MA-Sen: Jeff’s Election Night Projection Model

I’m as freaked out over Massachusetts as anyone – and I might be plenty angry/despondent/confused/hungover tomorrow morning.

But in the meantime, I’ve been working on a crude projection model using, in part, the “baseline” idea that Crisitunity and DavidNYC have made a part of SSP Election Night Tradition.

The model, as inputs, takes partial results from towns that have reported, and outputs a whole slew of numbers comparing the current situation to the “baseline” numbers.

Explanations below the flip.

So here’s the front end of the model:

http://spreadsheets.google.com…

You’ll see a few things:

• How Coakley and Brown are currently doing.
  

• How Obama was doing in 2008 with the same towns and parts of towns reporting.
  

• How a 2008 “baseline” Democrat on target to win by 1 vote would be doing.
  

• How Kerry was doing in 1996 against Bill Weld with the same towns and parts of towns reporting.
  

• How a 1996 “baseline” Democrat would be doing.
  

You’ll also see comparisons between Coakley’s performance and those of Obama, Kerry, and the two “baseline” Democrats.

Perhaps most significantly, you’ll see the “2010 Projection using 2008” line.

The model compares relative turnout between 2010 and 2008, and the relative performances of Coakley and Obama to project results from towns that have not yet reported.

The “2010 Projection using 1996” line does the same, except with 1996 data.

I’m not claiming this model is perfect. In fact, it’s pretty damn bad. I can think of a few glaring weaknesses:

• The fundamental problem of the ‘baseline’ idea: it assumes that every town will swing uniformly.
  

• Disparate turnout: this model compares turnout in aggregate, instead of at the town level. This may lead to an overestimation of turnout in areas with relatively low turnout (compared to 2008 and 1996) and the reverse in areas with relatively high turnout. This may potentially bias the projections in Coakley’s favor.
  

• Assumption of town uniformity: the model assumes that each town votes uniformly the same way, but…Jamaica Plain and Southie are not going to vote the same way, very simply. If a relatively Brown-friendly area of a town reports first, this will bias the projection in his favor. The reverse is true if a Coakley-friendly area reports first.
  

Incidentally, here are the blood and guts of the model: http://spreadsheets.google.com…

Update: In columns AS, AT, BB, and BC, you can see baselines for every town for both 2008 and 1996.

I threw in some junk results to test it, and so far I didn’t detect any coding errors. I don’t pretend that I’m better than the Associated Press – but I just want to have an idea of where we are at each point of the results phase.

Hopefully I’ll get a chance to keep this updated as results stream in tonight.

Lastly, if you live in Massachusetts, are reading this, and you haven’t voted (assuming you’re a citizen, not a convicted felon, etc..), what the hell’s wrong with you?!

Here’s hoping Coakley pulls this off.