2016 Postmortem
Related: About this forumLet's talk likely voter screens. We will see more polls similar to the Iowa one
First, my general disclaimer: I am a Clinton supporter, but I am also a statical analyst. When talking polls I try to be factual on why they're set up the way they are and where methodologies may fall apart.
Anyway, the Monmouth poll is a likely voter poll and not just a registered voter poll. Most polling that is occurring right now is for registered voters, which will begin to switch to likely voters as the primaries draw closer.
Registered voter polls are pretty simple affairs--if you're registered with a party then your answer is tallied.
Likely voter polls, on the other hand, attempt to answer the question of who will actually come out to vote when the election or caucus arrives. For proper statistical consideration you can't simply ask the person and take their word that they will come out, so instead the vast majority of polling agencies use past history. Specifically, has that person voted in the past. Then they ask are you planning on voting in the next primary.
Unfortunately, the last Democratic primary was 8 years ago, which means that only millennials who are 26 years or older now will trigger the past history requirement. All of the college-aged kids who have registered for the first time are not considered likely voters and won't be by most agencies. And it should be noted that specific age group is notoriously unlikely to vote in a primary.
So you can see why likely voter screens tend to older voters.
Monmouth appears to have a very aggressive weighting mechanism attached to their current poll, and I do believe it is underrepresenting turnout for the under 50 population (In 2008, 27 and under were 22% of the caucus demographics). But that does not mean this is a deliberate attempt to prop up Hillary.
The fact is, likely voter screens, as they roll out, will discount millennial responses. While you might not see something as outrageous as Clinton being +41 in Iowa, Sanders will lose points when pollsters switch from registered voters. It is one of the primary risks of counting heavily on the youth vote in polpoint
Edit: Case in point, the Loras poll just now being reported is also a likely voter poll.
Ed Suspicious
(8,879 posts)Is it a means of or technique used in screening out a demographic that you are not interested in polling? Would you mind going further into that basic concept of what is a voter screen? Thank you for the interesting post and the look behind the scenes.
Godhumor
(6,437 posts)If the pollster calls someone they are directed to end the conversation if the respondent doesn't meet polling requirements. So if a poll is of likely voters only, someone who hasn't voted previously will be thanked for their time and let go.
Ed Suspicious
(8,879 posts)Godhumor
(6,437 posts)Some pollsters also give directive to stop calls once certain demographic milestones are reached. So, once, say, 75 African American responces are collected the pollster will terminate further calls with that group to keep their overall sample weights intact. Not all do this, but some will, which is bothersome.
Others will simply adjust population weights after all outreach is done, which can be problematic for other reasons.
MohRokTah
(15,429 posts)It's a likely caucus goer poll, which is even more discerning than a likely voter poll.
A likely voter need only show up and cast a ballot, not terribly time consumer.
A likely caucus goer must show up and stay for the whole thing, a process which could take an hour or substantially longer depending upon who actually shows up and how familiar with the procedures everybody is.
Godhumor
(6,437 posts)My guess is that a caucus screen losses more people as a result of the screen than those who vote in primaries.
GusBob
(7,286 posts)People are complaining that 400 is too small
without making you type all day, how do they extrapolate the numbers
Godhumor
(6,437 posts)You can find it here:
http://www.democraticunderground.com/1251567623
GusBob
(7,286 posts)elleng
(130,878 posts)I don't read polls (usually,) and would add that it seems to me they're particularly inaccurate re: Iowa CAUCUS-goers, especially given what I've read here about how they're conducted.
HerbChestnut
(3,649 posts)Great Post.
Godhumor
(6,437 posts)Apparently I just can't help myself when this place goes ya-ya over polling, heh.