Revenge of the Petunias

Dave Freer has kindly replied to On Petunias and Whales: Part 9 in comments on the Mad Genius Club blog. This post is a reply. The format of my post is a bit ‘fisk’ like and I’m not keen on that because: The fisk approach I think always ends up being a bit aggressive When I see “fisk” now I can’t help but think of Vincent Donoforio’s portrayal of Wilson Fisk in the Netflix TV series of Daredevil. “To fisk” somebody definitely sounds aggressive. Having said that Dave has offered lots of points and I’d like to reply to them in turn and so the quote-reply style makes sense. In terms of quoting Dave’s comments, I am editing them to focus on particular points. This can lead to comments being distorted from the intended meaning and so I would recommend people read Dave’s reply in full in its original context. Dave’s comments will be bold and italic and be preceded by “DF:”.

DF: I’ll repost this on your site, but your principle flaw is that you skimmed until offended, and then did a magnificent job of deconstructing what I did not say. In other words you totally destroyed straw-Dave. Unfortunately straw-Dave and this Dave are not even vaguely same person. You started with an incorrect premise, which meant the edifice you constructed on it could not stand.

I did not find anything offensive in your original post. I actually like it and I think it was both brave and correct to examine the issues in this way. You pulled a dispute that was phrased in the language of US culture wars into a discussion that is phrased in the language of numbers and statistics. Consequently we can trade numbers rather than verbal barbs. Additionally I thought there was a socratic quality to the discussion that I kind of liked. I was wondering what places this would take conservative minded people if they followed the implications – particularly on some of the issues around gender and diversity that have generated so much debate. As for straw-Dave. I agree. I am arguing against an imaginary Dave. I don’t know you and I don’t know the contents of your head. I also have no particular interest in establishing the proposition “Dave Freer is wrong about something”.  What I was interested in was your argument. I even toyed with an early draft of Part 1 in which I avoided your name altogether to ensure it was a depersonalized argument. Unfortunately that made it both unreadable, even more straw-mannish and looked like I was avoiding crediting your own work to you. So yes Straw-Dave is who I was addressing because it was the argument, as I understood it and in terms of how people were using it that I wished to tackle. Of course the argument I understood from your text and the argument you intended to write may not be exactly the same because the written word is a noisy channel. For example if you only intended to demonstrate that some sort of statistical bias is present in the Hugos and that the source of that bias could be due to factors other than the prejudice of inner cliques or WorldCon voters overtly avoiding conservatives, then we disagree on very little. However the extent to which your argument therefore provides support for the two Puppy movements becomes more debatable. I’ll bring in two Straw-Daves: Strong Straw-Dave: is basically what I was arguing against. Strong Straw-Dave is arguing that the recent Hugo results provide a clear justification for the Puppy consternation against the Hugos. Weak Straw-Dave: is only saying that conservatives are doing less well than a person might expect  in the Hugos but the causes of that and hence the appropriate reaction are unclear. Real Dave maybe believes something else altogether.

DF: I repeat my suggestion of reading Prof Jonathan Haidt of NYU’s “The Righteous Mind” (Haidt is a psychologist, and a US Liberal) It’s an interesting read, but the nub you need is that US conservatives actually understand and know US Liberals quite well, but the converse is not at all true. It’s basically the Pauline Kael error (or as you have just displayed in your UK elections – the ‘nasty Conservative’ error). You are not actually the center of the demographic distribution, you just think you are, and this is a false perception heightened by a compliant media.

