Yesterday Dean Logan, the Acting Registrar-Recorder in LA County, released a report detailing a review they conducted of “non-partisan” (or “decline-to-state”) ballots in the 1% manual tally. To quote from the report:

In total, 1,820,758 ballots were reflected in the semi-official election results released Wednesday morning, February 6, 2008; 189,438 (10.4%) were non-partisan ballots. Our vote tallying system reported that 94,530 (49.9%) of these ballots included a vote in a party selection position (indicating a cross-over to either the Democratic Party or the American Independent Party) and a corresponding vote in a party Presidential candidate position.

…

Based on a physical examination of ballots summarized in Exhibit 1, we project that 24% of the non-partisan ballots cast were true non-partisan ballots with no cross-over party selection or Presidential candidate response position marked. Applying the same statistical model, we project 26% of the non-partisan ballots cast did include a mark in the range of response positions for Presidential candidates, but without a cross-over party selection. Based on the semi-official election results, this equates to approximately 49,500 ballots. Because the response positions were commonly assigned to candidates for all parties, it is impossible to determine with certainty for which candidate the voter intended to vote.

…

Based on our findings, we estimate that more than 70%, or approximately 140,000 non-partisan voters who intended to cross-over to cast votes in the Democratic or American Independent Party primaries or who chose not to cross-over and to vote a strictly non-partisan ballot did so successfully.

On average, 26% of non-partisan ballots, or approximately 49,500 ballots, appear to have been cast incorrectly. These ballots contained marks within the range of response positions for Presidential candidates, but did not contain a mark for the party selection. It is also noteworthy that the data compiled at the Congressional District level (Appendix D), as well as the precinct level (Appendix C), reveals a wide variation in the percent of incorrectly cast cross-over ballots. This variation strongly suggests that minimizing or eliminating errors in casting cross-over ballots can be achieved through improved voter outreach and poll worker instruction.