant about wildly divergent things, because they hold different principles as root principles, but that’s an aside.)

> People use SNS to socialize with other people for
> billions of reasons, politics being a small subset.

Aside from a likely exaggeration, no problem with this statement.

> I can and have been friends with people for
> thousands of reasons, not one of which involved
> politics. When politics interfere with otherwise
> convivial relationship I’m not at all surprised
> unfriending results.

Neither am I, particularly when the social media bond that originated the relationship has nothing to do with politics (this is an important point, down below). People react badly to surprise; finding out that someone that you thought of as a great fellow player of FarmVille is a closet anti-Semite or a tree hugger or any one of a number of things causes you to have to reevaluate your worldview. People don’t like to do that.

> My personal experience has been that liberals
> are intolerant of dissent in their own midst
> (viz any liberal who dares to stand up to the
> party line – Ethan’s recent OP a case in point)
> and in society at large.

My personal experience has been that liberals are much more concerned with group dynamic than conservatives are. You know, that whole “communitarian” vs. “individual responsibility” thing. So I don’t find this to be offhand an unlikely claim. However, it’s a claim, and your personal experience isn’t terribly strong evidence. Neither is this study. It’s certainly suggestive (I never claimed otherwise, you’ll note, from my first comment!) of deeper investigation. The theory (liberals are more intolerant of dissent in their own midst) is based upon at-least-first-glance reasonable assumptions – for certain types of dissent – sure.

> I’ve had liberals literally tell me not to
> think what I am thinking. That is a laugh. I’ve
> never met a liberal smarter than me, not
> measured by IQ, not by education and not by success.

Ward, statistically speaking this is a phenomenal claim. It’s certainly possible that you have a genius level IQ, and that you’ve been at school longer than every other liberal you’ve ever met, and you’re more successful than all of the ones you’ve met, but it’s pretty unlikely. In my case, it’s easy: I’m a pretty smart dude and within football-throwing distance of my office I’ve got people who are smarter than me by all three of those measures who are liberals, conservatives, atheists, deists, libertarians… you name it. I never have to worry about being the smartest guy in the room, I lost that when I started working here (probably earlier than that, to be sure).

I will say, however, that if you were an incoming graduate student in a social science and you said that to your advisor? They’d know that they were going to have to check your work a lot for bias. You’re coming across as powerfully impressed with yourself, here, Ward. Maybe you have a right to be, I dunno. But you’re going to be a shitty social scientist if you can’t approach a subject with a lot of humility, and you’re exhibiting exactly zero humility in this paragraph.

> I find this amusing Pat because back when you
> and I were talking about selection bias in the
> way temperature statistics were measured and
> gathered (and I showed you that literally
> thousands of reporting stations were “off the
> air”), you defended the practice of
> “interpolation” backed by statistics to make
> up the difference. You can’t have it both ways.

Ward, if you’re using this to defend extrapolating this study to “patterns of liberal behavior, generally”, I’m really hard pressed to interpret this paragraph any other way than, “You don’t understand social science.” This paragraph illustrates a serious misunderstanding, in shattering detail. (Also: you realize you just conceded that this can be a legitimate error correction method in some situations, yes? This sort of weakens your objection from the AGW thread, but that’s a digression)

If you’re not using this to describe “patterns of liberal behavior, generally”, then you’re misreading me and we’re talking past each other.

A physical instrument (barring instrumentation problems – yes, we can go down that thread some other time when we’re actually talking about AGW) basically measures, for the purposes of temperature, independent phenomena. That is to say, if I’m plonking down 1000 thermometers in 1000 places and taking temperature readings, I’m generating a map of local temperatures. If some reasonably small finite N of those thermometers goes out, and they’re in certain types of locations, I can still get a usable (although less useful and dependable) measurement by extrapolating the temperature in that location, by two estimations: the temperature in nearby stations, and the pattern of temperature variance between that location with the broken thermometer and a similar pattern elsewhere in the grid of instrumentation. So if I lose one temperature station in a desert, and I have six nearby in a geographic pattern that resembles a geographic pattern elsewhere (plus historical data), I can use all of these things together to guess with a reasonable degree of certainty what the missing station says. “Oh, station 6 went down, but for the last 12 years station 6 has been within a half degree of station 20. We figured out a while ago that’s probably because they are both desert stations within a mile of a mountain and downwind of a large body of water, so we have a geographical hypothesis that renders it likely that station 6 would be within a half degree of station 20 today, so we can write that in (well, in red, because we want to be sure we track the real data from the estimated data).”

The problem is that the differentiation here is that we have a large body of measurements, but they are limited to a set of closely related phenomena. We’re measuring temperature, on the planet Earth, with (hopefully) reasonable historical reliability. If a temperature station goes down, we can compare its historical record with the neighboring stations, and if they’re close enough, and follow a sustainable pattern, it’s not completely bonkers to fill in the blank. If we have an area where we’ve never had a station, but it is nearby other stations and it has geographical and atmospheric similarities to areas we do track, we can guess what the temp might be. It’s a far, far less reliable measure than having a temperature station there, but it’s not anywhere near the same as making things up. Micro-patterns do exist in climate, but the odds that one makes a long term difference in a baseline measurement like “global temperature average” is pretty small.

