To Really Screw Up An Election Count, Add Computers…
....and dumb people who don't know how to properly set up systems or check their errors.
While the Wisconsin recount slowly grinds on, I bring you two stories courtesy my membership in the European Spreadsheet Risks Interest Group (these stories don't seem to rise to the level of our horror stories, though.)
First, out of Waukesha, we have "Kathy's special program":
Waukesha County Clerk Kathy Nickolaus was using a software program created especially for her by the state Government Accountability Board when she made the huge error in compiling results for the State Supreme Court race between incumbent David Prosser and challenger JoAnne Kloppenburg. The special program was revealed in an email sent to other county clerks and released by the Kloppenburg campaign to buttress her claim that an independent investigation of Nickolaus’ office is needed.
....
The email, by Rusk County Clerk Denise Wetzel and addressed to other county clerks in Wisconsin, including Nickolaus, was sent on April 8, the day after Nickolaus revealed the vote-reporting error in a press conference.It reads, “Please note that the program Kathy uses IS NOT the new canvass reporting program that is in the (Statewide Voter Registration System) that we all have been using as of late. It is a completely different program that was created by GAB for Kathy to accumulate her votes prior to uploading them into the program that the rest of us use.”
....
“We ran this for the governor’s election,” she said, “and we ran the same system for the February election. We had no problems.”Under the system, Nickolaus said municipalities, including Brookfield, are given a template spreadsheet to fill in with election results and email back to her. Municipalities are instructed not to alter the spreadsheet – but on the night of April 5, when Brookfield returned its form to Nickolaus, she noticed it contained “extra columns,” complicating the process of importing it into the database.
The clerk said she called staff in Brookfield to again stress the importance of not altering the template.
Nickolaus, however, didn’t say this complication was the cause of Brookfield’s vote totals being omitted from the initial results released by her office.
Hey, blame an intern. That's what most cubicle warriors do.
But yeah, giving people a data entry template you're going to use to import into your huge database.... WELL DAMMIT YOU WRITE YOUR QUERIES TO CHECK THE FUCKING COLUMN COUNT AND HEADERS.
So somebody accidentally hit "insert column", just to the left of the vote count column, I bet. It would've been an empty column. Oh, I don't know - having data checks before entering them into your database is pretty standard.
Thing is, Kathy isn't a tech person. Indeed, it shouldn't require a tech person to run an election. AND IT DIDN'T - because Althouse is also not a tech person, and she noticed that there was a problem with the magnitude of the numbers being reported.
WHAT IT REQUIRES IS SOME FUCKING COMMON SENSE.
I think the Kloppenburg campaign would have a point of saying "You really shouldn't have someone dumb as a brick running your elections".... but they're not the most credible people to be doing this.
Let's move on, shall we?
Sure, here's another spreadsheet error in an election, this time in West Virginia (and to no effect):
Morgantown City Clerk Linda Little said a computer error led to a discrepancy in vote totals for council candidate George Papandreas.
The candidate -- who was defeated by incumbent Ron Bane -- actually received 99 more votes than was originally indicated in the total listed on the city's unofficial results.
....
Little said election officials had put the total number of votes for each precinct into a Microsoft Excel spreadsheet. The computer program skipped over an entry when it added the totals, she said.
I think they're trying to make this sound more complicated than it probably was. I bet somebody manually put in a =SUM(:) function, and either forgot to include the first or the last row. That's all. That's not a computer error -- that's a human error.
In both of these cases, the error was with the human, as it usually is.
For those who are interested in preventing some basic errors like these (and how to check), I recommend Patrick O'Beirne's Spreadsheet Check and Control. Also, I have a paper at the Society of Actuaries that can be useful for non-actuaries: Spreadsheet Issues: Pitfalls, Best Practices, and Practical Tips(PDF).
Common sense can be trained up, to a certain degree.
Sheer stupidity, well, can't do much about that but pray for wisdom....