I haven’t read the book but I am familiar with the idea and it was an issue that I considered during my analysis. A few things first though. I am definitely on the left and far enough on the left that I am under no illusions that I am at the centre of the demographic distribution. So I get what you are talking about because I also experience that issue of people in the middle understanding my politics less well than I do theirs. Additionally I believe it is also true that ideas from the left of centre are currently more available in a cultural sense than ideas from the right – so there is an imbalance in cultural discourse between left and right currently. I have my own views as to why that is the case but the margins of this book are too small to contain them. Thirdly while I am of the left, I’ve taken an active interest in American conservatism for several years now – again for reasons that are too long and dull to explain. So while I’m not an expert I am not as ignorant of the diversity of opinions in American conservative politics as my political leanings might indicate. However, to the nub of the issue. While it may well be true that US conservatives are more aware of the views of US liberals than vice-versa this does not imply that either group are particularly good at classifying people on some ideological spectrum. Indeed I would say that is a fundamentally had task and secondly it is shaped by ideology (i.e. the markers and proxies we use or regard as telling are shaped by our own ideological priorities and prejudices). Consequently I wanted to strongly minimize the extent to which I would make that error. I can’t completely – I’m looking at the world through pinko-lefty spectacles – but I want to make my assumptions check-able or independent. I did this two ways: Use your classifications of ‘red’, ‘white’ and ‘black’ winners when possible (with some caveats) Use a political typology based on statistical research. In this case the Pew Research Center’s typology based on survey’s of Americans on a range of issues.

DF: I’ve pulled together various figures here –mostly from Gallup, or Yougov rather than overtly partisan groups. Remember the ‘nasty conservative’ error – which means pollster tend to underestimate support for conservatives and conservative issues, because the polled are defensive about these issues.

This is a sound thing to do but it is important that we always keep in mind what we are counting and what those categories mean. “Conservative” as a category is not like “redhead” – the functional definition used in a given data source can lead to very different estimates. The range is sufficiently large as to result in quite different conclusions about the Hugos. Worse, there is an obvious error that can be made if there is a mismatch in the nature of the categories we use. If we use a narrow category to estimate the general population of conservatives and then use a generous category to classify the Hugo winners we will magically find the the Hugos are biased against the left and for the right 🙂 A Trotskyist I knew once was prone to this kind of error. They were convinced that most ‘working people’ were inherently socialist and hence a general-strike/revolution was just one economic crisis  away. On the other hand if you asked them in another context who was a socialist they rapidly withdrew to a no-true-scotsman style of reasoning (The USSR, China etc were “state capitalists”, liberals were just apologist for capitalism, social-democrats were just trying to make capitalism palatable etc etc) – in this double-think socialist were both a tiny set of people who subscribed to the newspaper he sold and at the same time the vast majority of working people.

DF: I think our starting point has to be false perception that ‘conservatives’ in the US are some tiny minority – perhaps 20%.

Whoah! Hold on. My argument was very much not based on conservatives AS A BROAD CATEGORY being a tiny minority. The terms I was using (e.g. ‘solid liberal’ and ‘steadfast conservative’) were taken from a statistically based typology. My argument was based on a specific sub-group of conservatives being a minority

DF: Actually the truth is almost the inverse. In 2011 40% of Americans considered themselves conservative. Just 21% considered themselves liberal. 35% considered themselves moderate, and the trend numbers of moderates is declining.

Yes. Back to an earlier point. What kind of data is this? Well it is data about how people self-identify with these labels. That is interesting data but precisely because of the shy-conservative issue you mentioned earlier it isn’t very useful for the Hugos. What we would need to use is the number of Hugo nominees who, if asked in that style of poll, would identify themselves as ‘conservative’. Now that is a tricky category, particularly when you consider examples like John Scalzi So the self-identification labels are not going to be a fruitful way of looking at the questions for several reasons. Consider the issue of what me call perceived-bias or something akin to stereotype threat. The puppy campaign itself promotes the idea that openly conservative authors may face prejudice or bias. Even if that is not true, an author who would in a confidential survey say they identify with the label “conservative” may be even more shy in the context of a public presence.

DF: Let me be clear here – I do not mean an individual who is openly and loudly supportive of one or two of the criteria I set out and as loudly opposed to the rest, but who is at least broadly supportive of all. Given your contention ‘more popular support in the US public than some conservatives realize’ – Gay marriage for example, would have passed electoral muster in the some of the various states it’s been put to the vote. It has repeatedly failed to do so, even in heavily ‘liberal’ states like California.

