BERN. The second SRG survey is a serious piece of work, and it is already being read as a result. It is not one. Fieldwork ran from 4 to 11 November. The midpoint was 7 November. The vote is on 29 November. The margin is plus or minus 2.8 points. A 51 to 46 split on the AHV tax, or a 49 to 46 split on arms exports, can reverse inside that band without anyone having lied and without anyone having changed their mind. Treat the fireworks ban, at 63 per cent, as the only question on the card that currently looks decided. Treat the other three as live.
There is a second mistake, and it is the Swiss one. The VAT rise for the 13th pension is a constitutional amendment. It needs the cantons as well as the people. A popular yes that is narrow can still fail if the small cantons say no, and gfs.bern's own reading is that the cantonal majority does not require a popular landslide to appear. Quoting 51 per cent no, and stopping, drops the hurdle that has decided more than one Swiss tax.
Campaigns move, and one just did
Anyone who thinks a mid-November lean is destiny should look at the Fairness Initiative. Yes fell 15 points. No rose 20. A proposal that had a lead now has a deficit, inside a single interval between surveys, because voters heard an argument about clashing tax systems and believed it. The fireworks initiative moved less, and it still moved 6 points toward no, which is what initiatives usually do when the cost is explained, and less than they usually do when the opponents are quiet and poor. The arms revision did not gain a single point of yes, and it gained 5 points of no from people who had been undecided. That is not a stable picture. It is a picture of a reservoir running out.
Turnout is the variable the percentages pretend to have already solved. Certain participation is 52 per cent, against a long-run average of 47.3. It is higher among older voters and among sympathisers of the Greens, the Social Democrats, the Free Democrats and the Green Liberals. It is lower among people with no party, and in the Centre's orbit. On the pension tax, older voters lean yes and younger voters lean no. On arms exports, older voters and men lean yes, younger voters and women lean no. A turnout that over-represents the already convinced is a different electorate from the one in the table. The pension piece has the party splits. The arms piece has the gender and age splits. Read them before you read the headline number a second time.
What a citizen can usefully do with a poll
Use it to see which argument is landing, not to decide whether your own vote is surplus. The no case on the pension tax is landing: each of the three opposing arguments cleared 60 per cent, and the yes arguments trailed by about 10 points. That is information about the campaign, and it is a reason for a yes voter to go, not a reason to stay home. The same is true, inverted, for a no voter on the fireworks ban, where the lead is large and the industry opposing it has not found a loud case. Staying home because a chart looked finished is how a 63 becomes a surprise, and how a 51 becomes an upset in the other direction.
Eleven days is a long time in a Swiss campaign and a short time in a Swiss postal vote. Many ballots are already on kitchen tables. Some have already gone back. A poll of people who still answer a telephone or a website is not a count of those envelopes. Gfs.bern said so, in the sentence under the charts: this is a snapshot, about three weeks before the vote, and it is not a forecast. The press will round it into a winner by lunchtime. You do not have to.