Statistics is the study of the collection, analysis, interpretation, presentation, and organization of data. Interpretation is the act of explaining, reframing, or otherwise showing your own understanding of something. Well that all sounds straight forward enough in our statistics-obscessed society and lord knows we’ve had enough of it in the UK over the last few weeks building up to election day. On May 7th the Nation trotted along to the polls.
So, after weeks of numbers, facts, hypotheses and interpretations, at 10pm on election night you could hear the collective gasps of an entire Nation that clearly, no surely not, it couldn’t be, the statistics had all been wrong! As Lizzie Bennett would have said, “I am all astonishment”, except I’m sure the Nation was not so polite in its living rooms. But how? Well that is easy. Humans are involved. Obviously. (Full video is well worth watching – https://www.youtube.com/watch?v=8X6udrxMvCQ, TomoNews US).
The Guardian predicted Conservatives and Labour on an equal footing. UK Polling Report and Electoral Calculus are also both licking their wounds and puzzling why everything was so inaccurate.
I’m not really an expert but spend a fair amount of time playing with data doing bits and pieces of research. A Radio 4 show today really intrigued me: ‘More or Less‘, that featured an election special. Listening to the pollsters – the people that publish the numerous and tedious polls about things – It got me thinking how the election was adulterated.
Problem number 1
Polls rely on sampling the population. They might grab a big bunch of people to question and then match responses to what they believe a ‘representative’ population to be. This is all based on previous experience, models and assumptions, and at this point, a tiny little bit of error can be introduced. Making big interpretations from small samples is always dangerous.
Problem number 2
Polls rely on asking people their intended actions. If you poll me now, I’d tell you when I visit the supermarket tomorrow I shall buy a packet of Haribo Tangfastic. But what it won’t predict that when I get there, if there is a 2 for 1 offer on Fizzy Sea Monsters, I might just change my mind. So basing anything on people’s claimed views and opinions adds a little more error.
Problem number 3
System corruption. We see this in any system of measuring and I’ve written about this before with the National Student Survey. The more organisations relies on the outputs of surveys for for strategies and decision making, the more corrupt the process of gathering will be. We see this in education and healthcare all the time. I haven’t quite thought through the polling business yet, but this might have come into play.
Problem number 4
The pollsters interviewed on the Radio 4 show admitted that while interpreting and publishing data, they keep an eye on each other’s conclusions. I guess we do this in science a bit and nobody really wants to be a maverick with outlandish views – there is always pressure to be part of the set. So if we have a bunch of polling companies all seeing what each other is doing and tempering their observations accordingly, it is generally going to end as an inaccurate amorphous mash. I call this the ‘plasticine’ effect – we may start off with individual ideas and colours, but it is easier to follow the crowd and we ultimately end up a brown sludge colour.
Problem number 5
The media. Notorious for not every accurately portraying or communicating findings, this just adds an extra layer of sludge. The result – the entire Nation duped and not at all looking the right way. A bit like the famous video clip of the crowd at Nasa watching the Space Shuttle take off. They were looking the wrong way and the clip hilariously shows the shuttle taking off behind them!
Conclusion?
I would probably conclude that someone somewhere knew about this all along. The prize however for best ‘Election Night Look of All Astonishment’ does how ever go to Al Murray at the Thanet South count.
