Yonivore
10-12-2004, 08:55 PM
Tea Leaves and Reading Polls (http://polipundit.com/index.php?p=4485)
The polls are out again, and like all entertainment, some are worth your time, and some are written by Zogby. Any vote is nothing more, really, than the total of the sectors forming the whole. That is, if you have good information on the demographics and groups, you can take apart a poll to see where the support came from, and rebuild the numbers to see how they will play out.
There was some discussion yesterday, about the difference between the CNN/USA Today/Gallup poll and the solo Gallup poll. The reason this is interesting, is that the two groups take the same respondent pool and answers, then build their polls in slightly different methods, sometimes producing different final numbers. This proves that a poll is only valuable if you know the components which built it, or else you are buying an opinion with nothing to back it up. I have been quite firm in rejecting those polls which will not show their work, for exactly that reason, no matter their previous claims or whether their results match my own opinion.
Of the polls released in the past week, only AP/Ipsos, Gallup and ABC News/Washington Post provided demographic details (CBS News has a new poll out, but has not yet released its internals). So, I am presenting the available information from those polls. I would mention before I begin, to note any sudden and unexplained changes in support levels, especially if one poll changes and the other does not. Ipsos released detailed demographics only for its October 7 report, so there is no shift to report for that poll. Bear in mind, also, that Ipsos used a peculiar and imbalanced sampling for its poll, which may skew results for their numbers.
The Gallup poll shows a 48-48 tie on October 4th, moving slightly to a 49-48 Kerry advantage as of October 11th. ABC News/Washington Post shows a 50-47 Bush lead on October 8th, improving to a 51-45 Bush lead on October 11th. Ipsos says Kerry leads 50-46, but I have addressed their methods in another article.
Many of the areas are agreed upon by the two polls. ABC shows support for Bush by men rose from 54% to 56%, while Gallup says it dropped from 54% to 53%, but that’s still fairly close. ABC says support for Kerry by man dropped from 42% to 40%, while Gallup says it dropped from 45% to 44% (note: this drop for both Bush and Kerry among men, indicates a rising number of uncertains among Gallup men). Ipsos says Bush leads Kerry among men, 50-46.
ABC shows support for Bush by women dropped slightly 46% down to 45%. Gallup says support for Bush by women dropped from 45% to 43%. ABC says support for Kerry by women dropped from 51% to 50%, while Gallup says it rose from 51% to 52%, still comparable. (note: this drop for both Bush and Kerry among women, indicates a rising number of uncertains among ABC women). Ipsos shows Kerry leading Bush among women, 54-37.
ABC shows support for Bush among whites rose from 55% to 56%. Gallup says support for Bush among whites rose from 54% to 55%. ABC says support for Kerry among whites dropped from 42% to 40%, while Gallup says support for Kerry among whites dropped from 43% to 41%. Ipsos shows a tight race here, Bush leading Kerry 49-47 (bear in mind, the sampling imbalance comes into play here).
ABC shows GOP support for Bush steady at 90%. Gallup says it dropped from 94% to 92%. Ipsos shows GOP support for Bush at 91%.
What’s interesting to me, however, are the areas where the polls do not agree.
ABC shows non-White support for Bush rising slightly, from 31% to 32%, both numbers well above his 2000 levels. Gallup says that between October 4 and October 11, support for Bush among non-whites dropped from 29% to 20%, his lowest showing this year, and an apparent outlier when compared to demographics from other polls which display this information (CBS/NYT, Pew, Newsweek, Marist, ABC, Ipsos). Ipsos shows non-White support for Bush at 27%. ABC shows non-White support for Kerry dropping slightly, from 69% to 68%, while Gallup says it jumped 6 points, from 68% to 74%. Ipsos shows Kerry’s support by non-Whites at 67%.
ABC shows support for Bush in the East dropping from 48% to 46%, while Gallup shows the numbers steady at 42% in the East. Ipsos shows support for Bush in the East at 38% (they describe it as “Northeast”, indicating a smaller region). ABC shows support for Kerry in the East rising from 49% to 51%, while Gallup shows Kerry rising from 54% to 55% in the east. Ipsos shows Kerry’s support in the “Northeast” at 57%.
