According to a Scripps Howard News Service story today, Bush’s Hispanic support in the national NEP exit poll has now been revised down from 44 percent to 40 percent.
Word of this revision came from an NBC official, elections manager Ana Maria Arumi. According to the story, Arumi says that:

…the exit poll over sampled in South Florida where Republicans are strong among Cuban-Americans.
For the revised figures the networks combined 50 state exit polls, which reflected more than 70,000 interviews.

This is obviously a step in the right direction and I can’t help but feel some vindication from it, but it does not answer some key questions about this particular survey snafu and actually raises some additional ones.
1. If the initial figure was so far off, why was that? Could it really all be from oversampling in South Florida? But what about the huge overestimate of Bush’s Hispanic support in Texas which was just revised downward in the last few days? Isn’t whatever caused that overestimate likely to have been part of the problem too? Has that correction of the Texas data even been incorporated into this new estimate of the national figure?
2. And if the Texas data were so screwed up–as the exit poll authorities now appear to admit–how do we know that there weren’t other states that were also seriously messed up and are now being uncritically incorporated into this new state-based national estimate?
3. If it is necessary to combine all the state data to get a reasonable national estimate for this particular demographic group, what about other demographic groups? Should we also use state-based national estimates for them? If not, why not?
4. Who’s making the decisions here anyway? The Texas revision was announced by AP and credited to Mitofksy/Edison, but this revision is announced by NBC, an NBC official is the one making the claim about South Florida oversampling and the networks are described as the ones pooling the 50 state polls (see above) to get the national estimate. What on earth is going one here?
5. Whoever is, or is not, in charge, at some point there should be an explanation forthcoming of what exactly went wrong, how exactly it was fixed and why exactly it was deemed appropriate to fix it in that particular way. At this point, all we can do is guess at all these things, which reduces one’s faith that the fixes they are currently implementing are really the right ones and are (finally) producing correct figures.

Or, more accurately, closer to where it was to begin with. I argued the other day that it was quite unlikely that Bush actually got 44 percent of the Hispanic vote, as the national exit poll claimed, and that the 59 percent share given him by the Texas state exit poll was particularly fanciful.
Now we have this AP item, showing a drastic downward revision in the Texas figure for Bush’s Hispanic support:

In the Nov. 3 BC-ELN–Texas Glance and BC-TX Exit-Poll Excerpts, The Associated Press overstated President Bush (news – web sites)’s support among Texas Hispanics. Under a post-election adjustment by exit poll providers Edison Media Research and Mitofsky International, 49 percent of Hispanics in the state voted for Bush, not a majority. The revised result does not differ to a statistically significant degree from Bush’s 43 percent support among Texas Hispanics in a 2000 exit poll.
The revised BC-TX-Exit-Poll Excerpts showed that 20 percent, not 23 percent, of all Texas voters were Hispanic. They voted 50 percent for Kerry and 49 percent for Bush, not 41-59 Kerry-Bush.

Quite a change and it affects not just the Texas Hispanic estimate, but the national one as well. As Steve Sailer correctly points out:

That reduction of 10 points in Texas would appear to knock almost 2 points off Bush’s national Hispanic share by itself (since the exit poll claimed that Texas accounted for 18% of America’s Hispanic voters), and the reduction in the Hispanic share of the Texas vote from 23% to 20% would reduce Bush’s national Hispanic share as well (because he still had more Hispanic support in Texas than nationally).

We shall see what further exit poll revisions do to the estimates of Bush’s Hispanic support. But my–and Sailer’s–estimate that Bush received around 39 percent, not 44 percent, of the Hispanic vote is looking better and better.

