Friday, June 20, 2008

parsing the primary

So, I suppose it's time to parse the primary data a little bit more. The problem with looking too deeply at the results from April is that there isn't anything to really compare the voting patterns to. The obvious comparison to the spring 2008 primary would be results from the spring 2004 primary which was the last presidential cycle. The problem of course is that in 2004, the primary had been locked up by the time the Pennsylvania primary came around. Contrast that with 2008, and no matter whether you thought the electoral results were already won by BO, the media played up the whole election as the biggest presidential primary in Pennsylvania ever Given that, any comparison of the two elections (2004 vs. 2008) is by definition an overinterpretation. Nonetheless, maybe there is something to glean from comparing the two elections.

As a starting point.... broken down by age, here is the graph I come up with for the number of voters in Allegheny county by single year of age in 2004 and 2008. Blue is 2008, Redish is 2004.

On the surface this already seems surprising because it makes it look as if the big bulge in voting was not just an increase in younger voters which is all the media really talked about. There was clearly a big, if not bigger, bump in the number of people who came out to vote between the ages of around 28 to 63.

Of course that is in the raw numbers, if you look at the percentage increase in the number of voters in 2008 compared to 2004, again by single year of age in Allegheny County I get this graph:

That shows the big bump up in younger, in fact of the youngest voters. Of course with percentages you get small number distortions when your denominators are so small, which is pretty much the case with voters under the age of 30, if not 40, in local elections. So interpret that graph as you will.

Is there something to make sense of here given all the caveats? Counting the number of voters does not allow you really figure out who was more likely to come out on election day. Especially around here, local demographics are so atypical that you get very different number of people in different age cohorts. So I have matched the voter turnout numbers used above to estimates I have on the population in the county in 2004 and in 2008.

What that gives me is the following graph which is the percentage of total population that actually voted, again by single year of age for Allegheny County in the two elections.

With that... I am not sure what it says in the end. Clearly the small number of younger voters in 2004 were given a big percentage bump up in 2008. I think it's much more interesting how many of the middle and older age voters came out. Nobody talked about that much, either during the run-up to the election nor even since. That spike in increased participation around age 27 in the 2008 primary is interesting and deserves some inspection. If I had to sum it up, I would say that the big increase in attention given the 2008 primary made the voter participation look much more like what is typical of a general election around here. What I think is probably the biggest question, and something I will look at when I have time, is whether the additional folks who voted 2008 were more likely to be the people who are likely to have come out in the general election anyway. Lots of younger voters come out in general elections even if they never make it to a primary. It makes a big difference if the folks who came out in the 2008 primary where those folks, or a lot of people really new to the political process and in the fall we can expect them to show up along with the folks who typically skip the primary, but make it to the general. The answer to that question across the country I bet will determine who wins in the fall is my call.


Anonymous Tom Lane said...

Very interesting graph. Are all the curves measured according to age at the time of the election? In other words, if I want to compare the behavior of the same group of voters, should I be looking at the graphs as posted or should I be shifting 2008 four years to the left? Whichever way it is, it'd be useful to show the other way too.

Saturday, June 21, 2008 9:09:00 PM  
Blogger C. Briem said...

the horizontal are ages and specifically ages at time of the vote. I think you are suggesting using birth year for the horizontal axis which I probably could. That has issues as well since its normal for an individual's voting behavior to change over time no matter. If I get time I will redo it, but you can just think of the first graph and shift the one time series left 4 years to do much the same thing.

Monday, June 23, 2008 8:41:00 AM  

Post a Comment

Links to this post:

Create a Link

<< Home