In this morning’s Metro there is a new voting intention poll for Harris. Harris were once one of the most regular UK political pollsters but left the scene sometime after 1997, becoming a leading online polling company in the USA as Harris Interactive. They made a surprise return just before the 2005 election and did very well, but since then we have only had one voting intention poll from them, though they have done other polling for Metro of a panel of working age Londoners (called something like UrbanLife). (That isn’t them anyway – BMRB conduct it!)

Today’s poll has topline voting intention figures of CON 39%(+4), LAB 30%(+10), LDEM 22%(+6) – changes are from Harris’s last poll in June 2009, which had “others” on a rather incongruous 29%, hence all the partie being up. We can’t tell a vast amount from the poll, since without a recent track record we don’t know how it compares to other pollsters figures, but we can at least hope that it’s a sign of regular Harris polls to come in the run up to the election.

I haven’t been able to have a dig around in their tables yet, but for those interested in methodology, here’s what I wrote in June 2009: “Harris are on online company with their own panel, like YouGov and Angus Reid. Their polls are weighted by age, gender, educational achievement, region and internet usage, but not it would seem by past vote or party ID. Instead Harris use something they call “propensity score weighting”, a proprietory weighting they say corrects for behavioural and attitudinal biases from different peoples likelihood to be online. Exactly how it does so, we don’t know.” Of course, I can’t guarantee they haven’t changed any of that since!

The poll was conducted between the 16th and 22nd and included 900 people (so quite a long time for a small number of interviews).


153 Responses to “Harris poll shows 9 point lead”

1 2 3 4
  1. Rob & co go to new thread

  2. Colin – The Treasury is repeated at 11.30pm and then will be on the I-Player .

    I hesitate to say it Colin, but make the most of the BBC. It won’t be able to do what it does for much longer if what you think you want comes about

  3. COLIN – yep, all consistent. The important figures are not the make up of the raw sample, but the demographics of the weighted sample. Are they being weighted to the same targets? On the whole the answer is yes they are (the YouGov figures are occassionally updated to take account of people joining the panel at later date when the balance of party ID has shifted, but that hasn’t been done at all recently).

    The recent change in the party ID voting doesn’t make any significant difference at all. It came about from the period testing the daily polling – on one day there was a freak result and when we analysed the sample the reason was the proportion of what we call the loyal Labour ID (that is, Labour identifiers who voted Labour in 2005) was higher than we’d got in all the other polls that week. We decided to split up the Labour ID group into those who voted Labour in 2005 and those who did not vote Labour in 2005 and weight them seperately. It means in practice YouGov’s political weighting is now an amalgam of party ID and recalled past vote weighting.

    In practice, while removing one possible cause of sample variation, the change doesn’t produce any partisan change in the voting intention figures: we reweighted about three weeks worth of data to see what effect it had, and in almost every case it made no difference at all.

1 2 3 4