There are three classes that voters can be registered as: “A” (for “active”), “I” (for “inactive”), or “S” (for “in suspense”). We have sixteen Kansas voter files from June 11th, 2015 to June 6th, 2017. I looked at each voter’s registrant class at the first time point. I then looked at the fifteen later versions of the voter file we have. If they did not appear in any of these, I counted them as “dropped.” Thus, these calculations include if someone was dropped and then registered again at a later date.

First are some raw numbers. The bolded letters in the first column denotes the three classes (i.e, “a”, “i”, or “s”). The next two columns tell us how many people were not dropped (“FALSE”) and were dropped (“TRUE”) by June of 2017. Each cell contains three values: N (the total number), Row(%) (the percentage of people of that class that were or were not dropped), and Column(%) (the percentage of people that were or were not dropped by class).

For example, there were 1,742,076 people on the voter file as of June 11th, 2015 (the very bottom right cell). While “in suspense” (code “s”) voters made up 1.96% of the file, they represented 13.84% of those dropped from the records. Additionally, we can see that 66.07% of voters “in suspense” in June of 2015 were dropped from the file within the two-year period.

 
Status at 2015-06-11		 Dropped by 2017-06-06?
FALSE		  
TRUE		  
Total
a
N
Row(%)
Column(%)		  
1485864
95.90%
94.10%		  
63511
4.10%
38.97%		  
1549375
88.94%
i
N
Row(%)
Column(%)		  
81657
51.50%
5.17%		  
76899
48.50%
47.19%		  
158556
9.10%
s
N
Row(%)
Column(%)		  
11584
33.93%
0.73%		  
22561
66.07%
13.84%		  
34145
1.96%
Total
 1579105
90.65%		 162971
9.35%		 1742076


This figure represents the row percentages from the table above:

Status by affiliation

We can see below that “unaffiliated” voters are overrepresented in the “in suspense” voters (as are Democratic and Libertarian, but to a lesser extent):

 
Party		 Status at 2015-06-11
a		  
i		  
s		  
Total
democratic
N
Row(%)
Column(%)		  
369140
88.78%
23.83%		  
40594
9.76%
25.60%		  
6046
1.45%
17.95%		  
415780
23.87%
libertarian
N
Row(%)
Column(%)		  
11277
82.26%
0.73%		  
1887
13.76%
1.19%		  
545
3.98%
1.62%		  
13709
0.79%
republican
N
Row(%)
Column(%)		  
713674
92.65%
46.06%		  
48792
6.33%
30.77%		  
7792
1.01%
23.13%		  
770258
44.23%
unaffiliated
N
Row(%)
Column(%)		  
455278
84.02%
29.38%		  
67275
12.42%
42.43%		  
19307
3.56%
57.31%		  
541860
31.11%
Total
 1549369
88.96%		 158548
9.1%		 33690
1.93%		 1741607


I’m also beginning to look at this relationship at the county level, so that I can control for other demographic variables. While those analyses aren’t ready yet, we can see the negative relationship between the proportion of voters registered Republican in a county and the proportion of voters “in suspense” in a county. Note that these analyses are done using the most recent voter file (June of 2017).


Since the table above indicates that this relationship might be driven by unaffiliated voters, we can also look at the positive relationship between the proportion of unaffiliated voters and proportion with “S” registrant status:

Dropped by affiliation

Lastly, we can also see below that “unaffiliated” voters are overrepresented in the voters dropped from the file:

 
Party		 Dropped by 2017-06-06?
FALSE		  
TRUE		  
Total
democratic
N
Row(%)
Column(%)		  
372029
89.48%
23.56%		  
43751
10.52%
26.92%		  
415780
23.87%
libertarian
N
Row(%)
Column(%)		  
12042
87.84%
0.76%		  
1667
12.16%
1.03%		  
13709
0.79%
republican
N
Row(%)
Column(%)		  
713280
92.60%
45.17%		  
56978
7.40%
35.06%		  
770258
44.23%
unaffiliated
N
Row(%)
Column(%)		  
481737
88.90%
30.51%		  
60123
11.10%
36.99%		  
541860
31.11%
Total
 1579088
90.67%		 162519
9.33%		 1741607