I think that is how I understood what you meant – a person who would broadly be supportive of each of the issues you listed. I worked with that and tried to match your issues with the typology data that was available. As for votes – votes aren’t statistically sound surveys of opinion. Other issues are at play, include voter turn out and expectations of winning or losing.

DF: You postulate that the positions I describe as defining the ‘liberal’ side enjoy far greater support among conservatives than I guessed: let’s actually look at the figures. Gay marriage: In the year I wrote this it was 30% among Republicans (71% of whom consider themselves ‘conservative’ – which makes it around 21% of conservatives support it ) Support for abortion among Republicans was 27% in the year I wrote the piece (again 71% of that =19%). Affirmative action – measured as support by those NOT receiving it for college entry, by race, was 22% (take 71% of that 16%). While some more support for some forms of affirmative action, even that peaks at 40%. (which includes the beneficiaries, which means the ‘support’ is lower still.)

I think that is more-or-less in keeping with the point I was making. I’d recommend looking at the typology data as well. A pro-abortion, pro-same-sex marriage, pro-affirmative action SF author could be a conservative. It isn’t likely but it isn’t so unlikely as to be absurd. Importantly on some of these major social issues, you will find some conservatives on the “left” side. Outside of the US this is even more the case.

DF: Militant Feminism – well as only 20% of the US population (23% women) identifies feminist, let alone militant, once again your imagination doesn’t live up to the facts. Among Republicans that sinks to 5%. Take 71% of that, and you’re down to 3.5%… and militant feminism is a fraction of that.

Again you are using labels that people identify with versus issues they take positions on. The shy-tory effect isn’t just for tories. Terms like “liberal” and “feminist” have some pejorative associations in the US.

DF: Socialism barely scrapes 20% in the total US population, and in conservative circles that would be almost zero (in other words, if you assume some moderates support it, it doesn’t even get all the liberal support.

Sure. I wouldn’t think there any conservatives who would fit the category “socialists” – but then how many of the US authors in your classification would identify as socialist? I can think of some prominent SF Hugo award winners who would have happily said “I’m a socialist” – China Meivielle and Iain Banks spring to mind – but American ones? That is a tougher call. if you have a count of Hugo winners who SELF IDENTIFY as socialists (i.e. they overtly call themselves socialists) that would be very interesting data.

DF: I think you need to re-think your ‘more popular support than conservatives think’. You have proved Prof Haidt absolutely correct.

I’m always happy to rethink it but as thinks stand the data you have shown is broadly consistent with the data I used. The differences come down to differences in what was being counted. More importantly I took significant steps to reduce the influence of the effect Prof. Haidt refers to.

DF: As your initial premise (above) fails spectacularly, because of your misapprehension of the size of ‘solid liberal’ base – I suspect what you mean by ‘solid liberal’ is what Gallup defines as ‘very liberal’ which is actually 6%, as opposed to 10% ‘very conservative’. Which means I erred on the side of caution, being nearly twice as generous to ‘very liberals’ than ‘very conservatives’. (I gave them equal weighting, and in fact the ‘very liberals’ should only have 60% of the ‘very conservatives’). If you mean ‘solid liberal’ to include ‘very liberal’ and ‘liberal’ then you must do the same for opposite side – which would leave your ‘solidly liberal’ at 21% and ‘solidly conservative’ at 40% – which is considerably worse for your argument. Your estimate has no substance or backing I’m afraid.

No guessing is required to determine what I meant by ‘solid liberal’. It wasn’t my term or my figure and the Petunia posts explain the source of the figure, the description of the group and how that group was identified. I used data based on correspondence on a series of issues and I did that because of how you described the two groups. The same goes for the ‘steadfast conservative’ group. Rather than going on labels people attach to themselves, I referred to a typology based on issues. The notion of position on a spectrum relating to a degree of correspondence of belief on some key issues is one you introduced in your argument. I concur with that aspect of your approach and attempted to use it to estimate the proportions. What I found is on several of those issues what you might perceive as being very leftwing was actually not that leftwing. Additionally on issues surrounding identity politics I believe there was a fair amount of evidence that views perceived as being quite radical by some conservatives have penetrated significantly to the right of the political spectrum. Consequently in a environment were recent political controversy and debate had been predominately focused on issues such as gender, sexuality and ethnicity, a conservative may perceive the environment as being more left-wing than it actually is.