By Alan Abramowitz
Perhaps you’ve seen or heard of an analysis by Michael Hout, a Berkeley sociologist, of the impact of e-voting in Florida. Hout and his associates claim that Bush did better than expected in the 15 Florida counties using e-voting. See the link below to their report.
I did my own analysis of their data. It does not support their conclusions. In fact, I find that Bush did slightly worse than expected in the 15 e-voting counties.
I did three things. First, I just compared the change in percent for Bush in Florida counties with and without e-voting. Contrary to their conclusion, Bush gained more support in counties without e-voting. Then I looked at a scatterplot of Bush2004% by Bush2000%. There is no indication at all here of any non-linearity in the relationship. Therefore, I cannot see why the Hout team added a quadratic term to the model. Then I did a regression analysis of Bush2004% with Bush2000% and a dummy variable for e-voting counties. The dummy variable had a negative but statistically insignificant effect. So if anything, Bush did slightly worse in 2004 in counties with e-voting when you control for his support in 2000. My guess is that this is because the e-voting counties tend to be in large metropolitan areas but Bush’s gains were greater in smaller, rural and exurban counties.

The new CBS News/New York Times poll suggests that, as indicated by the postelection DCorps poll, Bush doesn’t have much of a mandate for his policies and is unlikely to enjoy much of a honeymoon from a public that preferred him only marginally to John Kerry.
Bush’s overall approval rating in the poll is 51 percent and more people think the country is off on the wrong track (54 percent) than feel it is going in the right direction (40 percent). That’s a net of -14 on wrong track, actually slightly worse than recorded by CBS right before the election.
Bush’s approval ratings in specific areas, except for the campaign on terrorism, are all lower now than they were right before the election: 44 percent approval/48 percent disapproval on handling foreign policy; 42/57 on the economy; and 40/55 on the situation in Iraq. On the campaign against terrorism, however, his rating is 59/37, up 4 points since before the election.
The poll also finds more of the public uneasy (51 percent) than confident (47 percent)in Bush’s ability to “deal wisely with a difficult international crisis” and with his ability to “make the right decisions about the nation’s economy” (52/46).
On Social Security, by 51-38, the public thinks Bush is not likely to make sure Social Security benefits are there for “people like you”. Also, they don’t believe, by 51-31, that the Social Security system will be able to provide the proper level of benefits for them when they retire. However, the public is split on whether it would be a good idea (49 percent) or bad idea (45 percent) to let individuals invest part of their Social Security taxes on their own–Bush’s signature proposal in this area.
On corporate influence, two-thirds (66 percent) think large corporations have too much influence on the Bush administration, compared to just 19 percent who corporations have the right amount of influence and 4 percent who think they have too little (!).
On taxes, less than a third (32 percent) think Bush’s tax cuts since 2001 have been good for the economy (64 percent think they’ve been bad or made or made no difference) and only 31 percent think that additional reductions in taxes (another signature Bush proposal) would be good for the economy (62 percent think such reductions would be bad or make no difference). And, on the question of whether the temporary tax cuts passed in 2001 should be allowed to expire, more say they should expire (45 percent) than say they shouldn’t (41 percent).
On budget priorities, by more than 2:1 (67-28), the public thinks reducing the federal budget deficit should be a higher priority than cutting taxes. (No question was asked about spending on health care, etc. vs. cutting taxes, but that result would likely be even more lop-sided.)
On Iraq, for the first time since July, more say we should have stayed out of Iraq (48 percent) than say we did the right thing to take military action against Iraq (46 percent). Also, for the very first time, an outright majority (51 percent) says that the war in Iraq is separate from the war on terrorism (up 9 points since right before the election). Of those who say the war in Iraq is part of the war on terrorism (43 percent), 34 percent say it is a major part and the other 9 percent say it is a minor part. Finally, a plurality now say (46-45) that is not possible for the US to create a stable democracy in Iraq.
On the political parties, despite the Republicans’ gains in the 2004 election, the public now views the Democrats substantially more favorably (54 percent favorable/39 percent unfavorable) than they view the Republicans (49/46).
And as for that parity in party ID indicated by the NEP exit poll? It’s already gone, if it was really there to begin with. Confirming the Annenberg Election Survey results I wrote about a couple of days ago, the CBS/NYT poll now shows the Democrats with a 7 point lead on party ID (36-29).

