Reckless Talk by Polling 'Experts'

You know what they say about polls...well I don't believe they are all lies and statistics about Canadian opinions can be very useful, when considered in a balanced way.  Apparently if you are the head of a polling firm, such as Nik Nanos, however, you are able to infer trends from a very short sequence of data.  Just taking a look at the chart in this article with a short, narrow peak in the past two weeks being called a new normal is a bit of a laugh.  How can a few points define a new normal? 

What Nanos is talking about here is an intuition that opinion polls have a kind of inertia that keeps a party's support around a certain level until some news or institutional change comes along to upset it.  That seems a reasonable idea, but just like financial markets, detecting that we are currently in one of these steady periods is much easier to determine in hindsight than at the moment.  To speculate that two weeks of polls at some level indicate a new level of support does not match the evidence of how much variance there can be over the long wrong in party support.  This is especially true for a period that saw a momentary public certainty that an election was imminent, strong evidence of an economic turnaround that may well change like the wind again in weeks to come and the unprecedented humanizing spectacle of the prime minister playing the Beatles on piano.  I would think this couple weeks are very likely to be an anomaly given these events.  But it doesn't matter what I think, when the head of a company conducting these polls expresses that opinion it carries more weight and makes me wonder if preconceived ideas are entering into the polls that political pundits rely on so heavily to have something to talk about.

Personally I think its completely inappropriate for pollsters to speculate on where trends are going, they should behave as analytical statisticians who describe in an unbiased manner where we are now and what we can say with a large degree of certainty about the present data given historical trends.  This loose talk of a 'new norma' is unscientific, biasing towards future polls and makes one wonder at the non-partisan nature of the polling process itself.

No comments:

Post a Comment