Friday, November 05, 2004

Today's Slate features an article arguing that terrorism, not moral values, is the real reason for the Democrats' loss.

I may be missing something, but I just don't understand the argument:

Much has been made of the fact that "moral values" topped the list of voters' concerns, mentioned by more than a fifth (22 percent) of all exit-poll respondents as the "most important issue" of the election...Voters who cited moral issues as most important did give their votes overwhelmingly to Bush (80 percent to 18 percent)...

Terrorism was cited by 19 percent of voters as the most important issue, and these citizens gave their votes to the president by an even larger margin than morality voters: 86 percent for Bush, 14 percent for Kerry...

These differences hold up at the state level even when each state's past Bush vote is taken into account. When you control for that variable, a 10-point increase in the percentage of voters citing terrorism as the most important problem translates into a 3-point Bush gain. A 10-point increase in morality voters, on the other hand, has no effect.
I don't understand: voters who cite "terrorism" and "moral values" go for Bush at roughly similar margins. How exactly does an increase in "terrorism" voters translate into a Bush gain while the same increase in "moral values" voters does not? What variable is he controlling for?


At 5:05 PM, Blogger Julia said...

I think the point is that the people who cited morality as their top concern would have voted for Bush even if 9/11 had never happened, while the 19% who cited terrorism represent "new" Bush voters--or something. But don't quote me on that.

At 5:22 PM, Blogger alex said...

You think so? Its an argument that makes sense, I'm just trying to understand if its the argument he makes here.

Mostly, I'm mystified by the following sentence: "A 10-point increase in morality voters, on the other hand, has no effect."
That makes no sense to me: 10% more voters at the polls who say that morality is the most important issue, breaking for Bush by an 80-20 margin, should make a difference. And its not a theoretical issue.

At 6:32 PM, Blogger Julia said...

When I read the article as a whole, this did indeed seem to be what Freedman is saying. I think that the 10% increase might be attributable to voters in states with gay marriage amendments on the ballot, where issues of morality might have been more at the forefront that in other states. Nobody I know who voted for Bush did so for reasons directly related to moral issues (not that this is necessarily a representative sample by any means).

At 6:47 PM, Blogger alex said...

Hmmmm. I really don't think that this is what he is saying. I mean, he says "When you control for that variable, a 10-point increase..."

Of course, real elections don't control for any variables, so he must be talking about some regression that he is running. That is, there was no 10% increase in real life - its just that "a 10% increase in X leads to a 3% increase in Y" is his way of summarizing the correlation coefficient he got.

I just cant understand what regression he is running since his entire description of it is one vague sentence.


Post a Comment

<< Home