According to the 2004 NEP exit poll, Democrats and Republicans were dead-even on party ID (37-37) in the 2004 election, a 4 point shift from the 39-35 Democratic advantage registered by NEP’s predecessor, VNS, in the 2000 election.
Did a shift of this size really take place in partisan allegiances of the American electorate? Given how much the NEP poll apparently had to weight down Kerry voters and weight up Bush voters to conform to the election result, there are certainly reasons to be cautious about that poll’s measurement of a characteristic so closely correlated with the presidential vote. It is also possible the NEP’s measurement reflects less a change in underlying sentiment among the electorate and more a change in who showed up at the polls on election day.
It doesn’t exactly settle the issue, but it’s worth drawing people’s attention to a report on party ID trends recently released by the Annenberg Election Survey. According to the report, in about 45,000 interviews of registered voters (RVs) conducted from December, 1999 through January, 2001, Democratic identifiers led Republican identifiers by 33.7 percent to 29.9 percent, a 3.8 point Democratic advantage essentially identical in size to that measured by VNS in the 2000 exit poll.
Annenberg conducted about 68,000 interviews of RVs from October 2003 to mid-Novmber, 2004 and found only a slight diminution in the Democratic party ID advantage to 2.8 points (34.6 percent Democratic to 31.8 percent Republican). That’s quite a different story than the one implied by 2004 NEP exit poll and, given the huge sample sizes in the Annenberg study, is certainly worthy of consideration.

A November 21st Washington Post article notes that Colorado provides lessons for the Dems. The article begins:

When Democratic state chairmen gather in Florida next month to lick their wounds from the Nov. 2 election, their agenda will include a careful study of one bright spot in a generally sorry performance: Colorado, a solidly red state that went almost completely blue this year.
Despite a large Republican advantage in registered voters and the popularity of President Bush, who carried the state easily for the second time, Colorado Democrats picked up a U.S. Senate seat and House seat that had been considered safe for the GOP. They reversed Republican majorities in the state House and Senate to take control of the legislature. And they backed expensive ballot measures that passed by large majorities despite opposition from the GOP.

A Nov 20th New York Times Op-Ed piece notes that environmental measures received wide support in last months election. Here are some excerpts.

Though nobody seemed to notice, Republican and Democratic voters seemed to be of similar minds on one issue this election: the environment. Across the country, in red states and blue states, Americans voted decisively to spend more money for natural areas, neighborhood parks and conservation in their communities. Of 161 conservation ballot measures, 120 – or 75 percent – were approved by voters. Three-and-a-quarter billion dollars were dedicated to land conservation.
…So what’s the story? Simply put, these measures unify Americans. It’s hard to be against new parks and trails, or to disagree with wanting to protect farms and forests from development. What’s more, voters have learned that these measures often provide local solutions to water-quality problems: preserving natural natural lands in watersheds can help protect drinking water sources or reduce storm-water runoff.
…True, this year’s election didn’t turn on environmental issues. But the voters sent a message anyway: whether we’re red or blue, we all have a little bit of green in us.

