Sunday, September 30, 2012

Iowa 2012 election analysis

In 2004, Republicans had just over a 4,000 edge in 'active' registered voters, and won by a similar margin of 10,000 votes. In 2008, much like Nevada, Democrats surged to a huge edge of 106,000 'active' registered voters. This lead to a 146,000 blow-out win for Obama in 2008.

So how is 2012 shaping up? Republicans currently have a 12,000 edge in active registered voters, bigger than they did when Bush won in 2004.

In spite of the Iowa poll averages, Romney looks to have the advantage this year.

I notice that Rasmussen's most recent poll shows a Romney edge of 3 points. This looks about right. Not from confirmation bias (looking for a poll that agrees with my opinion), but because that result agrees with the objective evidence from non-polling sources.

Nevada 2012 election analysis

In 2004, Nevada Republicans had a small registration advantage: 434,239 to 429,808. By 2008, the Democrats had built a huge registration advantage: 625,134 to 513,629. These changes were reflected in the presidential voting, turning a small Republican edge of 21,000 votes in '04 to a blowout Democratic win of over 120,000 in '08.

Nevada has an interesting system of differentiating between 'active' registered voters and 'inactive' registered voters. Essentially, 'inactive' registered voters are those in which records indicate that they may have moved. They can still vote if they can show that they are legally able to do so, but I would think that such voters would be significantly less likely (though to what extent I don't know) to actually vote. I believe that, like increasing (or decreasing) voter registration, it is a proxy for enthusiasm.

In 2008 almost all of the Democrats' advantage was in 'active' voters, with about 101,000 of their advantage coming from them and only about a 10,000 edge in 'inactive' voters. This year, however, while the total Democratic advantage is only slightly smaller, their advantage in 'active' registered voters has been almost halved to 55,700. About half of it is now in 'inactive' voters. There are 127,000 'inactive' Democratic registered voters compared to only 76,000 'inactive' Republican voters.

One final issue is the growth of the Independent American Party. When you look at their site, it is clear that they are essentially Tea Partiers. If you look at only the active registered voters as of August 2012, Republicans, when combined with the IAP, are essentially tied with the Democrats.

Nevada looks to me to be in play as of the August data. Obama was clearly in much worse shape than he was in 2008 (though of course he can afford to lose quite a bit, considering the blow-out double digit win).

Saturday, September 29, 2012

Ohio 2012 election analysis

As I stated in my earlier post on Florida, I am attempting to do a 'sanity check' on the state level polls coming out recently. A surprisingly large number of them seem to have results outside the realm of probability, so I decided to see if I could find predictive data outside of polling to see what it might tell us.

What I have stumbled upon is a comparison of historic voting patterns among various large segments of registered voters, with current voter registration. With Florida, I compared historic voting patterns of registered Democrats with that of registered Republicans and using those to forecast a vote using current registration levels.

Unfortunately, Ohio does not post these types of data, so I have to use a slightly different method - comparing the voting history of various counties with the current number of registered voters in those counties.  Certain counties have historically been much more Democratic oriented and others have been more dominated by Republican voters. To the extent that this pattern holds, you can estimate which party is better situated this year than in previous elections. This is somewhat more labor intensive than was the case with Florida, but I think that it is similarly helpful.

There are 88 counties, so I will be focusing on some of the largest counties to see if we can see evidence for which way this election will be going compared to 2008.

Obama won Ohio by 262,000. Cuyahoga county by itself provided Obama with more than 258,000 of that advantage. So what is going on there this year?  This is the total registered voters as of 9/7/12:

Democrats  -   346,642     38%
Republicans -  126,280     14%
Other          -   431,372     48%
Total               904,294

Looks pretty dreadful for Republicans, right? Well ... in comparison to 2008, its actually a huge improvement. Voter registration figures for November 2008 show:

Democrats  -       395,514     36%
Republicans -        91,416       8%
Other          -       625,527     56%
Total           -    1,112,507

In their most important county, Democrats have suffered a net loss of over 83,000 from their advantage in registered voters, from an advantage of 304 thousand in 2008 to an advantage of 220 thousand earlier this month.

Does this indicate a state that is going to be MORE favorable to Obama this year vs 2008 as many polls are showing, or one that is going to be considerably LESS favorable?

I had planned to do the same comparison for other counties (Franklin and Lucas counties in particular), but I can't find similar data on their websites. However, there is a recent Fox News story which details the declining registration in Democratic strongholds in Ohio including not just Cuyahoga but Franklin and Hamilton. Interestingly, they mention that this decline in Democrat standing is holding over at least 8 'key' swing states including Iowa, Florida, and New Hampshire.

This simply makes sense - Obama has been a disaster, and 2008 was a Democratic wave. How on Earth could he improve substantially in swing states as many polls are stating? In the real world, Republicans are increasing their number of registered voters in these swing states, while registered Democrats are declining.

Who are you going to believe, common sense plus data on literally millions of American voters actual actions, or are you going to believe divergent polls based on merely thousands of respondents (with highly debatable samples)?

UPDATE:  As I said, I wasn't able to get the same information from other counties, but I was able to download a CSV file from Hamilton County with the current registration figures. In 2008, Obama won Hamilton County by 7 %, nearly 30,000 votes - 225,213 to 195,530. Hamilton County currently shows a total of just of 554,000 registered voters. What is really interesting is that right now, by sorting the 'party code' column, I see that registered Republicans outnumber registered Democrats by more than 2 to 1 (though they are both swamped by officially unaffiliated voters). There appear to be a total of just 43,932 registered Democrats in the county vs 97,419 registered Republicans.

UPDATE 2: I have now downloaded a voter registration file from Franklin County, and it's more of the same. Keep in mind that it is of limited utility since I can't compare it to registration data as of 2008, and the unaffiliated voters make up the biggest percentage by far. With those caveats in mind, there are over 791,000 registered voters in Franklin County. Of those, 102,605 are registered as Republicans, while there are only 67,915 registered Democrats (leaving over 620,000 to the smaller parties and 'non affiliated'). This is in a county that Barack Obama won by 116,000 votes (over 20 percentage points 59.6 to McCain's 38.9).

Florida 2012 election analysis

The polls are all over the place this year, and something like half simply make no sense whatsoever. Obama is now a known factor, and has been a disaster. Romney, whatever his faults, is an improvement over McCain. Yet Obama is supposed to be doing twice as well vs Romney as he did vs McCain? I'm sorry, but that is inconceivable, and I DO know what that word means. How can this be checked without reference to polls? Using one poll (say Rasmussen) to 'prove' that another poll (say, Fox or MSNBC) is wrong doesn't make any sense. All you are doing is showing a textbook case of confirmation bias. The only way to judge is if you are somehow able to look at the issue by looking at data that is not artificially weighted, adjusted, etc. 

Where can you get such information, you ask? It depends on the state, but there is a surprisingly large amount of raw data available that I believe can be used as a sanity check. By this I am referring to votes in actual elections (rather than estimated "likely" voters), and I am referring to party preference actually reported to the state in voter registration forms. I am also referring to the TOTAL registered voter population, not a small sample taken from it, which may or, (much more likely) may NOT be representative.

I've chosen to look at Florida first, because it is not only an extremely important state in terms of electoral votes, but because it furnishes a long record of voter registration figures, including partisan breakdowns. One of the arguments that Republicans have been making about the polls is that they seem to be ignoring the evidence from the 2010 elections. The pollsters, by and large, respond that presidential years are very different. I think that each side overstates their case a bit, but to be as fair as possible to the polls, I will only consider presidential years in this post.

 In 2004, Florida had 10,301,290 registered voters:

Democrats -      4,261,249   41.4%
Republicans -    3,892,492   37.8%
Other  -             2,147,549    20.8%

So, notice, Democrats had over a 3 1/2 point advantage in registration, nearly 370 thousand registered voters.

In the 2004 election, Bush won by over 380 thousand votes. The actual vote totals were:

Bush -     3,964,522
Kerry -    3,583,544

Notice that, in comparison with registered voters, Bush's vote totals were slightly greater than the number of registered Republicans (102%), while Kerry's were significantly lower (84%) than the number of registered Democrats. In 2004, either Republicans were much more motivated to vote, or they were winning a huge percentage of 'Other' affiliations, or some combination of both.

This all changed for 2008, right? Well ... to some extent, but not nearly as much as I would have thought. First, and most obvious change was in the huge surge in Democratic voter registration, and the big advantage they derived from it:

Democrats  -        4,722,076   42.0%
Republicans -       4,064,301   36.1%
Other  -                2,461,257   21.9%

So in 2008, Democrats now have more than a 650 thousand registration advantage, nearly 6 points. How much do they win by? Considerably less:

Obama  -     4,282,367
McCain -     4,046,219

What jumps out at me is that McCain, like Bush in 2004, pulled in essentially the same number of voters as registered Republicans (99.8%), while Obama, like Kerry, pulled in a significantly lower percentage of voters/registered Democrats. He did improve over Kerry's figure (91% vs 84%), but this was a 'wave' election! In spite of that, McCain managed to cut over 400,000 off of his registered voter disadvantage!

So where does this leave us for the upcoming election? In terms of registered voters, much better shape than 2008*:

Democrats  -      4,581,056   40.0%
Republicans -     4,137,890   36.1%
Other   -             2,727,594   23.8%

Republicans have maintained their 2008 percentage, while Democrats have dropped 2 full percentage points. If Romney does as well as McCain did in garnering as many votes as registered Republicans, Obama has to do as well as he did in 2008 to get it even. If Obama does even slightly worse, say, 89% of registered Democrats, Romney wins.  If Obama some how, some way manages to maintain his 91% from 2008 and Romney makes even a slight improvement on McCain (i.e. does as well as Bush did in 2004), Romney wins.

As the country stands, I simply can't imagine Obama doing BETTER than he did in terms of Democratic turnout in 2008. I also don't see how Republicans don't turn out better for Romney than they did for McCain. Especially in Florida. The polls showing Obama out to big leads there aren't worth the pixels they are printed on. Best case scenario for him is a close nail biter. It's more likely that Romney wins it going away - he's going to have 10 or 20 times the resources McCain did.

Look at every other election metric besides: Republican governor. Republicans control the State Senate better than 2-1. They control the State House better than 2-1. The US House delegation is Republican at better than 3-1. The US Senate is the only place the parties are even, and that seat is in jeopardy.

Hopefully tomorrow I'll be able to look at Ohio. It is much harder than Florida because they don't have nearly the same availability of records online, and they don't have partisan breakdown. However, I think I can do something similar with county registration and voting records.

*The figures from the Florida Division of elections that I downloaded were dated July 16th of this year.

UPDATE:  One more point. 2008 is far and wide referred to as a Democratic 'wave' election. This is reflected in Florida by not only a huge surge in registrations, but by a large increase in Democratic participation. It makes sense that these two are linked - when an unusual number of Democrats are motivated to go out and register for a particular party, shouldn't they also be more motivated to actually go out and vote? Sure. However, this year the Democrats are faced with the opposite problem. As they are LESS motivated to register this year, doesn't it now follow that this lack of motivation also applies to registered Democrats? As their registration advantage falls back to the historical norm, should not their turnout likewise fall back to normal?