I took a quick look in our archives and gathered data from all our elections this past year. There were 13 of them in total; 3 judicial, 3 general, 6 special, and 1 run-off. Here is the data:
The average number of votes cast is 54.7, with a low of 27 (February 2015 Special Speaker Election) and a high of 90 (May 2015 General Election). Unsurprisingly, special elections, with an average of 46 votes, almost always receive fewer votes than both general and judicial elections, with respective averages of 72 and 47 votes.
The average number of votes cast during the last 24 hours is 4, with a low of 0 (July 2015 Judicial Election) and a high of 13 (May 2015 General Election). Excluding that outlier, all other elections have between 1 and 6 votes during the last 24 hours - except for a few cases, this is not counting votes edited within that time period.
Looking at these together, if we eliminated 5th-day voting and assumed that all citizens who would vote during the extra 24 hours would instead not vote[note]An excessively restrictive assumption, I think, since some people vote at the last minute but would do so regardless of the length of the voting period. I think it's best to examine this from a "most votes prevented" perspective to determine the most damage it could do, though, so that is what I am doing here.
[/note], we would lose an average of 7% of the votes that would otherwise be cast in any particular election. Again, the minimum here is 0% (July 2015 Judicial Election) and maximum is 14.4% (May 2015 General Election).
I also examined the decisiveness of these votes - that is, whether their casting affected the result of the election.
In exactly two cases, the late votes did affect the election results. These two are the May 2015 General Election, in the race for speaker, and the May 2015 Run-Off Election, also for speaker. The former was tied at 40-40 with 13 votes cast or edited[note]As I am not an admin, I cannot see the prior content of any particular edited post. It is thus impossible for me to determine if any individual edited post did indeed affect the result of the election. However, given that the vote was a tie, and given the high number of late-edited votes, I deem it highly likely that these late decisions did affect the result.
If an administrator wished to go through that thread and confirm my proposition, that would be an interesting project.
[/note], and it led into the latter, which ended at 35-37 with 5 votes cast or edited during the last 24 hours. In this case, a shortened voting period may have affected the results.
In 9 elections, late votes had zero effect on the results.[note]In the interest of full disclosure, I will admit that I did not carefully examine these elections to determine how many people edited their votes during the last 24 hours. The lead was strong enough for these, and my recollection of that being the case even early on, means that I am confident enough to say that late votes did not decide these elections despite not checking this data. Someone else is of course free to do so if they wish.
[/note] These elections were generally very lopsided, with a huge majority of the votes going to a particular candidate or candidates, and the number of late votes cast would not have changed the result even if they had all been for the runner-up candidate.
In two elections, the November 2015 Judicial Election and the March 2015 Judicial election, the number of votes or edited votes (6 and 4, respectively), could have been enough to change the results but did not do so in these instances. To explain with an example, in the former, the winner with the 3rd most votes had five more than the runner up, with the fourth most. 6 late votes could have been enough to put the 3rd placer above the runner up, but in this case, the 3rd place candidate was not in all of them and subtracting all six would not have put the runner up in 3rd. Similarly, in the latter, the 4 late or late-edited votes would have been enough to let the 4th placer overcome the 3rd placer, with a separation of only 2 votes, but in this case were not cast such that that was true.[note]I hope this makes sense and I'm explaining it okay. I can try again if it's confusing.[/note]
The conclusion that I draw from this is that, at its worst, shortening the voting period may affect the results of competitive elections and will almost certainly affect the results of highly competitive ones. However, in highly competitive elections like the one for speaker last May, anything could affect the results - lengthening the voting period would almost certainly have changed the outcome, but we are not advocating for giving people more time. Additionally, this is the worst-case scenario; I think in actuality the results would be less noticeable, as most people (if you can describe anything less than 5 as "most") who would have voted during the last 24 hours would instead vote earlier.
