Brian Arbour sends along some analysis of early voting in the Texas primary. The numbers come from the Texas Secretary of State, which is releasing data for the 15 largest counties in Texas. Brian writes:
bq. I measured the turnout increase from 2004 against demographic characteristics that have differentiated the two Democratic candidates to this point–% Hispanic, % Black, % Bachelor Degree, and Median Income. The numbers below measure the number of voters through February 25, 2008 (7 days of early voting) to those of February 29, 2004 (also 7 days of early voting).
bq. Early voting numbers show that turnout is up strongly in counties that have demographic characteristics that favor Barack Obama. Turnout is up only modestly in counties whose demographics favor Hillary Clinton.
Below the fold are the scatterplols of the percentage increase in turnout by the various demographic characteristics, as well as some necessary caveats.
Brian offers these four caveats:
bq. There is no evidence yet that these “new early voters” are really “new voters.” They well could be those who usually vote on election day, who have decided this year to vote early.
bq. Each of these three variables is correlated with each other. The three South Texas (and thus, most likely Clinton friendly) counties—Hidalgo, El Paso, and Nueces—are always at one end of the scale. These are the counties that have the smallest increases. The turnout measure could just be measuring something about political culture that makes voters willing to vote early rather on election day.
bq. We don’t know who or where in each county people are voting early. While these data are consistent with an Obama surge, it could well be that turnout is high in areas in these counties where Clinton will do well. Or that demographic groups that favor Clinton—such as middle income women—are the ones creating the increase.
bq. We have no data yet on who these people are voting for. Thus, I’m not measuring the really interesting question. We’ll still have to wait a week to get an answer.
Brian concludes:
bq. These numbers don’t prove anything yet. But, they would like these numbers better at Obama headquarters than at Clinton headquarters.