Heres an excellent article on panelbase and how it leads people to vote in a pre-determined way, and why the results are usually outliers.
Its partly to do with the fact they will take payment and put up any old nonsense and framing of questions...
From another article regarding scottish independance...
*In his now regular independence polling blog at whatscotlandthinks.org, John Curtice, the psephologist at Strathclyde university and the Scottish Centre for Social Research (ScotCen), wrote this week that the Panelbase poll is an outlier.
Panelbase is known, he said, for favouring "yes" support in comparison to other pollsters. That might explain why the SNP has hired it instead of YouGov to do its polling. Curtice said:
as we have repeatedly pointed out, Panelbase's polls (as hitherto conducted for Sunday Times Scotland) have consistently produced results that are more favourable to the Yes side than those of any other pollster.
On average previous Panelbase polls have put Yes on 36% and No on 45%. The average reading across everyone else's polls is Yes 32%, No 53%.
Indeed, Panelbase had a very different approach to the structure of its questions in its contract with the SNP: it asked the "how will you vote" question after first leading its respondents through two positive questions about trust in the Scottish government versus Westminster, and whether they agreed that Scotland "could be a successful independent country."
This formula is crucial, Curtice wrote:
All survey researchers are aware that the responses they get depend can not only depend on the exact wording of the question they ask, but also on what questions have been asked immediately beforehand. There is good reason to believe that this proved important in this case.
By prefacing the referendum voting intention poll with two questions that elicited a response favourable to the Yes side, some respondents could well have been cued into saying Yes when they otherwise would not have done so.*