DF: However – I will quote the document you’re criticizing – “Guys, you can argue about the figure, but essentially you’re splitting hairs. Take it down 0.00005, or up 0.4 (beyond that allows no undecided votes) – the outcome is the same.” And I included your 50% calculation. “To forestall the ‘oh but 1/7 is too low’, let’s run it at a ridiculously high ½, which means only red and black balls. No neutral, but they are equally probable. You can’t get more generous and possibly claim that there is no ideological bias in Hugo awards. ½ x ½ x ½ x ½ x ½ . – that is still unlikely to happen by chance more than once in every 32 years. There is a 97% chance that won’t happen. Do more complex stats, it’s just as implausible.”

I didn’t think this particular argument made sense. It only makes sense if we adopt a quite different notion of conservative which would be anybody to the right of the median/mean position on a symmetric distribution. That isn’t a crazy thing to do but it changes all the data we are looking at. It changes who is white, red and black in your analysis and would even change how you would have to evaluate each winner. So in that particular scenario you model with no white-balls and an even split of red and black balls it would be surprising to find no black-balls but you haven’t done a count of red-and-black on that model. It may well work out 50-50 and actually it would be very hard to tell because the figures would be highly sensitive to how we classified people near the political centre. Either way that analysis does not address the point I was looking at. Identifying people by some core issues the percentage for the “red” could be quite high, high enough to make the proportion of reds you identified to within the bounds of chance. You may say that if the size of the red proportion is large then the size of the black proportion should also be large. However you have already identified that people from one position on the spectrum are less skilled at evaluating the political positions of people on another part of the spectrum. It stands to reason that your ability to classify conservatives (based on the correspondence of positions on a set of issues) is likely to be less error prone than your red classification.

DF: Thank you. I was doing my best to keep it simple. Please read my comment in the document you’re criticizing – “Nor do I think no outspoken left-winger should ever win. If that was happening, I’d be here fighting to see they got a chance.” I have a track record of this.

Not a problem. I only discussed these issues as I thought others might and say that your overall argument was wrong because of x, y or z when these were either minor nitpicks or necessary simplifications of a complex process.

DF: Er. No. Sorry –see my quote above (in answer to 2) running the numbers at 0.5. Besides the facts, established my rebuttal to your (1) that a realistic figure for ‘red balls’ would be 6% IF the figure for ‘black balls’ is 10% – and it still does not matter. You are confusing ‘possible’ with ‘probable’. It is not impossible. It is highly improbable.

My reasoning is spelt out in Part 5 and my reasoning for a higher estimate for “reds” is spelt out in Part 3 and Part 4. The probabilities for testing the hypothesis are also outlined in Part 6. I haven’t checked my working so there may well be errors.

DF: My choice of years was actually influenced by working backwards and running out of time. However I was busy with a larger analysis which showed the Awards gradually become more and more and more dominated by one political wing, year on year. I had a rather nasty computer crash, and will have to do it all over again. I will, but it is slow.

Sounds interesting. I think what you need for a tighter case is present which authors you are classifying as “red” and on what basis. For example if each of the red authors have overtly self identified as a socialist then you have a slam-dunk case. However, if each of the authors are people that a steadfast conservative would call a “socialist” then your case is not so hot.

DF: Mike Glyer – using less rigid metrics than I did (so probably counting people as ‘black ball’ I would consider ‘white ball’ – he described his method as ‘subjective’ – but he does have an exhaustive knowledge of the people in the field) found 19 conservatives winning Hugos (not Novel Category, ALL categories) in the last 20 years. I think you would find most of those 19 are pre-2004. Mike did not do a liberal winners number – but if it is to match the demographic it should be not more than 60% of that 19 – 11, with all the rest being ‘moderate/neutral. This is clearly not the case. I can arrive at 11 in three years, not 20, and that is being extremely stringent about ‘red ball’.

I posted a message to Mike Glyer at 770 about his figure to get a better sense of which categories he was referring to (e.g. did he include Dramatic presentation? Just the awards that have a single named person as a winner? Did he include artists? Editors? I’m not sure). Without knowing the breadth of awards he was thinking of, it is hard to evaluate. I have posted a short post on this figure earlier when I didn’t know the source. So, yes – Mike Glyer is a pretty good source for Hugo history but no, that figure by itself doesn’t help us much.

DF: I don’t know if you’ve worked this out, but I assume you would accept that there is bias in the number of women winning Hugos? Well – over the same period I make that as rough count of 57. So given that women make 50% of the population, and Conservatives 40% because Mike’s definition of conservative is more generous than mine – you then establish that bias against conservatives is actually 2.7 times as bad. Am I suggesting we ignore women? No. But we cannot ignore something worse, and notice women’s participation.

My reply to this bit got a bit long. I’ll write a separate reply.

DF: Perhaps you’d like to reconsider that? The evidence is overwhelming.

OK I have reconsidered and at the moment I still say the evidence is slim.

DF: This is straw Dave I’m afraid. I quote: “So let us take the hypothesis ‘There is no ideological bias in Hugo awards,’ which we’d love to prove true and test it mathematically to do so.” I wasn’t trying to establish what the sources of the bias were. I was simply establishing there was bias. I have actually written about the sources in several other places. But the point of the exercise was utterly dismiss the ridiculous contention that the outcome reflects a null-bias situation.

It may be straw-Dave but if your conclusion is only that there is a statistical bias against conservatives in the Hugo awards then you don’t have a very good argument in favor of Sad Puppies. For example you start your essay with this:

DF: Or why, if the people ‘really the only reason ‘Ancillary Justice’ won a Hugo Award’ was that it was much better than ‘Warbound’, that there is no ideological bias at all in the entire Hugo Award process, and that, indeed, they personally didn’t know what so-and-so politics were, are right, then Ann Leckie and her fellow winners should buy a lottery ticket and become a multi-millionaires overnight.

However a finding of some general statistical bias in the Hugo winners does not contradict the proposition that Ancillary Justice was much better than Warbound nor does it contradict that people didn’t know (or perhaps care) what so-and-so politics were. In your introductory paragraph you seem to be using “ideological bias” in a more narrow sense. In reality a statistical bias in the Hugo results could have many causes – including that for various reasons that conservatives simply are not producing as much high quality genre fiction. Put another way the bias can lie in the pool that is being drawn from or from how people are drawing from that pool. Even if we find bias it is not possible to distinguish the two. If the source of the bias is essentially a social/cultural one that means SF/F authors in general are unlike the general population politically (and there are reasons to think this may be the case) then the source of the bias is there being a shortage of conservative writers. Well I suppose Sad Puppies might help promote conservative writing but in that case SP would be an example of affirmative action – and that is something that I suspect many SP supporters have an ethical and ideological objection to. You have stated that you don’t support affirmative action for conservatives. So, I’ve considered straw-Dave, because straw-Dave is making a case for Sad Puppies and a case for why Ancillary Justice may have won for reasons other than being better than Warbound. Non straw-Dave is just showing some overall bias to the left in SF/F writing.


8 responses to “Revenge of the Petunias”

  1. Thank you for posting this. I am very time pressed at the moment, and I can’t give it the attention your comments really deserve. (I’ve just wasted rather a lot of my allotted internet time on Mark. You and I may differ, but at least you are pleasant about it. I apologize for the somewhat rough tone – his ‘attacking’ stance set my bristles up – and I made the mistake of thinking it was you.) I will go back your original post and reply there I hope tomorrow, about your conclusions. Sorry – deadlines.

