Elections Director Sherril Huff puts this claim at the center of her election campaign:

Staff reports to the Canvassing Board demonstrated continued improvements by elections officials in such areas as inventory reconciliation of ballots received and counted, and voters credited with voting. Using Six Sigma quality standards, the Elections Divisions' reconciliation reports balanced with 22 inventory discrepancies,

A summary of what I've found:

* Huff's office provided me with a Mail Ballot Report and Summary and Detailed Poll Canvass Reports, an Early Voting Report and what appears to be a report on Provisional Ballots that were sent to the canvassing board. They were unable to provide me with a comprehensive Provisional Ballot report, nor could they provide me with the summary reconciliation reports that are required by RCW 29A.60.235 and to be "publicly available at the auditor's office or on the auditor's web site".

* I also received and reviewed extensive voter database records, including the current list of voters, lists of all absentee and provisional ballots issued in the November 2008 election along with their disposition, the "voter history" file, which lists all voters credits, and an audit log of all changes to the voter database from mid-2004 to the present. To Huff's credit, I requested these files over the holidays and received them within 2 days after the staff returned from their holiday break. These are the same sorts of files that Dean Logan told me would take several weeks to produce, and in fact took him nearly 7 months.

* My review of the data records indicates that ballots to voters reconciles reasonably well county-wide, but not necessarily by district. For example, more provisional ballots appear to have been tabulated in some legislative districts than there were provisional voters registered in those districts and vice versa. This could happen if a voter, say, who lives and is registered in Seattle casts a provisional ballot at a Redmond polling place. In such cases, the submitted ballot is supposed to be duplicated onto a ballot from the voter's home district, with only the common races marked onto the new ballot. [See canvassing board rules, sec. 6.2(b)]. If the ballots are not correctly duplicated, it would be possible for a voter who is registered in one district to have their vote counted in a race in which he is not eligible to vote.

I found a net discrepancy of 56 ballots when reconciling by legislative district. e.g. the 1st L.D. had a net 21 more provisional ballots counted than there were provisional voters, and the 45th L.D. had a net 12 more ballots than voters, etc. Some other districts had a net excess of voters to ballots. Of course, the net numbers can mask more incidents of the out-of-district voters having their ballots counted in the district cancelled by an equal number of the opposite incidents.

The absentee ballots also do not appear to reconcile by district. It's harder to get a more exact number, as there is an excess of voters credited over ballots counted, due to empty envelopes and spoiled ballots.

I'm open to the possibility that my analysis was based on incomplete data or was otherwise incorrect. But ensuring that absentee and provisional ballots are tabulated in the correct district is an important task. I would assume that the Elections Office would prepare reports to demonstrate that this is done correctly. If the Elections Office wishes to disclose reports showing a different discrepancy between votes and voters at the district level I will happily post those.

* The data records also indicate that some voters who submitted two ballots had both ballots counted:

43 voters are recorded as having cast both a poll ballot and a provisional ballot, with 1 voter getting both ballots counted, a 2.3% error rate.

UPDATE: I also notice that there were an additional 6 provisional voters who had their ballots rejected on the grounds that they had voted a poll ballot; but they were not actually credited with voting at the polls and the polling site reconciliation reports do not indicate any unexplained extra ballots at those precincts. So I guess the total error rate on preventing double-voting at the polls and by provisional is 14%!

268 voters are recorded as having cast both an absentee ballot and a provisional ballot. In 19 of these cases both were rejected, although it appears that in some of these the provisional probably should have been accepted, i.e. when the absentee ballot was not signed. In 29 cases the provisional was accepted and the absentee rejected, in violation of WAC 434-253-047(6) which gives precedence to the absentee ballot. And 10 voters are recorded as having had both ballots accepted. The error rate is between 4% - 17%, depending on how you count.

9 voters are recorded as having cast both a poll ballot and an absentee ballot, with 4 voters getting both ballots counted, a 45% error rate.

Given that so much infrastructure is supposedly deployed to prevent voters from voting more than once, one would expect that the error rates on the relatively few situations where voters submit multiple ballots to be near 0, but it isn't.

I welcome any documentation from the Elections Office which shows that my analysis is not correct.

Again, the 2008 election was administered much better than the 2004 nightmare. Huff deserves a share of the credit for that. But her wildly optimistic campaign boasts of only "22 discrepancies" and "Six Sigma quality standards" (which implies a mere 3 errors per million) don't seem to be nearly true.