Read the full story here.
2012 Voter File:
The initial voter data for 2012 obtained by The Reader was not useable for this initial analysis because it did not have polling location for each voter listed. It did have a large amount of voting data that would prove useful later.
The Reader obtained a new file with voters and polling places linked below. This file contained 314,969 voters and the date suggests it is from a polling place mailing list as of 2012-03-05. This data has all voters and their polling place. Shape files are also available upon request.
2008 Voter File:
In order to do a comparison of the voter’s distance from their polling places, The Reader obtained a 2008 voter file from just prior to the 2008 general election with linked polling places. This file contained 315,257 voters and is linked below.
Both files are large and the Election Commissioner requests that voter files not be used for commercial purposes. If it’s your intention to contribute to this analysis in the public interest, please email us the details of your interest to share access to these files.
Note: All GIS processes were conducted in ArcMap 10.
All GIS data from the Election Commissioner included 2012 polling places and precincts along with 2010 polling places and precincts. Deciding to use the 2008 voter file for comparison to 2012, a new 2008 polling place file was created based on the 2008 voter file. A point file was created by using the 2008 voter data and running a pivot table on polling places and address. With a list of all polling locations in 2008, we were able to geocode (place on the map based on address). A new polling places file was created because the polling places file for 2010 didn’t match the voter file for 2008. Some polling places changed between 2008 and 2010.
Data Creation Process:
The first step was to geocode all of the voters in the two file (2008 and 2012) to generate a distance to polling place value and determine what Census Tract each voter was in to give them demographic data from the census. We used an address locator created from a 2011 street file and parcel file for Douglas County. Douglas County GIS has wonderful data available here.
After running the geocoding process in ArcMap 10 we were able to match the voters to the rates below:
Total in Voter File 315,257
Total Matched 312,169
Percent In This Analysis 99.02%
Total in Voter File 314,969
Total Matched 312,497 (312,499 were actually matched and two were later remove)
Percent In This Analysis 99.22%
Some voters we not matched because their address could not be found. While there are likely some remaining voters who have been incorrectly matched in both years, the majority of the address were matched with a 100% match rate. The mismatched voters will not have a dramatic impact on the results.
Straight Line Distance
The next step in the process was to give each voter a distance to their polling place. With no quick and easy way to give each voter a distance based on their travel by road, a straight line distance method was used. In this method, each polling place and the voters connected to that polling place were placed on the map by themselves. This consisted of one point for the polling place and a point for each voter connected to it. These two data sets were then spatially joined so that “Each point will be given all of the attributes of the point in the layer being joined that is closest to it, and a distance field showing how close that point is (in the units of the target layer).” In order to speed this process up, multiple polling places were done at once as long as they were far enough apart as to not allow a voter to be joined to the wrong polling place. These exported files contained only a few polling places and their voters with a distance. Once each polling place was done for the given year, they were joined back together to form a new voter file with all matched voters and a distance to their polling place.
This process was done using the matched voters in 2012 and the 2012 polling place shapefile provided for the 2012 data. The 2008 data was done using the matched voters in 2008 and the 2008 polling place shapefile that we created. The resulting data, if viewed for all voters at one polling place in ArcMap 10 would look like the screen shot below if shown with the shortest distance in green and the longest distance in red.
The final step in the setup of these data sets was a process of joining census data to each voter. Because we did not have individual demographic data for each voter, we used Census data from the 2006-2010 American Community Survey (ACS). We downloaded a number of tables from the American FactFinder website http://factfinder2.census.gov/faces/nav/jsf/pages/index.xhtml that pertained to the demographics that we wanted to use to analyze change in distance. These tables include data on poverty, race/ethnicity, employment, education and median household income. These data tables were condensed and combined into one file that was then joined to the census tract GIS layer in ArcMap 10. Once joined to the census tract GIS layer, voters could be joined to census tracts and, in turn, joined to the data pulled from the ACS. This process, similar to joining the voter points to polling points, uses voter points and census tract boundaries to join voters to the census tract that “it falls inside”. In other words, the voter was given all attributes of the ACS based on what census tract that they fell inside of. Below are the fields of census data added:
Below is an example of what that might look like for one polling place. All of the voters from the Beadle Middle School are in census tract 74.60 and thus were give the poverty percentage, minority percentage and median household income of that census tract according to the 2006-2010 ACS.
