YouGov: Where 1 Mirror reader = 8 Guardian ones
| YouGov’s newspaper weightings | Unweighted number | Weighted number | “Value†of each reader in the poll |
|---|---|---|---|
| Express/Mail | 417 | 301 | 0.72 |
| Sun/Star | 206 | 429 | 2.08 |
| Mirror/Record | 114 | 309 | 2.71 |
| Guardian/Indy | 173 | 62 | 0.36 |
| FT/Times/Telegraph | 224 | 170 | 0.76 |
| Other paper | 217 | 232 | 1.07 |
| No paper | 655 | 505 | 0.77 |
Could newspaper weightings be causing distortions?
The big polling news overnight was that the YouGov dailly poll reported a share of just 9% for the Lib Dems alongside 40% for the Tories and Labour. In doing so the firm was even more out of line with all the other firms that have recently carried out surveys.
| Pollster | Last LD share |
|---|---|
| YouGov (online) | 9% |
| ComRes (phone) | 16% |
| Angus Reid (online) | 15% |
| Populus (phone) | 15% |
| ICM (phone) | 16% |
When this happens repeatedly then it’s worth looking again at the methodology to see if there is anything that might be driving this – and in the table at the top I have reproduced the data that was published yesterday for the the poll announced on Tuesday evening.
Ever since YouGov’s arrival on the polling scene nearly eight years ago a key weighting has been the newspaper that those on its panel say they read. The firm has target weightings based on sales and readership figures and adjusts responses in line with the target.
A serious challenge is that far fewer readers of the red-tops, particularly the Daily Mirror, respond to YouGov’s invitations to take part in its polling. This means that the views and party preferences from those that do take part have to be weighted up – in this case by a factor of 2.71. Compare that with the Guardian/Indy readers who over-respond leading them to be scaled down to just 0.36.
So in this poll the views and party preferences of each individual Mirror/Record reader were worth nearly eight times as much as the Guardian/Indy ones.
This happens day after day after day though not always on the same scale with red-top readers on its panel being invited to take part in polls far more often than those who read other newspapers. Could being asked repeatedly have an impact on their responses and could it mean that the most politically minded are most inclined to take part?
Basically in this poll were the 114 Mirror/Record readers properly representative of the the 309 they were deemed to be when the results were being processed?
Whatever having a segment that almost always has to be scaled up by a considerable amount creates an inherent challenge.