[T]he KABC findings should be considered deeply flawed.

In fact, SurveyUSA has been very wrong in LA primaries before. Just three weeks before the 2001 primary, SurveyUSA said that Hahn was beating Villaraigosa by 12. Villaraigosa ended up beating Hahn by 6.

Our own poll, conducted 2 weeks ago, shows Bob tied for second with Hahn at 20 percent - a jump of 10 points for Bob since December - while Antonio's poll apparently shows that Bob is in second, 5 points ahead of Hahn.

As a result, we expect Hahn's political consultants to launch their televised attack on Bob very soon.

There is plenty more evidence about SurveyUSA and a lot of it follows - but rest assured, we are confident that these results don't accurately reflect the momentum of the campaign.

SurveyUSA does not use a live person to administer their surveys. The questions are asked by a computer-generated digitally record voice. Responses are recorded via the keypad on the respondent's phone. In the words of Michael Traugott, chair of the Communications Studies Department at the University of Michigan and an expert on political polling, one of the limiting factors in this kind of polling is that the polling firm, he said, "has no idea who is on the other end of the phone. It could be a kid or someone who doesn't live at that address."

SurveyUSA chooses who to call on the basis of their phone number, rather than by selecting from a list of actual voters. This kind of sampling, known as Random Digit Dial, relies on voters to tell the truth about whether they are likely to vote in the election in question. In high turnout elections, this is not much of a problem - the respondent really is likely to vote. But in low turnout elections like the one happening next month, non-voters are more inclined to lie and give the socially desirable response - of course I plan to vote - than admit that they never actually make it to the polls.

In low turnout elections, RDD sampling introduces bias. Candidates with low name recognition are even less well known by non-voters than they are by likely voters. In other words, an RDD survey in a low turnout election will overstate the strength of a well-known candidate - like Jim Hahn or Bernard Parks - and understate the strength of a less well-known candidate - like Bob Hertzberg.

Taken together, these methodological flaws suggest why the SurveyUSA survey is showing Bob in 4th even as today's LA Times reports on Bob being one of the three top candidates"locked in an increasingly complex game of chicken."