About Invalid Ballots in Burlington 2009 IRV mayoral election

Extracted (almost) verbatim from two posts by Juho Laatu

There were 8984 votes out of which

The percentage of ballots that were not filled properly was small.

The percentage of votes that failed or could have failed to indicate all the relevant preferences that the voter had seems to be higher (part of the 606 and 2458 votes). This is not catastrophic though. Maybe people will learn, or maybe it is acceptable to have even this kind of numbers in the long run.

(Some of the voters may also be bullet voters by nature ("All or nothing", "My candidate is the best and all others are out of consideration", etc.) Maybe they want to send this kind of message in their vote (even though they understand that they will cast a weak vote).

It seems that ballots were not too complex to fill properly, and most voters also filled them well enough to be counted also at the last round. One remaining concern is that someone said that the level of participation was low. (The number of votes in the Mayoral race, 8984, was 43% of the number in the US presidential and VT governor races held a few months previously.) The complexity of the IRV method may be in some degree responsible for this.

More data on the problematic ballots (=gaps)

(1=one mark, 0=no marks, 2=two marks, five numbers for the five columns)

"Regular votes" with one candidate in the leftmost columns:

10000: 1475 bullet votes
11000: 1891 votes that marked two candidates
11100: 1775
11110: 873
11111: 2853 votes that used all the slots

Regular empty votes:

00000: 4

Regular votes but with ties:

21000: 1
11200: 1
21110: 1
21111: 2

Votes with one one slot gap:

01000: 3
01100: 1
01111: 2
10100: 6
10110: 1
10111: 3
10121: 1 (with a tie)
11010: 3
11011: 10
11101: 38

Votes with one bigger than one slot gap:

00001: 3
10001: 14
10011: 6
11001: 17

In general the number of irregularly marked ballots (ties and gaps) is quite small (113). If we count the 1-slot gaps as minor errors (that could happen to anyone) then 46 bigger errors remain.

Some random observations:

Return to main page