ABC shows support for Bush in the Midwest steady at 46%. Gallup says it shot up strongly, rising from 39% to 52%. Ipsos shows support for Bush in the Midwest at 42%. ABC shows support for Kerry in the Midwest dropping from 50% to 48%, while Gallup says it plunged, from 58% to 43%. Ipsos says Kerry’s support in the Midwest was at 53%. That should explain why both candidates are spending a lot of time there.
The strangest numbers, I think, come from the South. ABC says support for Bush in the South rose from 54% to 58%, which would make sense after his improved 2nd debate. But Gallup says support for Bush plummeted in the South, from 61% to 51%! There is nothing anywhere to explain this plunge, though I suspect Gallup may well have sampled many more urban respondents in the South, where Democratic support is more concentrated. Ipsos says Bush’s support in the South is at 56%. ABC says support for Kerry in the South dropped from 43% to 40%. Gallup says support for Kerry in the South shot up 8 points, from 37% to 45%, but again, there is no reason which would explain this sudden rise in this demographic. Ipsos says support for Kerry in the South was at 41%.
In the West, ABC says Bush dropped a little, from 50% to 48%. Gallup shows a sharper drop, from 51% to 46%. Ipsos says Bush’s support in the West is at 41%. ABC says Kerry’s support in the West also dropped, from 48% to 46% (again indicating rising number of uncertains), while Gallup says Kerry’s support in the West rose from 47% to 50%. Ipsos says Kerry’s support in the West is at 55%.
Among Democrats, ABC says Bush rose slightly, from 12% to 13%. Gallup disagrees, saying Bush fell from 10% to 8%, while Ipsos says Bush claims 4% of Democrats (remember the heavy West coast sample). ABC says Kerry has dipped from 87% to 85% among Democrats, while Gallup says he has risen slightly, from 89% to 90%. Ipsos says Kerry has 91% among Democrats, but again, the imbalanced sampling appears to have caused this.
Among Independents, ABC says Bush has dropped a little, from 50% to 48%, but still leads Kerry in this group. Gallup says Bush has risen from 40% to 42%, but still trails Kerry in this group. I do not mention Ipsos’ numbers for Independents here, because with only 49 respondents out of a total of 944, the number is too small to be statistically valid. ABC says Kerry’s numbers with Independents has fallen from 47% to 44% (again, note the fall in both Bush and Kerry among Indies, meaning a rise in uncertains), while Gallup shows Kerry falling with Independents from 54% to 50%.
The Age breakdown is difficult to compare, because different ranges were used by different polls. ABC says that among 18-30 year olds, Bush stayed steady at 48%, while Kerry climbed from 49% to 52%. Ipsos says 48% of 18-29 year olds prefer Bush, while 46% prefer Kerry. Among 31-44 year olds, ABC says Bush dropped from 56% to 55%, while Kerry dropped from 42% to 39% (again, the double drop means more uncertains). Ipsos says that for 30-39 year olds, 43% prefer Bush, while 55% prefer Kerry.
Gallup says for the 18-49 year old age group, Bush dropped from 50% to 48%, while Kerry rose from 46% to 48%. For the 40-49 year old age group, Ipsos says Bush leads Kerry, 50% to 47%.
ABC says that among 45-60 year olds, Bush rose from 45% to 50%, while Kerry fell from 53% to 46%. Gallup says that 50-64 year olds, Bush fell from 52% to 49%, while Kerry rose from 46% to 48%. Ipsos says for 50-64 year olds, Bush takes 45%, while Kerry claims 49%.
ABC says for voters 61 years or older, Bush dropped from 51% to 48%, while Kerry rose from 44% to 47%. Gallup says that for voters 65 or older, Bush rose from 43% to 48%, while Kerry dropped from 52% to 50%. Ipsos says that for voters 65 or older, 41% prefer Bush, and 55% prefer Kerry.
When one poll says something different from another, either could be right. But when a poll shows sharp changes in a short time for no apparent reason, I’d suggest the reader take the results with some salt. Just as there is not question the first debate did some damage to Bush, but only a little, the second helped Bush, but again, only a little. The third debate, now that we see the effects of the previous debates, will be watched by many people, and is therefore important. It’s interesting to me, however, to see the apparent rise in uncertains. That looks to be working in Bush’s favor, but he still has to win these voters over. If President Bush is the same confident, relaxed man we saw last Friday, that will happen.