Now, on the other hand – at the risk of sounding like I’m talking down, here – when you’re measuring people, you’re not measuring the frickin’ weather. As enormously complicated as the weather system is, in short term measurements it’s a hell of a lot more predictable on a small set of known variables than people are. Controlling for dependent variables is a lot harder when you’re measuring people. If you have a sample that’s all men, and you’re measuring some behavior, you better be goddamn sure that whatever you’re measuring is either dependent upon the fact that they’re men, or completely independent from sex-correlation or sex-linkage.

If you’re studying a subpopulation of a larger set, you have two ironclad rules: you draw no conclusions about that larger set without some awesome supporting evidence, or you extrapolate to the larger set only if you have some awesome theory that links the subpopulation to the greater whole.

This study is indeed measuring “liberals who use social networking” vs. “conservatives who use social networking”. But the adoption rate between the liberals and the conservatives is so large that you have to understand that effectively we have to assume that we’re not measuring closely related phenomena. There are fundamental differences between “liberals” and “conservatives” that affect how they adopt social media, and since social media is itself a social phenomena, those fundamental differences are going to have a large impact on any pattern of social behavior inside social media, when comparing those two groups.

This isn’t “we’re measuring the temperature of deserts and temperate rainforests in Africa and Asia and we’re using the results to guess as to what the temperature differences would be between deserts and temperate rainforests in South America and North America”. This is, “we’re measuring temperate rainforests on Earth and deserts on Mars”. There’s a category error, there. You see why you can’t use this to generalize?

This paragraph is worrisome:

> Chris, it isn’t just FB, but all SNS. However
> they weren’t including such sites as THIS ONE.
> If they were they might have gotten further
> information but of course the option of
> “defriending” or anything of the sort is
> unavailable. Therefore the only thing left
> is Social Networking.

You’re jumping through a hoop and assuming something very large, there.

> Personally I could contend (and a study
> could well prove me right) that more liberals
> use SNS than conservatives because more
> liberals are un and under employed, therefore
> have more free time. Something tells me you
> wouldn’t like to know about that study either. 😉

There’s a whole lot of assumptions in that paragraph that look extremely dodgy to me, Ward. I’d love to see how you would construct such a study. The act of building this study itself would probably reveal quite a bit about your thought process.

> Even more obviously, people who self-select
> as liberals tend to skew younger and
> younger people tend to be more involved in
> SNS. Likewise people who tend to be
> conservative tend to be older (hence the
> old saw: If you’re not liberal when you’re
> young you have no heart, if you’re not
> conservative when you’re old you have no
> brain). Age alone can explain these
> numerical differences to a large extent.

Age alone can also explain a difference in patterns of behavior that have nothing to do with liberalness at all. Here’s let’s use this bit, which I won’t contest, and you offered.

Theory: older people are more likely to adopt social media as a way of connecting with existing friends and family (Grandma only joins facebook to see the grandkid pictures that her daughter puts up there). Thus, their “friends” list is much more likely to consist largely or exclusively of people they know outside of social networking, with whom they are both familiar and bonded. Younger people are more likely to adopt social media as a way of gaining new friends (looking for like interests, gaming buddies, girls/boys). Thus, their “friends” list is much more likely to consist largely of people with whom they have a much more tenuous connection (a single connection such as like interest in a hobby, school membership, etc.). The younger folk are thus more likely to add new friends, and dissolve existing friends, in general.

Hey, Pew could have tested this theory, provisionally, by asking four additional questions in this survey: how old are you, why did you join social media, how many friends do you have on your friends list that are only online friends, and how many friends do you have on your friends list that are family or old offline friends.

> Finally and I’m finished with this discussion
> here, I have seen liberals say the vilest
> things to me and to conservatives hands down.
> Perhaps some conservatives have done the
> same but that hasn’t been my personal nor
> reading experience. Moral superiority isn’t
> a given, it must be earned, and earned
> constantly. Staking the moral high ground
> and then using it to shit on one’s
> opponents is beneath contempt. I will
> criticize that behavior every time I
> witness it.

When people around here blog about vile things that some supposed conservative said somewhere, and you jump in the thread, how often is it – in your own estimation – to criticize that behavior, and how often is it to point at some supposed liberal somewhere that did the same thing, or something equally atrocious?

Suggested further reading, if you’re actually interested in understanding social science, particularly how to design, implement, and interpret a survey instrument:

The Craft of Research, Booth, Colomb, Williams, ISBN 0-226-06567-7
Social Research Methods, Neuman, ISBN 0-205-45793-2
A Primer In Theory Construction, Reynolds, ISBN 0-205-50128-1Report