Doug Ericksen’s base is the good ole boy network in Ferndale and the rest of “out in the county.” I wonder how Seth Fleetwood is planning on cracking that nut?

Just keep on wondering, Walter.  What candidate exposes their campaign strategy to the public?

Uh-oh. Another snippy version of Wayne Farber has shown up to bedevil me.

I would call it good news, indeed. I know others are pounding Riley for the size of the “n” but I’m not so concerned if it was a good “n.” National samples are often based on n = 1000. If your n had been 100, I doubt you would be getting so much grief. The bottom line is how unbelievably close Seth is on the scale of “who would you vote for today” in spite of relative lack of name recognition, and he has over 3 months to go. As Riley said on his blog, Ericksen’s numbers won’t be going up after 16 years. This is Seth’s race to lose. I’m stunned and ecstatic to see such numbers this far from election day. And, as someone mentioned in the comments on Riley’s page, we’ll see how good the numbers are when we get the primary results. Thank you, John and Riley, for doing this.

I was astounded to read a Republican Party smear piece asserting Seth Fleetwood is an IRS deadbeat and mooching off the uber-rich.

I was subsequently unsurprised to dig in and find it’s a bunch of hooey ginned up by some lying Republican sacks of poo.

I’ll vote for a South Hill liberal with a record of integrity before I would ever sully my ballot for the cretins whose opening campaign volley is a untruthful smear badly told.

Geez, it’s a little early for an October surprise, and if you can’t do something that holds water, don’t bother.  Idiots.  If I was a Republican, I would be ditching the dimwit operatives of that unfortunate piece lickety-split.

While I like what the poll seems to show, the reality is n=85 renders it meaningless, as its variance would be very large, given the sample size versus the population size.  You want an n=300 or so for a meaningful read.

In addition, I am always skeptical of polls done by someone with a bias in the outcome.  They can be done well—that is, they can deliver an accurate snapshot of where likely voters are at the moment in time they were polled—but it’s very easy to let bias creep into the questioning, the phrasing, all the myriad details that can influence a respondents answer.  In addition, you need to make sure that the poll is representative of the larger population; extremely difficult if not impossible in a sample this small.

Nonetheless, I am cautiously optimistic about Seth’s chances.  He’s smart, doesn’t get flustered in a debate setting, and understands the issues.  Plus he has Julie Fleetwood on his side.

However, while I disagree with the accuracy of the mailer I got from Senator Ericksen, it is well put together and messaged to resonate with the average voter. 

A winnable race, but lots of work to do.

While I appreciate your points about a small sample size, that is a fair point there, I reject your premise that someone with a bias is unable to conduct a fair poll. We here at NWCitizen have endeavored to be completely transparent about the methods of this poll - the phrasing of the questions are exactly as listed above and you can judge for yourself how that would bias the voter.

We have another poll going out in September that will provide another glimpse at this race.

Hi Riley,

I don’t doubt your intent to be unbiased, but there are many ways bias can creep in.  I used to do polling, and have some training and experience; not just making stuff up.  For example, you say your questions are unbiased, and you provide them—which is helpful in addressing one vector through which biases can enter the poll.  However, I don’t know how you picked those sampled.  Are they representative of the County as a whole?  If so, that may introduce a bias, as the County as a whole may not reflect the subset of County residents who vote.  And if your sample is reflective of the subset who votes, how to you smooth the results to address infrequent vs. frequent vs. reliable voters?  What steps have you taken to analyze your respondents to ensure that subgroups—minority voters, voters by city or subarea of the County, etc—are proportionally represented?  How about those with landlines and no cell phones, and vice versa?  Are you addressing a likely higher turnout of the East Indian community because Satpal Sidhu’s presence in another district race, which will likely result in a significant increase in turnout in that community, and which will likely benefit Seth?

These are just a few examples of ways in which bias can creep into a poll.  Since by its nature bias tends to leave us a bit blind in certain areas, and we all have biases, it suggests that skepticism of results from a biased source is a good idea.  And being something of a numbers geek, and familiar with construction requirements for statistically valid polling, I can tell you that it is pretty much impossible to have a valid sample with a p=85 for the poll you are conducting. That does not mean the information isn’t interesting, or potentially informative, but it is definitively not reliable.

What I don’t understand is why, with the low cost of robocalls, you did not quadruple or quintuple your call numbers to get a larger sample which—with a lot of additional work on the raw data—could potentially produce a more meaningful and reliable result?  Something to consider for next time.

If you want some insights on how to strengthen your effort and results, call me.  I’m happy to help you get a more robust product.

We aimed for a sample size of 500 - we did 2,000 calls and previous robopoll efforts had netted a 25% response rate, however in the dead of summer, our response rate was much worse. I was pretty disappointed by the results, but we were left with the choices of sit on the data or release it. We released it. I am working to raise the money for a larger effort in September and I’ll give you a call for your donation when we get closer.

I do know how to balance a sample - I did receive quite a bit of training at this, thank you. While we did not adjust for the increased turnout of the Sikh community, we did get a sample that was proportional to the county-city divide, gender and age. While it is not a perfect measure, I was confident enough in our work to go forward with it.