The polls are out again, and like all entertainment, some are worth your time, and some are written by Zogby. Any vote is nothing more, really, than the total of the sectors forming the whole. That is, if you have good information on the demographics and groups, you can take apart a poll to see where the support came from, and rebuild the numbers to see how they will play out.
There was some discussion yesterday, about the difference between the CNN/USA Today/Gallup poll and the solo Gallup poll. The reason this is interesting, is that the two groups take the same respondent pool and answers, then build their polls in slightly different methods, sometimes producing different final numbers. This proves that a poll is only valuable if you know the components which built it, or else you are buying an opinion with nothing to back it up. I have been quite firm in rejecting those polls which will not show their work, for exactly that reason, no matter their previous claims or whether their results match my own opinion.
Of the polls released in the past week, only AP/Ipsos, Gallup and ABC News/Washington Post provided demographic details (CBS News has a new poll out, but has not yet released its internals). So, I am presenting the available information from those polls. I would mention before I begin, to note any sudden and unexplained changes in support levels, especially if one poll changes and the other does not. Ipsos released detailed demographics only for its October 7 report, so there is no shift to report for that poll. Bear in mind, also, that Ipsos used a peculiar and imbalanced sampling for its poll, which may skew results for their numbers.
The Gallup poll shows a 48-48 tie on October 4th, moving slightly to a 49-48 Kerry advantage as of October 11th. ABC News/Washington Post shows a 50-47 Bush lead on October 8th, improving to a 51-45 Bush lead on October 11th. Ipsos says Kerry leads 50-46, but I have addressed their methods in another article.
Many of the areas are agreed upon by the two polls. ABC shows support for Bush by men rose from 54% to 56%, while Gallup says it dropped from 54% to 53%, but that’s still fairly close. ABC says support for Kerry by man dropped from 42% to 40%, while Gallup says it dropped from 45% to 44% (note: this drop for both Bush and Kerry among men, indicates a rising number of uncertains among Gallup men). Ipsos says Bush leads Kerry among men, 50-46.
ABC shows support for Bush by women dropped slightly 46% down to 45%. Gallup says support for Bush by women dropped from 45% to 43%. ABC says support for Kerry by women dropped from 51% to 50%, while Gallup says it rose from 51% to 52%, still comparable. (note: this drop for both Bush and Kerry among women, indicates a rising number of uncertains among ABC women). Ipsos shows Kerry leading Bush among women, 54-37.
ABC shows support for Bush among whites rose from 55% to 56%. Gallup says support for Bush among whites rose from 54% to 55%. ABC says support for Kerry among whites dropped from 42% to 40%, while Gallup says support for Kerry among whites dropped from 43% to 41%. Ipsos shows a tight race here, Bush leading Kerry 49-47 (bear in mind, the sampling imbalance comes into play here).
ABC shows GOP support for Bush steady at 90%. Gallup says it dropped from 94% to 92%. Ipsos shows GOP support for Bush at 91%.
What’s interesting to me, however, are the areas where the polls do not agree.
ABC shows non-White support for Bush rising slightly, from 31% to 32%, both numbers well above his 2000 levels. Gallup says that between October 4 and October 11, support for Bush among non-whites dropped from 29% to 20%, his lowest showing this year, and an apparent outlier when compared to demographics from other polls which display this information (CBS/NYT, Pew, Newsweek, Marist, ABC, Ipsos). Ipsos shows non-White support for Bush at 27%. ABC shows non-White support for Kerry dropping slightly, from 69% to 68%, while Gallup says it jumped 6 points, from 68% to 74%. Ipsos shows Kerry’s support by non-Whites at 67%.
ABC shows support for Bush in the East dropping from 48% to 46%, while Gallup shows the numbers steady at 42% in the East. Ipsos shows support for Bush in the East at 38% (they describe it as “Northeast”, indicating a smaller region). ABC shows support for Kerry in the East rising from 49% to 51%, while Gallup shows Kerry rising from 54% to 55% in the east. Ipsos shows Kerry’s support in the “Northeast” at 57%.
