In the recent Ipsos Mori poll, the voting intention for May’s Local Election was along similar past polling in Scotland (in Westminster and Holyrood). The SNP are in the mid 40% range, with the Conservatives around 20%, Labour in the high teens, and so on.


As is a statgeek’s wont, I browsed the additional data and something caught my eye. I noticed that the age group breakdown had some interesting bits and pieces. And then I caught myself, and reminded myself to not get too interested in the minutiae of polling data.


As per usual, there’s a fairly hefty level of support for the SNP in the 16-24 and 25-34 age groups. Of some surprise was the fact that 25% of 16-24s seem to be inclined to vote Conservative. Then I was checking myself for several reasons.

  1. The poll is just a snapshot in time, with two months until the ballot.
  2. The weighted sample total is 753, with a margin of error (MoE) of 3.5%
  3. The sample size for the 16-24 age group is 75, with MoE of 11%
  4. The two older groups make up 80% of the age group poll data

Now it could be that the older folk are greater in number, and it’s probably the case that older folk are more likely to turn out, but I’m not holding my breath on anything a poll says as being representative of an election, two months before the ballots are cast.

What I find very strange is the continued habit of polling companies to separate the age groups the way they do. If we assume that voters might go all the way to 100 years of age (90-110 being a fair guess, split the difference), then here is how the groups are split:

  • 16-24 – 9 years (11%)
  • 25-34 – 10 years (12%)
  • 35-54 – 20 years (24%)
  • 55+ – 46 years (54%)

Granted, the populations at the higher end of the last group taper off, but surely it would be more representative to have 10 year intervals, such as 16-25, 26-35, 36-45 and so on, and at the very least, if there is going to be a xx+ category, let’s start it at 65 so that people can get a handle on the pensioners, as opposed to the non-pensioners.

We can all agree that pensioners’ priorities differ substantially from others. Would six groups instead of four be so difficult? Perhaps the 75 sample in the smallest group highlights that dividing people into small groups for polling purposes invites MoE criticism.

So make the sample sizes bigger and more representative. It’s that easy. A collection of small, bad polls can easily be replaced with an occasional large and reliable one.

One last point. There seemed (to me) to be a glaring omission from this poll. Almost 12% of 2012’s voting intention went to Independents, and that’s the problem with local election polling that prompts people from a pre-prepared list of national parties. Or maybe all the good folk in the Highlands and Islands have changed their vote to SNP? Or maybe the polling companies don’t sample off the mainland so much.

A quick re-jigging for fun:


I fully accept that the 12% voting intention for Independents in 2012 will be different in 2017 (or maybe it won’t, but it probably will). It would be nice to know how many people from the Highlands and Islands were polled by Ipsos Mori.

I recommend you don’t trust small polls (less than a thousand after weighting and intention to vote is factored in), and that you check the regional and population variations are reasonably representative. Quite frankly if polling is that important, do it properly.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s