Is Polling Worthwhile?

Please Click the chart to See in Full Resolution

Usually polls only come up in the media when an exciting one comes up; recently there has been a lot of shouting about the Lib Dem position, with the cliché response that polls are not always accurate or that there are X number of years until the next election.

Because the media only look at the 'exciting polls' we tend to lose the real picture; above is a graph of all the polls conducted in 2010/2011, there are times when the peaks and troughs are so close together that it becomes almost impossible to read. The weeks preceding the election had such a density in polling that it's hard to gauge what's really going on.

This graph demonstrates clearly how a singular poll can be worthless, noting the significant peaks and troughs; the margin of error is usually around +/- 3. The problem with the above graph, of course, is that multiple polling companies results have been used all of whom use different weighting techniques, sample sizes and methods of dealing with "don't knows". Note that ICM, for example, have consistently shown slightly higher polls for the Lib Dems whilst YouGov have tended to produce the lowest Lib Dem Polls; this is due to the fact that ICM assume that x% of don't knows will vote how they claim they did in the last election whilst YouGov tend to not reassign don't knows in this way as they think that there is too much assumption in this area.

It is useful to appreciate regular polling; it give us a far better picture of what is going on with their poll. The importance of polls is looking at the moving average; a singular poll showing the Lib Dems at 7% means nothing, but if you put it in a series of polls with a range of 7-13% you can see that the underlying support is more likely to be around 10%.

It is also worthwhile to note that polls are not meant to be prophetic, rather they are gauging support at that moment in time; so the Oldham East & Saddleworth Polls may prove to be worthless (especially as they are likely to have modified some of the voting intention, and that only 1 has been done by each company involved). It is also very difficult to grasp localised polling; attempting to poll a specific group or locality draws problems itself, the larger and more generalised your polling the more accurate it is likely to be. Survation's poll of Oldham East & Saddleworth was a complete waste of time, a poll of just 250 people is simply not large enough.

The value of polling is that it gives us a relative impression of public support; since the election polls have suggested a consistent decline in Lib Dem Support and a gradual increase in both Conservative and Labour support, although the moving average from the last two weeks suggest the first time Labour have produced a clear, and consistent, lead in years but we will know more about the underlying position by next week.

Sorry that was a bit rambly but meh.