Between 2000 and 2004, there was a net shift of about 4 million votes in Bush’s direction. On November 9, I posted a brief analysis of where Bush’s vote gains came from by state. I now have had a chance to conduct some analysis of the county-level results from 2004 and they provide some very intriguing insights into the spatial distribution of Bush’s vote gains. These insights challenge the conventional wisdom about where Bush’s vote gains came from and appear to contradict the exit poll’s findings from this year about the spatial distribution of Bush’s gains.
First, my analysis finds that only about a quarter of Bush’s net vote gains from 2000 to 2004 came from the nation’s “ideopolises” (see this piece in Blueprint magazine for an explanation of the term) and three-quarters–the vast majority–came from less technically advanced metro areas and from rural areas.
Second, my analysis shows that Bush made gains across the board (sometimes less, sometimes more than the conventional wisdom has indicated) when you examine counties sorted into 10 categories, going from most urban to most rural. (This analysis uses the rural-urban continuum codes developed by Calvin Beale of the USDA’s Economic Research Service.)
Starting with the most urban counties, those that are central counties of large (1 million or more) metro areas, Bush improved his vote margin by 2.4 percentage points (i.e., he narrowed his margin of loss to about 55-44). His gains in these areas accounted for about 19 percent of his total net vote gain.
In fringe or exurban counties of these large metro areas, Bush improved his winning margin by 6.7 points (to 62-38). But because these exurban areas contain far fewer people than the central counties, Bush received only 13 percent of his vote gains from these counties.
More important to Bush’s vote gains were medium-sized metro areas (250,000 to a million in population), where he improved his winning margin by 3.5 points (about the national average). But because of the large number of people in medium-sized metro areas, Bush received over a quarter (26 percent) of his net vote gain from these counties.
In small metro areas (less than 250,000 population), Bush improved his margin by 2.4 points and received about 8 percent of his net vote gain.
Turning to nonmetro counties, which are typically considered “rural” and which have urban concentrations that range from a high of 50,000 a low of under 2,500, Bush did the best in nonmetro counties that are adjacent to a metro area and have an urban population of between 2,500 and 20,000. In these counties, he improved his margin by 6.4 points and received 15 percent of his overall net vote gain.
In the other five types of rural counties (see the rural-urban continuum codes cited above), Bush improved his margin by from 2.2 to 4.2 points and–putting all these other rural counties together–received 18 percent of his net vote gain.
How do these findings match up with the exit poll findings on Bush’s performance in different types of areas? Not very well at all.
Consider this: the exit polls say that Bush’s margin was compressed both in their “rural areas” category (shrinking by 3 points) and in their other nonmetro category, “cities and towns, 10,000 to 50,000 population” (shrinking by an astonishing 19 points, from a 21 point to a 2 point margin).
It’s very hard to square this with the findings cited above on Bush’s gains in all categories of nonmetro counties, from the most rural to the least.
Or consider this: the exit polls say that Bush improved his margin by an incredible 24 points (going from a 71-26 deficit to 60-39) in “cities of over 500,000” population and improved by an almost as stunning 17 points (going from a 57-40 deficit to dead-even) in “cities and towns, 50,000 to 500,000 in population”. But a glance at the findings above for the different metro categories fails to find anything even remotely consistent with these shifts.
While the exit polls use different categories (cities of different sizes, suburbs, etc.) that are not county-based, it would take a hell of a story to reconcile these findings by pointing to the differences between categories.
So put a big question mark by those exit poll spatial findings. They just don’t square with analysis of the actual votes that were cast and where Bush made his gains.
Much more of this county-based analysis to come! I’m just getting started and will shortly be taking a look at some of the more interesting states in ’04 election.

As the conventional wisdom settles in about the 2004 election, it is, as always, subject to challenge in many important ways.
Alan Abramowitz does some important spadework on this conventional wisdom in his slide show, “God, Guns and Gays: Testing the Conventional Wisdom About the 2004 Election“. I think you’ll enjoy it and find it a source of much useful (and some surprising) data.

