I've been trying to take the survey to check, link doesn't work. They must know I'm a republican.
Okay, look.
Polling is a business. Telephones and calling staff cost money. A website costs less.
Zogby rationalizes it by saying cellphones and caller ID render traditional calling invalid as a measurement tool. So he goes interactive, applies some sort of weighting to address those not online and makes his estimates.
Of far more curiosity is these wide varying swings in other polls as to what the party mix is in the sample. They are calling at random and asking people their party affiliation and getting these wide swings.
There is something wrong in all this and I don't know what it is. People are not changing their affiliation over 3 and 4 week periods. The polls prior to August showed heavy Dem sampling, then in late August and September they showed heavy GOP sampling. Now the recent Newsweek poll shows heavy Dem sampling.
It is said the Newsweek poll sampled only the west coast on one of the days and no one particulary on Saturday, which means most sampling was Friday night. This could have skewed it.
Perhaps Rasmussen has it right. He forces a mix and holds it constant. All variations thereafter are truly changes of opinion. That's why he has been so stable in his results all year.