Relationship between name order and votes in an election

#1
I'm analysing the results of an election with 42 candidates, and wish to test for a relationship between the order presented on the ballot and the number of votes each candidate received.

So far I've just used a simple bivariate model. Candidate ballot position (1st to 42nd) is the independent variable and number of votes as a dependent variable. Is it acceptable to use an ordinal as the independent variable in this way?

Of course neither variable is normally distributed. For the original data p= 0.0043 If I change this to a Box Cox distribution p=0.0034.

Is there a better approach?
 
#3
OK under 'correlation' on my software
I've found it outputs in a table format Spearman's rs=0.00268 between rank position and Votes (the two columns I've listed below)
with an option for Permutation p=.0031
So presumably that can be treated like a convention p value, yielding significance for the slope at the 1% level?

Here's the raw data, with the order shown on the ballot and vote data aligned. Perhaps someone can confirm the above figures?

1 21355
2 4148.2
3 6448.56
4 14566.46
5 2862.24
6 2197.41
7 13271.67
8 3660.93
9 12570.5
10 10830.89
11 689.47
12 551.66
13 932.12
14 3269.99
15 8883.66
16 1027.46
17 7801.11
18 9080.61
19 298.57
20 12535.89
21 648.41
22 509.27
23 885.4
24 1001
25 583.9
26 681.62
27 1882.44
28 2444.14
29 1945.89
30 15668
31 400.04
32 8973.9
33 355.03
34 5201.17
35 850.27
36 3472.99
37 4336.09
38 537.09
39 594.14
40 386.76
41 627.48
42 1571.74
 
Last edited: