Archive for May, 2015

Laziness and Pointless Hypotheticals

May 28, 2015

A few days ago, quite a few commentators lashed out at the pre-election opinion polls for being grossly inaccurate.  Surely most folk would conclude that the shaking of the fists is justified.  But I cannot myself bring about the energy to join such mindless rage.  However, the thought that an investigation by the British Polling Council is being put forward to find out why the polls failed so spectacularly when the answer is rather simple… that is the kerfuffle making me alternate between chuckles and fumes.

Why not join in?  Surely the polls were so horribly wrong that somebody somewhere needed to be shouted at.  Well, I also believed the polls.  All the way up to the vote I was juggling several coalition hypotheticals in my mind and what this would mean for British foreign policy (my single issue for voting*) as well as the integrity of the British union and the struggles of the poor and working classes of the union, like some mad game-theory analyst.  But unlike a lot of hacks I cannot aim my disappointment at anyone other than myself.

Anyone could have taken two seconds to study the source of the polls in order to find out that there was indeed the source of all the polls as opposed to any kind of plural.  That alone ought to have made anybody suspicious.  And those suspicions would have been validated upon further inspection: sampling methods were skewed and had no real way of assessing what voters were really thinking.  Apparently the hopeless sways to the old socialist Left exhibited by Labour and their choosing of a pitiful shambles of a Labour leader had given the party a strong standing against the Tories, and in the face of a Scottish National Party onslaught in Scotland and a United Kingdom Independence Party vote-grab in Northern England.  Nothing out of the ordinary there, so we thought.

That is what I should have done: dismiss the polls as part of a larger consensus and hence refuse to waste time on coalition hypotheticals.  Jeremy Paxman said more or less the same thing.  If it were a single poll claiming a hung parliament I would have found it easier to dismiss, as I did  with the poll published a week before the Scottish Independence referendum claiming both sides were neck-and-neck (though I did not dismiss it completely since I still feared the momentum of the movement).  On that occasion I had a strong suspicion that such a poll would not have been able to pick up on the signals of the silent majority and that the No side would carry the day.  They did by at least ten percent.  I honed on this conclusion because the Yes side were more likely to make their voices heard on the matter and that would bias the polls.

Unfortunately I dropped my guard for this general election.  In this case I thought there was a degree of legitimacy behind the polls because they had all come to the same findings.  A hung parliament was going to happen and there were seemingly no sways of public thought throughout.  Surely the polls cannot all be wrong, can they?  That was my train of thought and I went with it.  Keeping a polemical, anti-consensus guard up requires practice, and I evidently need more of it.

I am hoping there are some statisticians out there who have picked up on what I am going to say next.  Skeptics and scientists love to talk about why we should dismiss anecdotes – the holy preacher standing near two metal beams that were left standing in the shape of a crucifix among a collapsed building due to an earthquake, claiming such beams prove the existence of the particular god Yahweh and Jesus Christ, is being flippant.  Like-minded folk such as myself point this out whenever we can, and we also hammer down the additional observation that among a collapsed building that was once full of metal beams, had there not been a rearrangement in the form of a probable crucifix that would have been the truly extraordinary event.  Plot out the occurences of crucifixes with the occurences of other objects formed by beams on a graph, and you will get blips here and there when you look for crucifixes.  It is to be expected, not to be startled at.

Similarly, metadata involving tests with medicine will inevitably include some faulty studies that gave conclusions more unusual than others, showing as odd blips away from the general correlation.  Scientists expect this and that is why we have metadata in the first place: the study of studies weeds out the bad science.

Why then, when it came to these polls, were there no blips at all?  Why was no crucifix of any kind standing among the rubble?  You should already know by now: all the polls were using the same unreliable testing methods and sources.  This is what should have rang so many alarm bells within the journalistic profession.  Yet nobody took the simplest of steps to verify their sources.

And I likewise was too lazy.  Though even now, I seem to be among a few people saying so.  What does yelling at the polls achieve, besides avoiding having to point the finger at ourselves?  What use does it do to have a set group of fallible (or worse) researchers do all the thinking for us so that we do not have to?  And why set about a committee to investigate the polling standards when the simplest of all explanations is that bad journalists publish before verification and readers have a tendency to read what conforms to their viewpoints, not what challenges them?  Though I should add the exception to that final question of the masochistic fool who is so annoyed by a website he will read it and froth at the mouth on a daily basis, but only after funding the website every day via advertising to support how much he hates it.

We all know why the exit polls are the most reliable of the lot: the ability to ask voters as they come out the booths means you are not limited to having to track those voters down to all corners of the country.  They are already there waiting, and voters who decide not to disclose how they voted will most likely be divided evenly among all parties (unless there is some toxic political atmosphere that makes people genuinely fearful for speaking of their vote on a particular party, which was not the case here).  Therefore, a safe bet can be put on their accuracy give or take the tightly contested seats, and it shows.

Yet in the final hours before the official result, these were the polls that received the most hostility.  Tories are going to get the majority?  What fool would believe a poll like that?  Not just any silly poll, an exit poll!  There is something wrong here and we shall be proved right in the next few hours!

We need to stop being so bloody stupid.  Using common sense is more than enough to make bad polls fall away.  Internal investigations are unnecessary, thank you very much.

* And as for this single issue, it seemed to me that, despite the Tories’ dangerous tendency to be isolationist such as during the Bosnian invasion by Serbia and Croatia for example, they were the only party that best considered matters of principle over fighting Islamic State, and I can still remember the way in which they refused to be pushed around by Assad and Gaddafi.  I would have voted for them as the lesser evil, despite my deep hatred of everything else Tory.  Labour gave the impression of not giving a single fuck about foreign policy, doing everything they could to avoid any associations with Blair at all costs (except when it came to Islamic State when Milliband said that strikes against them were justified but only because he had to say it), though I ended up voting for Labour because I figured that the only practical power my vote really had was to resist the Scottish National Party onslaught in my constituency (it failed – our long standing Labour seat in Cumbernauld got pounded by the Scottish National Party having over double the votes Labour had).  And I should also give credit to the Liberal Democrats’ foreign policy in their manifesto that I wasn’t expecting, which went into the most detail about Syria and the need to take the side of those fighters who declared themselves enemies of both Islamic State and Assad.