Worcester's Weblog - Poll Gap: Internet v. Rest; Up-dated Reflecting Sunday's Polls

MORI chairman Sir Robert Worcester analyses the latest opinion poll data.

Congratulations to the YouGov founders and shareholders who will be wealthier on Monday, when their float delivers about 16318 million valuation to the fledgling company. YouGov has a turnover of just under 1632 million, so valued at c. 9 times their turnover, and with a profit last year of c. 163600k, a 30 x profit price. Well done!

In the interview with Peter Kellner in the Evening Standard last week, he extolled their enviable record on getting elections right, save Europe last June, but not mentioning the London Mayor's figures and ignoring completely their problems with the fact that President Kerry isn't after all in the White House, as both YouGov, polling for the Economist, and Harris International, the only other internet pollster, thought would happen, the only two of the 18 polls released in the final two days of the campaign to be so far out, as the table below shows.

All 16 telephone polls were within 177 3% of the election result of the share of each party adjusting for different reporting policies; 12 were within 177 2% and 5 were within 177 1%, yet I still hear many people here talking about how the polls got the American elections wrong.

Are we about to see a replay of the American experience?

So far in this British election, with under two weeks to go, there is evidence of a gap between the phone pollsters, NOP (2 phone), ICM (5 phone), Communicate Research (1 phone), Populus (1 phone) and MORI (3 phone and 1 face to face), and the two Internet based polls, (6) and BPIX, organised by the academics Sanders and Whitely. BPIX (2). haven't told us who does their Internet panel poll yet. So, 23 polls in all, 14 phone/F2F and 9 internet.

Since the start of the election each group has exhibited a remarkable consistency, but significantly different from the other group.

Of course, the gap may close between now and May 5th, but according to our track, trying to average out the rolly pollys so as not to count them more than once each per four-day 'fresh' sample, we find that the phone/F2F polls average 32.5% for Tories, 39.2 for Labour, and 21.0 for the Liberal Democrats, a 6.7 point lead for the Labour Party, whereas the two Internet pollsters, YouGov and BPIX's sub-contractor, have the Conservatives at 34.8%, Labour at 36.3% and LDs at 21.4%, or a 1.5 point lead for Labour.

There's a gap then at the half-way mark of 5.2 points gap between the different methodologies; just 1.2 points over what there was in the USA in November 2004 when the internet folks elected President Kerry by a six point lead over George W.

ICM's survey of issues asked "Which political party is putting forward the best policies on...?, and found Labour with a lead of:"

• 23 points on "the economy generally" (14% rated it "most important")
• 4 points on "health services" (21%)
• 13 points on "the fight against terrorism" (4%)
• 12 points on "education" (13%)
• 9 points on "Europe" (4%)
• 4 points on "taxation and public services" (15%)

Of the eight questions asked, the Tories led on only two, but "taxation and public services" is a double-barrelled question, as some respondents would have had different answers on taxation and public services, and might well have said the Tories were ahead on one and not the other. Tories did lead by:

• 1 point on "law &order" (13%)
• 9 points on "asylum and immigration" (9%)

On "Who would make the best Prime Minister?" ICM's 14 point lead for Mr Blair over Mr Howard stayed consistent, with a 15 point lead this week.

YouGov, Peter Kellner, one of Britain's latest millionaires (as of tomorrow), reports in today's Sunday Times that their poll is in fact re-interviews with some of YouGov's internet panellists first polled two weeks ago. He finds, as we did decades ago when he was on the Sunday Times as a journalist and I was their pollster, considerable 'churning' under the surface of the fairly stable headline figures. Some 18% of their panellists answer the voting intention question differently than they did just a fortnight ago, two thirds of the movement between the "don't know" and "will not vote" category but about six percent of the total sample switching directly from one party to another.

It is sad to see that the great tradition of panel studies' conduct and analysis has been pretty well lost due to tighter financial stringency on the part of the newspapers, as a great deal was learned from panels, repeatedly calling back week after week on the same respondents, which MORI did for several elections for the Sunday Times, and for the Indy on Sunday/Sunday Mirror in the last election, from which much hidden movement was measured, and the reasons for the movements obtained through the use of open ended questions probing the reasons for the movement.