    One of the things I realize I was less than clear about was the exclusivity of the two subsets (red and black ball). So not only are the red ball supportive of those issues (and I am pretty sure I said publicly supportive/advocating rather than being “Socialism is good, it works well in XYZ, rather than ‘I am socialist’ – of course if your are a socialist I assume you support it.) but they are strongly opposed to the other subset’s pet issues (so for example a black ball would support the death penalty, and a red ball oppose it).

    Something people do get confused about I’d just like to clarify: I would fail to categorize as either red or black ball. I am not putting this up for self-interest as it has little or no effect on my reader-base. (it is quite plausible that if right wing sf/fantasy became as dominant as left wing is now, it would hurt my sales. I benefit by the present situation.) I do this because I believe discrimination is wrong, foolish and deprives us of great authors, regardless of who the target is. I do come from a situation which I see as similar (Apartheid SA, where my family took the equally unpopular with ruling class position.) I do love the genre, and find many of the argument being advanced for ‘the way it is fine and good’ as being remarkably like the rationalizations as to why a small proportion of the population were better and should enjoy special treatment. The truth is that is not sustainable – a fact I feel is supported out by year-on-year drop in sales of sf/fantasy being reported by traditional publishing. (Meanwhile, several of the pups -excluded from traditional publishing are having bumper years and growing sales). If trad pub sales were rising, I’d accept that maybe everything was fine and good. I may be wrong about what needs changing… but more of the same isn’t working.


  2. Ok – trying to avoid the lengthy fisk type treatment, but there is no honest way to deal with your assessment short of point by point. It is not intended to be unpleasant or confrontational. I mark your text with #

    #’For example if you only intended to demonstrate that some sort of statistical bias is present in the Hugos and that the source of that bias could be due to factors other than the prejudice of inner cliques or WorldCon voters overtly avoiding conservatives, then we disagree on very little.’

    My first step was to display that there was some sort of statistically supported bias. You have to realize you’re stepping into a debate that’s been going on for about ?5 years or more. I first looked at it back then, because I wanted to disprove what I though a ridiculous contention that was the case (John Ringo made the claim. I’ve had my differences with John, but in this case he was right, and I apologized for doubting him). At this stage we thought all that was worth doing was pointing it out, that people would say. ‘we don’t like them, but they deserve a fair go’. Of course that proved wishful thinking. I emigrated and life was full and busy. I know it went on being denied, but until Larry’s performance last year I’d largely forgotten about it. The post you have deconstructed was a response to various other comments – going back some time.

    I now think (and I think I can support) that the result is NOT endogenous (caused by conservative/moderate readership, or authors), or desirable.

    I believe it to be the end product of a series of things INCLUDING inner cliques (who are largely peripheral to worldcon). The latter would be irrelevant without the foundation of prior events. In other words, it would not exist without the discrimination being endemic at most of the traditional publishing houses, and magnified by the client support structure around traditional publishing (ergo retail and the review establishment). When you add the very small numbers needed to win nomination (so for example the staff and attending authors of one large publisher (where paying for membership is tax deductible expense) voting can without another vote win). Where ‘client’ groups who depend on grace-and-favor exist, you have a situation in which this can and will happen. There is a high level of interconnectedness among the ‘inner circle’ – conspiracy is perhaps the wrong word, but co-operation is fair comment.

    The discrimination with the grace-and-favor cliques was more FOR those of their own and in the cliques, than against anyone, IMO. This meant some excellent authors who would even have been politically perfectly acceptable – but just weren’t connected were never on the inside track (this is why when Brad was looking for recommendations, I put forward a couple of Australian authors, and ASIM).

    However, when the political aspect was once again raised last year, there was a knowing and willful effort to exclude – even though this would prove the point that such exclusion happened. Yes, strategically very stupid, but then people who believe they are very powerful often are.

