|
and after each debate, abc indicated that they sampled more republicans than democrats. In the exact same percentages each time 30 percent democrats, 39 percent republicans. I doubt GAllop has altered their formulas as they plan a single strategy each year and then base ALL of their polling that year on what they believe is the actual percentage that will happen on election day based on prior years. THat is, they use the same percentages for every poll for the entire election season. They do not chage the percentage of people they try to poll, they make calls and when they get to a point where 39 percent of the people they call are repuplicans they stop counting republican calls and when they reacn 30 percent dems they stop. Which is why you see different numbers of people polled for each poll. They try to make their polls fit their predetermined rate of anticipated voter percentage.
All pollsters set up a model of the percentage they think are going to vote REpublican and Demoocrat and then try to poll in those percentages. Which is dumb, and presupposes how many people will vote in each election. A smarter thing wouls be to preselect a set number...1500 people and keep calling until you polled 1500 people and not presuppose anything.
I worked in administration for a laarge libraary system that regularly did polling and they always got the results they wanted by only polling people who cam into the library about bond issues to build new libraries. They would always get results that said 80 percent of library users approved of floating a bond issue to build new libraries. But whe the bond issue was floated it would LOSE by 80 percent, because the library only pollled among the 20 percent of the community that used the library and not the entore tax paying population. That is an extreme example, but it explains what happens with gallop. If gallop polled equal numbers of democrats and republicans you wouild get an actual closer number of what thr public are actually thinking. To preset your expectations skews all polls. Only Zogby and a few independent pollsters even attemt to account and adjust for this, and even they do not do it as well as they could.
They actually try to make predictions of who will vote from past elections, and do not do that well.
|