ABC shows support for Bush in the Midwest steady at 46%. Gallup says it shot up strongly, rising from 39% to 52%. Ipsos shows support for Bush in the Midwest at 42%. ABC shows support for Kerry in the Midwest dropping from 50% to 48%, while Gallup says it plunged, from 58% to 43%. Ipsos says Kerry’s support in the Midwest was at 53%. That should explain why both candidates are spending a lot of time there.
The strangest numbers, I think, come from the South. ABC says support for Bush in the South rose from 54% to 58%, which would make sense after his improved 2nd debate. But Gallup says support for Bush plummeted in the South, from 61% to 51%! There is nothing anywhere to explain this plunge, though I suspect Gallup may well have sampled many more urban respondents in the South, where Democratic support is more concentrated. Ipsos says Bush’s support in the South is at 56%. ABC says support for Kerry in the South dropped from 43% to 40%. Gallup says support for Kerry in the South shot up 8 points, from 37% to 45%, but again, there is no reason which would explain this sudden rise in this demographic. Ipsos says support for Kerry in the South was at 41%.
In the West, ABC says Bush dropped a little, from 50% to 48%. Gallup shows a sharper drop, from 51% to 46%. Ipsos says Bush’s support in the West is at 41%. ABC says Kerry’s support in the West also dropped, from 48% to 46% (again indicating rising number of uncertains), while Gallup says Kerry’s support in the West rose from 47% to 50%. Ipsos says Kerry’s support in the West is at 55%.
Among Democrats, ABC says Bush rose slightly, from 12% to 13%. Gallup disagrees, saying Bush fell from 10% to 8%, while Ipsos says Bush claims 4% of Democrats (remember the heavy West coast sample). ABC says Kerry has dipped from 87% to 85% among Democrats, while Gallup says he has risen slightly, from 89% to 90%. Ipsos says Kerry has 91% among Democrats, but again, the imbalanced sampling appears to have caused this.
Among Independents, ABC says Bush has dropped a little, from 50% to 48%, but still leads Kerry in this group. Gallup says Bush has risen from 40% to 42%, but still trails Kerry in this group. I do not mention Ipsos’ numbers for Independents here, because with only 49 respondents out of a total of 944, the number is too small to be statistically valid. ABC says Kerry’s numbers with Independents has fallen from 47% to 44% (again, note the fall in both Bush and Kerry among Indies, meaning a rise in uncertains), while Gallup shows Kerry falling with Independents from 54% to 50%.
The Age breakdown is difficult to compare, because different ranges were used by different polls. ABC says that among 18-30 year olds, Bush stayed steady at 48%, while Kerry climbed from 49% to 52%. Ipsos says 48% of 18-29 year olds prefer Bush, while 46% prefer Kerry. Among 31-44 year olds, ABC says Bush dropped from 56% to 55%, while Kerry dropped from 42% to 39% (again, the double drop means more uncertains). Ipsos says that for 30-39 year olds, 43% prefer Bush, while 55% prefer Kerry.
Gallup says for the 18-49 year old age group, Bush dropped from 50% to 48%, while Kerry rose from 46% to 48%. For the 40-49 year old age group, Ipsos says Bush leads Kerry, 50% to 47%.
ABC says that among 45-60 year olds, Bush rose from 45% to 50%, while Kerry fell from 53% to 46%. Gallup says that 50-64 year olds, Bush fell from 52% to 49%, while Kerry rose from 46% to 48%. Ipsos says for 50-64 year olds, Bush takes 45%, while Kerry claims 49%.
ABC says for voters 61 years or older, Bush dropped from 51% to 48%, while Kerry rose from 44% to 47%. Gallup says that for voters 65 or older, Bush rose from 43% to 48%, while Kerry dropped from 52% to 50%. Ipsos says that for voters 65 or older, 41% prefer Bush, and 55% prefer Kerry.
When one poll says something different from another, either could be right. But when a poll shows sharp changes in a short time for no apparent reason, I’d suggest the reader take the results with some salt. Just as there is not question the first debate did some damage to Bush, but only a little, the second helped Bush, but again, only a little. The third debate, now that we see the effects of the previous debates, will be watched by many people, and is therefore important. It’s interesting to me, however, to see the apparent rise in uncertains. That looks to be working in Bush’s favor, but he still has to win these voters over. If President Bush is the same confident, relaxed man we saw last Friday, that will happen.