How a +3 Romney lead is actually Obama +1.5 (see below)
The value of a poll is supposed to be, "If the election were held today"... who would win. But the election isn't being held that day so certain factors aren't present. Namely, the urgency from all sides to get people to vote, in addition to the conclusion of political debate. Therefore to be accurate a poll must forecast what election day turnout among different groups will be, if that matters. In some elections, a particular candidate is winning/losing within all demographic groups. But in our Presidential system and most important contests members of different groups vote differently. Women/men, blacks/whites/latinos, young,middle aged, older.
The value of a poll is supposed to be, "If the election were held today"... who would win. But the election isn't being held that day so certain factors aren't present. Namely, the urgency from all sides to get people to vote, in addition to the conclusion of political debate. Therefore to be accurate a poll must forecast what election day turnout among different groups will be, if that matters. In some elections, a particular candidate is winning/losing within all demographic groups. But in our Presidential system and most important contests members of different groups vote differently. Women/men, blacks/whites/latinos, young,middle aged, older.
So when looking at certain polls, it is wise to think (maybe research) will election day turnout and results mirror these results.
Take a Florida American Research Poll that shows Romney leading Obama 49 46
black vote 9%
white vote 73%
I must give them credit for showing the cross tabs but here's where I disagree:
black vote 12% 86% for Kerry
white 70%
black vote 13% 96% for Obama
white 71%
That would be a 30% drop in black turnout from 2008
That would be a 25% drop in black turnout from 2004
This is obviously not plausible.
Looking at the last Florida American Research Poll that shows Obama leading Romney 50 45
black vote 12% 91% for Obama
white vote 72%
magic huh? When the percentages equal past turnout, Obama is winning.
In the past survey black = 10.92% in Obama column
In the current survey black = 8.19% in Obama column
============================================
2.79 percent that is not politically defensible
In addition we'd have to subtract a little from Romney's column.
Furthermore Latino turnout in 2008 was 14%
This poll doesn't give Latino breakdown but another poll did 61-31% Obama
So going back to the last poll, using the same vote percentages, but weighting the results to the 2008 turnout which is conistent with the 2004 turnout:
% R O
73 whites 56 39
13 balcks 5 91
14 latinos 40 60 (gave a little to Romney)
Romney Obama
whites 40.88 28.47
blacks .65 11.83
latinos 5.65 8.4
==============================
47.18 48.7
Demographics are steady and must be heeded. Party id can change. Ideology id is fluid.
No comments:
Post a Comment