The good news is that the complete set of all results, by precinct, for every county and city with a separate board of elections, was posted at the State Board of Elections on Dec 1
. Every jurisdiction is in CSV format so it ports easily into Excel or any other spreadsheet or database. The difficulty is that there are 112 separate files and no cross correlations. Individual ballot lines are each treated separately in the files.
To answer a question like how did Durbin do by CD or Rauner by Legislative District requires finding all the precincts in the smaller district then finding the same precincts for the other race of interest. Its quite time consuming in a simple spreadsheet like Excel. It's a more reasonable task with a relational database like Access.
In any case, the SBE isn't going to create the correlations. The number of possible correlations between offices is enormous. Who is to say that US Senate by CD is the most important one to do? Since DKE did this for the 2012 data, I won't be surprised if they do the same for 2014. Of course they get to choose the correlations in that case.
Thanks, muon. I'm doing IL next and these links have been pretty helpful. They've been easier than going to each county/municpal site for results.
The most time-consuming part was Chicago, where there were about 130 split voting districts that I had to assign. I don't have time (nor really the patience, even by my standards) to split them individually, so I gave them to whichever CD they cast the most votes in; 'rounding to the nearest precinct' is how I usually deal with those situations. Thankfully, the split precincts were pretty lopsidedly split towards one CD, so it was obvious where to assign them. Yes, it leaves some margin for error, but most were heavily Dem anyway and we're talking about 130 out of 2,000+ precincts.
I'm working with some of the data-crunchers from DKE. I'll probably be the only one doing an exhaustive set of Senate/Gov races by CD, at least for the next few months. DKE seems more focused on breaking results down by LD.