This process was done for the 2008 and 2012 voter files using the 2006-2010 ACS data. Because we wanted to maintain the same areas in the distance comparison between years, it was important that we used the same ACS data for both voter files. It was initially suggested that we should be comparing the voter data by precinct between 2008 and 2012, however, because these boundaries changed between the years, a comparison would between them would not work. Using the census method, voters counted in >50% minority census tracts in 2008 and 2012 should be people in the same geographic area.
Once these modified voter files were finished, the data was exported into excel for analysis. These two files are available under the same guidelines as above in regards to voter files.
Pivot Table Comparison
The primary method of analysis is the pivot table feature in excel. The pivot table tool allows you to summarize data by defined columns and rows. In all cases my column was the average distance of voters to polling place as defined by my row, or, the defined area based on census tracts of specific median income, minority percentage, poverty percentage, education percentage, employment percentage or county quadrant. Using the table below as an example, the pivot table allowed us to calculate the average distance of every voter from their polling place (as defined by the straight line distance calculation defined above) in all census tracts where less than <10% of people age 25 and over (according to the 2006-2010 ACS) had a bachelors degree. We also did this for the remaining data breaks as they were initially defined by for minority percentages ( < 10%, >= 10% and < 20%, >= 20% and < 50% & >= 50%).
These analyses were conducted for both the 2008 and 2012 voter files and then compared between years. Because the same ACS data was joined to voters in each year, the same areas are being compared between years just with slightly different voters and changes in polling place distance. This analysis can be seen in the file named Comp_2008vs2012. Any additional census data (or commute time data) could be joined to the voter files by census tracts and then analyzed. Additionally, the data could be broken down in any number of ways that has not been previously done by us.
The Analysis can be summed in this document: VoterComp_2008vs2012
And is narrated below:
Voters in the northeast and southeast quadrant of the county as divided roughly by 72nd and Dodge streets experienced a 43% and 45% increase respectively in the straight line distance to their polling place from 2008 to 2012. Voters in the northwest and southwest quadrants experienced a 19% and 21% increase in the straight line distance to their polling place respectively from 2008 to 2012. The northeast and southeast quadrants made up about 35% of the total voters in 2012.
Voters in Census tracts with median household income between $0 and $24,999 experienced a 30% increase in the straight line distance to their polling place from 2008 to 2012. Voters in Census tracts with median household income between $25,000 and $49,999 experienced a 53% increase in the straight line distance to their polling place from 2008 to 2012. Voters in Census tracts with median household income between $50.000 and $74,999 and those over $75,000 experienced a 14% and 17% increase in the straight line distance to their polling place from 2008 to 2012 respectively. Voters in census tracts with incomes below $50,000 made up about 43% of the total voters in 2012.
Voters in Census tracts with less than 10% of the population over the age of 24 with a bachelors degree or greater experienced a 68% increase in the straight line distance to their polling place from 2008 to 2012. Voters living in Census tracts where the population over the age of 25 with a bachelors degree or greater was between 10% and 20%, 20% and 50% and over 50% experienced a 35%, 32% and 12% increase in the straight line distance to their polling place from 2008 to 2012 respectively.
Voters in Census tracts with 20% to 50% and over 50% minority populations experienced a 47% and 53% increase in the straight line distance to their polling place from 2008 to 2012 respectively. Voters in Census tracts with less than 10% and between 10% and 20% minority populations only experienced a 27% and 9% increase in the straight line distance to their polling place from 2008 to 2012 respectively. Voters in census tracts with over 20% minority populations made up about 40% of the total voters in 2012.
Voters who were Republican experienced a 24% increase in the straight line distance to their polling place from 2008 to 2012 while Voters who were Democrat and Libertarian experienced a 32% and 41% increase in the straight line distance to their polling place from 2008 to 2012 respectively.
Voters in Census tracts with less than 10% of the population over the age of 17 in poverty experienced a 17% increase in the straight line distance to their polling place from 2008 to 2012. Voters in Census tracts with 10% to 20% and greater than 20% of the population over the age of 18 in poverty experienced a 45% and 62% increase in the straight line distance to their polling place from 2008 to 2012 respectively.
Voters in Census tracts with less than 10% of the population over the age of 15 unemployed experienced a 23% increase in the straight line distance to their polling place from 2008 to 2012. Voters in Census tracts with 10% to 20% and greater than 20% of the population over the age of 15 unemployed experienced a 61% and 39% increase in the straight line distance to their polling place from 2008 to 2012 respectively.