The British General Election has just finished and one of the principal losers was the opinion polling industry. Yes, those people whose projections politicians invest such faith in screwed up again. The polls had been showing that the two main parties, Conservative and Labour, where basically tied, From these poll figures the pundits, really the classic example of fools with tools, estimated that no party would have an overall majority and that some form of cooperation with other parties would be required in order to form a government. With both parties having a credible chance of actually leading the next government. With Labour having a better chance, as they would be leading the supposed anti-austerity majority.
Well the pollsters got it wrong. The Conservatives won a slim majority and the Labour party had a horrible night, not least in their previous Scottish strongholds. Scotland on election night resembled a pretty gory episode of Game of Thrones, with long time favourites and strongmen dispatched with brutal suddenness.
But not all of the pollsters got it wrong one of them did have a poll that reflected the actual result. And their case gives us an object lesson about filters and biases. The pollsters, Survation, conducted a poll the day before the election and their results were very close to the actual vote count for each party. However they did not publish the results. Why?
Well here are the actual words of their CEO:
The results seemed so “out of line” with all the polling conducted by ourselves and our peers – what poll commentators would term an “outlier” – that I “chickened out” of publishing the figures – something I’m sure I’ll always regret.
Survation fell foul of the bias that is “herd” mentality; that behavior were by members of a group feel safer going along with the thoughts and actions of their peers , the“ herd”, rather using their own analysis. They would rather be wrong together than take the risk of being wrong individually no matter what their evidence reveals. Like lemmings they’d rather jump off the cliff rather than be in the minority safe above ground. They don’t want to be the child that exclaims that the Emperor has no clothes.
The lesson here for PM’s is that when we are in the Orient stage of the OODA Loop we have to be careful of the biases that can impinge on our analysis. And also we have to trust our methodology and have the courage to call it as we see it. If the results appear to be an outlier then by all means check and double check but eventually you have to release them. Playing it safe by following the herd often isn’t right, just ask the lemmings.