Polling clarity

For a long, long time I have ignored polls. Not because they aren't helpful, but because there are so many reasons why individual polls jump around, and it was so tricky to see the underlying trend. Pollster.com fixed that problem for me, and it is now the only place I go to see what is happening in US elections. The clarity of a clear message, but the high-resolution of individual data points for those who look a little closer.


I was inspired, and have basically copied this model for the upcoming UK election. Hence my new blog: Polling Clarity.


I use exponential smoothing to reduce the noisy pattern you would get if you drew a line between every poll that came out, and try to pick up just the underlying pattern.

It does of course make a difference how reactive I make the trend line. I can set it to zig zag around and follow all of the latest trends, but this would be a mess and would miss the point. Instead I have found what I think is a good reactivity mix of taking some of the new information that an additional poll provides, and yet having regard for the other recent polls, and so trying to represent the real underlying trend.

I will try and update it with each new poll, for the duration of the election, whenever it should be announced.