There are two lines of analysis that are typically used to justify the claim that the 2004 election result was somehow stolen by the GOP. The first is various bits and pieces of “evidence”–the precincts in Cuyahoga County, Ohio with more votes than registered voters, the counties in Florida (Baker, Holmes) with huge Bush margins but big Democratic registration advantages, etc.–that supposedly indicate vote tampering. I find this evidence profoundly unconvincing and think Farhad Manjoo and others have it basically right: there’s not a lot of there there. Vote tampering does not appear to have happened on the scale necessary to affect this election.
The second line of analysis invokes the now-infamous early releases of the NEP exit poll data, which showed Kerry with a 3 point national lead, solidly ahead in Ohio and also leading in Iowa, Nevada and New Mexico. The reasoning, laid out most clearly in a paper, “The Unexplained Exit Poll Discrepancy“, by Steven Freeman of the University of Pennsylvania, is that exit polls are very accurate surveys and highly unlikely to produce the results referred to above by chance if the real world results truly were +3 Bush, etc. Therefore, the reasoning goes, our measurement of the real world (the actual vote counts) must be wrong and the original exit poll results right. Conclusion: there’s something very funny going on with this election.
But there is a huge problem with this line of reasoning. The exit polls have always drawn samples that are off the real world results and have always had to be corrected (weighted) to eliminate bias, reflect new turnout patterns and, in the end, just flat-out conform to the election results. This year is no different (though it is possible that the magnitude of these corrections has been greater than normal).
Here is my understanding of how the exit poll samples are weighted, based on what I have been able to ferret out so far. (No doubt, I’m not getting it entirely right, but it’s damnably difficult to track down good information about this–exit pollsters have never made much effort to publicly explain and document their methods.)
1. Samples are weighted to correct for oversampling of precincts (for example, exit polls have historically selected minority precincts in some states at higher rates than other precincts) and for non-response bias (exit poll interviewers try to keep track of refusers by sex, race and age).
2. Samples are weighted to correct for changing turnout patterns in the current election, since the sample design is based on past turnout behavior.
3. Samples are, in end, simply weighted to correspond to the actual election results. This is done by first weighting exit poll results in sample precincts to the true precinct results, as they are known, and then weighting the overall sample to the overall election result, once it is known.
At what point are these various weighting procedures performed? That’s difficult to say because of the lack of public documentation of exit pollsters’ methods. But it appears to be the case that weighting of flavors one and two takes place at least partially during the day (and continuously through the day), while the third flavor naturally has to wait until actual election results start to become available.
So where were we in this extremely complicated weighting process when those first +3 Kerry exit polls hit the CNN website? Who knows? (And exit pollsters have not exactly clarified the issue since).
But it’s certainly clear that those data had not yet been weighted (or at least very much) to reflect the actual election outcome (again, part of standard exit poll procedure, not anything peculiar to this year). But how much had they been weighted to reflect the other factors (1. and 2.) mentioned above?
Possibly much of this weighting had already been done. If so, then the rest of the sample correction–that took their data from +3 Kerry to +3 Bush–was done by good old-fashioned weighting to the election outcome. Or perhaps it was some combination of additional weighting for factors 1. and 2. plus weighting to the election outcome.
Who knows? Again, exit pollsters don’t seem to be particularly eager to share this information. Nor do they seem particularly eager to clarify how common it has historically been for exit poll samples at that time on election day to be that far off from the actual election result.
The issue of historical comparisons is an important one. Part of what has led to the brouhaha over this year’s exit poll is people’s lack of knowledge about how exit polls have been conducted in the past.
Consider this. The unweighted–completely unweighted–data from the last four presidential elections before this year are as follows:
1988: Dukakis, 50.3; Bush, 49.7
1992: Clinton, 46; Bush, 33.2
1996: Clinton, 52.2; Dole, 37.5
2000: Gore, 48.5; Bush, 46.2
President Dukakis? Obviously, the unweighted data have always been highly problematic and–interestingly–have always shown a strong Democratic bias. Now these unweighted data from past years do not, admittedly, correspond to where we were in the weighting process on election night this year when the +3 Kerry poll hit the ‘net–those data had presumably already been weighted to some extent to correct for factors 1. and 2.–but it is still food for thought.
Of course, it’s entirely possible that exit poll samples this year, controlling for similar points in the weighting process, were more “off” than in past years. I can’t say at this point and I urge the NEP to make the appropriate historical comparisons available to answer the question. But, even if so, this is hardly evidence of skulduggery in the real world; much more likely it reflects the enormous–and perhaps increasing–difficulties of conducting surveys of this complexity in a rapidly changing country.
Of course, additional inaccuracy in the exit poll samples this year (if true) is not a development completely devoid of implications. It could mean that some of the specific results from the survey are less reliable than in the past. (I personally have my doubts about some of the numbers, like those for Hispanics.) But that’s a far cry from assuming an election has been somehow stolen or tainted.
My advice: calm down and concentrate on what’s really important–beating them next time.