I had posted this earlier, but it is relevant to Chicagoland.
The concept of Cumulative Vote Share is that as you add in precincts, one-by-one, you can calculate the share of the vote at that point in the count. This is quite familiar from election night. The order the votes are accumulated can be significant. In Pennsylvania, the votes from Philadelphia are reported first, and the initial vote share is extremely Democratic, the rest of the night the Republican share will climb.
Imagine that all precincts used paper ballots, and the election clerks in each precinct were equally efficient, such that the could count one ballot in some period of time. Let's assume 6 seconds per ballot, or 10 ballots per minute, then if there were 405 ballots cast, it would take 40m30s to count the ballots, at which time they would call the results in. A precinct with 505 ballots cast would take 10 minutes later to count.
At the county board of elections, the precinct results would be added in and the cumulative vote share reported. This is the blue line in the chart. So it might be reported that with 20% of the vote counted, Rauner has 38.7%.
Conspiracy theorists believe the fact that Republican CVS climbs as larger precincts are added in proves vote fraud since it is "common knowledge" that larger precincts are in Democratic-leaning cities. This is not true. For example, the largest Cook County precinct is in Hanover, which is as far as you can get from Chicago and not be Kane County.
They also overlook that turnout in most elections, and particularly non-presidential elections are in lower economic status areas, which typically are more Democratic. With fewer votes cast, they could be counted sooner, and added into the cumulative vote share. Places like Evanston and Oak Park have relatively high turnout, and are strongly Democratic, so the turnout differential is not strictly speaking partisan. But turnout in Cicero was horrible, and it was very Democratic.
But imagine that the county board of elections called the precincts. They might assume uniform turnout, and call the precincts in order of registered voters. With fewer voters, there will be less votes to count, so they will likely be done by the time the call is made. If precincts are ordered by number of registered voters, the red curve results.
It is quite level, with just a minute Republican trend in larger precincts. There may be a delay in adjusting precincts. If an area is declining in population, there may be resistance to closing familiar polling places. If an area is increasing in population, or at least remaining steady, they may not want to create additional polling places. Areas that have lower turnout, are also likely to have lower registration.
This chart shows the relationship between Rauner%, and Turnout%. The bubbles are colored based on Rauner%, and bubble size is based on registration.
It may be easier to see this chart directly in the spreadsheet.
Cook County Rauner 2014There is a bit of optical illusion. The countywide turnout was 48.5%, and Proviso and Worth are just below that, with Thornton, Bloom, Bremen, and Niles a couple of points lower. There is a lot of overlap in the blue cluster around 60% Rauner (Ormond is hidden behind Palatine, for example), such that some of the lower turnout areas such as Schaumberg may have more visual weight, with their turnout perceived as being at the lower end of the circle, rather than as its center.
Overall, the more Republican townships have a bit higher turnout than most of the more Democratic townships, and there are some very low performing Democratic areas. I would interpret the turnout in Rich, Thornton, and Proviso as indicating that black turnout is less tied to economic status than that for white and Hispanic voters.