« Wrangling the legal doggies | Main | Detailed analysis wanted »

January 25, 2005

Analyze this

The other day, Stefan Sharkansky at Sound Politics posted the self titled "definitive analysis" of King County's votes vs voters issue. For those out of the loop, there are more votes than credited voters. The discrepancy at last published count was about 1800.

I am unaware of the process being used by King County to sort out the discrepancy. But it would seem to me that there are four basic things one would need to know:

1. All duly registered voters as of November 2.
2. The assigned precinct for each registered voter.
3. Which voters actually cast ballots.
4. To which precinct the cast ballot was credited.

Unless you're Stefan...who somehow thinks that if he justs comes up with all the registered voters and how many votes were cast in the precincts..he's got the analysis pegged.

A commenter on the blog (Scottd), pointed out the problem:

Impressive analysis, but the bottom line is you don't have the one set of data you need to defend your conclusions. You don't have a complete list of voters who were credited with voting in Nov. 2004 along with the precincts where their credit was assigned.

You've done impressive work attempting to reconstruct this list using data from publicly released voter lists and other sources. And, you acknowledge some possible sources of error in your list, but you don't do much to quantify that error. Even if your list is 99.5% accurate, it would still contain nearly 4500 errors -- and it's quite possible that your error rate is higher than that. Here are three possible error sources you need to consider:

1. While you picked up voters from the Dec. 29 file who were dropped in the Jan. 7 version, I don't see any attempt to pick up voters who were registered on Nov. 2 and dropped before Dec. 29. I estimate that KingCo dropped nearly 60,000 voters from its rolls over the last year. (They added approximately 110K new voters, yet, their total voter registration only increased by 52K.) That works out to an average of nearly 5000 voters per month. From Nov. 2 to Dec. 29, it's possible that nearly 10000 eligible voters were deleted from KingCo's records. I don't know how many of them voted in the election, but neither do you. However, this one source of error is potentially as large (or larger) than the number of mismatched voter/ballots you report.

2. Your assumption that precinct assignments in the Nov. 2 file are correct is suspect. It's quite possible that some of these assignments were in error and corrected in later lists. In fact, this is practically a certainty, since one of the reasons we have provisional ballots is to deal with errors in precinct assignments.

3. Your attempt to allocate provisional, absentee, and poll ballots in the final recount is pure guesswork, as you admitted in your post:

I used the counts of each type of ballot per precinct from the machine recount, but the precinct total from the manual recount, realizing that the totals would still be off by 1 or 2 in some precincts.

Uncertainties of +/- 1 or 2 voters in some precincts could easily add up to total errors in the hundreds when you consider than KingCo has over 2600 precincts.

To borrow a phrase, it would seem that the margin of error in your analysis is at least as large as your conclusions.

Congratulations to scottd, who had his final post expunged and is probably now banned. LOL

For those of you looking for a "definitive analysis" on the Washington Gubernatorial Election and it's various foibles, I doubt you'll find it from the fingers of Mr. Sharkansky. This is not the first time his "analysis" on these matters has been less than shaky.

The only way I can see to truely analyze this issue properly is to have the four pieces of information that I outlined above. I can't rightly see how that can be done without the precinct sign in sheets (voters in Washington sign into the voter roll sheet at a table in their polling place) that outline who actually voted and which precinct they're in. And then there has to be a way to figure out which precincts each ballots were credited to.

I've picked up a partner in this endeavor in the form of Torrid Joe of Also Also blog. This sort of analysis isn't my strength, so I'm hoping Joe can hand hold me through it. LOL

Joe is in the process of making some contacts and gathering up the needed information. We'll keep you posted as to how we do.

Posted by Carla at January 25, 2005 06:51 PM