My usual technique for the Texas presidential primary polls is to compose my sample of voters with a history of voting in primaries, and then limiting the responses to those who say they will vote in the party's primary that they have previously voted in. It's a tight screen to reflect low turnout. The general election will see five times the number of voters as both primaries combined.
As we get to the point that more voters are tuning in, I decided to do a comparison poll of the population that does not vote in primaries, but does vote in the general elction. In the last presidential election, 7.4 million Texans voted, compared to about 840 thousand in the Democratic primary and 690 thousand in the GOP primary. One might expect the GOP to have a larger primary turnout than the Democrats since they have dominated recent general elections, but that's not the case. Only 15% of Bush voters had voted in the primary while 30% of Kerry voters had. This means that this non-primary voting population is tilted heavily in favor of Bush voters since they are a larger percentage of a larger number. Sixty-five percent of this population voted for Bush and only thirty-five percent voted for Kerry.
The line of questioning was slightly different than the primary polls. Rather than ask which party and then listing that party's candidates, I asked if the respondent had a favorite candidate, then which party that candidate belonged to and finally listed that party's candidates. I did this to focus on candidate appeal rather than party allegiance. After the candidate questions, I asked the same issue questions that were in the primary polls. General election voters, being less interested in politics, were less likely to take the poll, so I only got 375 responses, despite making 50% more calls than I usually do for the primary polls. Margin of error is larger, especially after splitting the sample three ways into both parties and 'no favorite'.
Of the respondents with a favorite, the Democrats did better than the 2004 number would indicate. Rather than a 65/35% split, Republicans only led 55/45%. It's likely that the fluctuating GOP race caused a few GOP voters to be undecided rather than picking a favorite, so that may skew the numbers a bit. However, when the numbers are broken down by gender, a clear split is observed. Men vote Republican 60/40%, not far from expected, but women vote almost exactly 50/50%. There has been speculation that Clinton may be attracting Republican women, but I can't confirm that from this data. Clinton received 33% of all women's support while the GOP frontrunner Giuliani received only 16%, followed by Obama at 12% and Romney at 10%. Among men, Giuliani received 23% to Clinton's 19%, Obama's 12% and Romney's 11%. It shound be noted that both Clinton and Giuliani do better among women within their respective party primary voters. Thompson and Huckabee trail McCain among both sexes.
Ron Paul supporters have suggested that he is attracting Democratic crossover voters, but there is no evidence in these results. He received 3% of the total vote. Among only those who picked a GOP candidate as their favorite, his percentage is almost identical to his support among GOP primary voters.
The Issues
I asked a series of issue questions, first asking respondents to rate six issues on a scale of 1-9 on how important it was for the next president to address, followed a question on Iraq troop levels. For the Democratic primary poll I had included some initial analysis using averages, but the distribution of ratings is not really well suited to averages. In that poll. Barack Obama supporters were the most likely to support a troop increase, but they were also the most likely to support quick withdrawal. An average value loses this information. Instead, I've put together a series of bubble graphs that graphically represent the percentages of each rating. I've combined the data from different polls into a single graph to compare the ratings of each party's primary voters with the ratings of general election voters by party preference as well as undecided. As noted above, the undecided voters are heavily weighted with past Bush voters. A second series of graphs compares the ratings for the supporters of the first and second place candidates for each party. Since the Republicans have a tie for second, there are total of five candidates. This series includes candidate supporters from both the primary and general election polls, though most are primary voters.
Many graphs after the break - dialup beware! |