    #Yes. Back to an earlier point. What kind of data is this? Well it is data about how people self-identify with these labels. That is interesting data but precisely because of the shy-conservative issue you mentioned earlier it isn’t very useful for the Hugo

    That was why I looked the surveys on the specific points (which are often raised as parts of political campaigns in the US) they supported, and also what political parties they supported, rather than merely taking it from there. Because a vote is not self-ID. There is a high probability with each of the US parties that voters will substantively skew that way. Also given the shy-conservative situation it may be hard to ID conservatives… but outspoken US liberals are usually quite easy. Keep in mind I belonged to SFWA for years, and still belong to two other groups of US writers. When all of the nominees in a category are people I know fairly well, and whose opinions I have read for years, I could do so fairly well. To get a red-ball tag you had to be outspoken. When I was in doubt I did look up their websites, and do linked searches to the names. When in doubt still I tagged ‘white’. Sometimes I will still be wrong and they’re loudly selling themselves as what they’re not. As we’re talking about perceptions, that’s really irrelevant. The shy con/ outspoken lib situation just can’t exist when all the nominees are outspoken, and that happened quite often.

    You’ve been somewhat critical of the criteria I chose – I could as easily have chosen finance or migration – but this was just harder to find data. What I did was to try and pick ‘hot button’ topics where the people would have loudly advocated for and against. My rule of thumb for site guesswork was at least three issues (either for or against)

    OK. I am out of time again. I’ll try and have another go tomorrow.
    Thank you for your effort and interest. This was a mammoth project, which I undertook thinking it would finally stop the denial, so we could move on to the how, and why, and how heal this – because it is bad news no matter where you sit (it reduces money to lubricate the system, and could very well produce a backlash). All I seem to have done is paint a target on my back. Oh well. Situation Normal.


    • Thanks for replying. You are correct about the scope of this. My posts on your argument actually started out as brief comment I was going to write on File770 – something like Dave Freer has some interesting points but I don’t think he has considered…
      I ditched the initial comment and started writing a slightly longer one off line. That go too long and at that point I decided to revive a plan for a blog project I’d been thinking of 🙂


  3. Gay marriage: In the year I wrote this [support for gay marriage] was 30% among Republicans (71% of whom consider themselves ‘conservative’ – which makes it around 21% of conservatives support it )

    This is such blatant nonsense mathematically that it really calls into question Dave Freer’s ability to perform any sort of statistical analysis. Consider the following parallel statement (with made-up numbers):

    Socialism: In the year I wrote this [support for socialism] was 20% among Democrats (20% of whom consider themselves ‘socialist’ – which makes it around 4% of socialists support it)

    30% is the probability that a randomly selected Republican supports gay marriage. 71% is the probability that a randomly selected Republican considers themselves ‘conservative’. If you multiply these together to get 21%, you get the probability that a randomly selected Republican supports gay marriage AND considers themselves ‘conservative’ assuming that support for gay marriage and ‘conservative’ self-labelling are statistically independent. This is very different from the statement that 21% of conservatives support gay marriage — indeed the assumption that allows you to multiply the numbers together asserts that 30% of conservatives support gay marriage.

    Quite simply, there is not enough information in the 30% and 71% stats to determine the level of gay marriage support among conservatives. It does allows us to set bounds on those levels of support, but those bounds are very loose. In the extreme situation of all non-conservative Republicans (29%) supporting gay marriage, that leaves 1% of Republicans that must be accounted for by conservative supporters of gay marriage. This 1% of Republicans corresponds to ~1.4% (1/71) of conservatives. In the opposite extreme of no non-conservative Republicans supporting gay marriage, the entire 30% of Republicans who support gay marriage must be drawn from the conservative ranks. This works out to 42% (30/71) of conservatives supporting gay marriage. All we can say given that 30% of Republicans support gay marriage and 71% of Republicans are conservative (without any further assumptions) is that the percentage of conservatives who support gay marriage is somewhere between 1% and 42%.

    This kind of high-school-level mistake is typical of the (lack of) mathematical sophistication in Dave Freer’s analysis. I have neither the time nor the inclination to go through his reasoning and point out all its flaws, so I’m thankful there are people like Camestros willing too do it for me. Thanks, guy!


%d bloggers like this: