September 09, 2004
Mayoral Poll Shows Potter With Strong Lead Over Francesconi
But A Quarter Of Likely Voters Remain Undecided
There's a KGW report from this morning on a recent public opinion poll which includes some numbers from the local Mayoral race between Jim Francesconi and Tom Potter. In the main, we're posting about it here because the story itself doesn't quite directly reflect what the poll itself indicates.
So to start with, here's what the KGW report says the poll says about the Mayoral race here in Portland:
Former Portland police chief Tom Potter holds a nearly insurmountable lead over city commissioner Jim Francesconi in the race to become Portland mayor. Since the May primary, Potter has extended his lead among likely voters. About 49 percent of them said they would vote for Potter while Francesconi attracted 25 percent of the potential vote. But another 25 percent are undecided.
"I think it's going to be awfully tough for Francesconi to make up that 25 percent gap," Riley said. "This thing is well outside the margin of error."
That Riley is one Mike Riley, the pollster responsible for the survey, described as a "random-sample telephone poll ... directed at voters, registered as of July 2004, who indicated that they are 'very likely' to vote in the November 2nd 2004 election."
In the case of the Mayoral numbers, only 157 people were surveyed (a comparatively small number, and at least partly responsible for the +/-8% margin of error), but more importantly, we wanted to point out what the poll results themselves (pdf) say about this race.
The question asked of the Portlanders surveyed was: "In the race for mayor of Portland, which of the two candidates will you vote for, Jim Francesconi or Tom Potter? (If Undecided) Which of the two will most likely get your vote?"
Of those 157 respondants, 26 people (17%) said they would vote for Francesconi and 13 people (8%) said they would likely vote for Francesconi, while 63 people (40%) said they would vote for Potter and 14 people (9%) said they would likely vote for Potter. Meanwhile, 39 people (25%) said they were undecided and 2 people (1%) said they would vote for someone else.
In the KGW report, they conflate the definite votes with the likely votes to end up with their 49% for Potter and 25% for Francesconi. In reality, the only certain numbers (using "certain" in that soft way reserved for public opinion polls, especially those with such a small sampling) are the definite 40% for Potter and the definite 17% for Francesconi. Arguably, those "likely" votes for each candidate could be considered at least somewhat persuadable.
Of course, the real unknown in these numbers is the full quarter of respondants who remain entirely undecided in this race. The open question, we suppose, is whether or not those respondants who did support a candidate are an accurate reflection of how the undecideds will shape up once they decide, in which case it'd be a done deal.
Either way, between the earlier-reported numbers from some internal poll that suggested Potter had a 30-40 point lead and this new Riley poll, the numbers continue to suggest to us the rationale behind the Francesconi campaign's recent swing towards negative campaigning. Conventional "wisdom" in politics, we suppose, is that when one is that far down, it usually means that one is not attracting voters and so the "only way to go" is to try to drive voters away from one's opponent.
Comments (6)
raging red on 09 Sep 2004
b!X, I thought you might be interested in this bit of info: I just received a flyer in the mail from the Francesconi campaign all about "the REAL Tom Potter," discussing two decisions he made as Police Chief. What's so interesting about that? I left Oregon in March. They sent this to me in West Virginia. I don't even know how they got my address or figured that I'd be voting in Portland's mayoral race this year.
Kari Chisholm on 10 Sep 2004
What's astonishing about that poll is that KGW thought it was reasonable to even have a conversation about a poll that had a 187-voter sample. I mean, are they kidding? That's microscopic!
In a recent article in Campaigns & Elections magazine (not available online w/o $$$), pollster Michael Cohen gives this guidance on sample sizes:
1000 - bulletproof
800 - acceptable by the press
600 - low end of what is acceptable by the press
400 - low end of what you should do for strategic purposes; not for publication
300 - only to decide whether or not to run
A sample size under 200? Totally unacceptable.
Kari Chisholm on 10 Sep 2004
er, I mean, 157-voter sample. Ugh.
Jeff on 10 Sep 2004
Kari's right--that's a ridiculous sample. But even worse, the "methodology" section reveals exactly nothing. Pollsters have several hoops to jump through before their findings are considered valid--and only the first is numerical. (Statistically speaking, you achieve decent reliability after something like 50 responses, but that ignores wide variation in political polling, which I should discuss outside of this parenthetical.)
In politics, your numbers don't mean a thing if they're not REPRESENTATIVE. If I go to the Red and Black Cafe and collect responses from the first 1000 people who come through the door, I'll have fantastic statistical validity. But the numbers will be useless. In order to acheive RELIABILITY, I have to have sample a group that looks like the voters.
All Riley idd was find "very likely" voters. But again, going back to my Red and Black example--I could have 1000 very likely voters there, too. But Riley doesn't tell us if it has found a representative sample--just a random one. If it wasn't representative, they might have weighted it, but there's no mention of that, so we don't know.
But we do know that this poll had Kerry and Bush tied, where a recent Zogby poll has Kerry up by ten. And Zogby, as we can see from their methodology, weights results "to ensure the selection of participants actually reflects the characteristics of the voting population."
Thus, FAR BETTER reliability. This Riley poll appears to be crap.
doretta on 10 Sep 2004
No, actually, what you DO want is a truly random sample of sufficient size from whatever your target population is, in this case, a random sample of likely voters.
If you get all your voters from the Red and Black, that is obviously not a random sample. You don't get reliable numbers by trying to be representative. All this weighting stuff is highly suspect and easily manipulated. My impression is that the only reason they do it is to justify not asking enough people to get statistical reliability, i.e., because it's cheaper.
All that statistical stuff about error and confidence levels assumes a random sample. Those measures tell you how likely it is that your random sample is truly representative.
a user on 14 Sep 2004
As I posted on blueoregon, jeff, Zogby's poll must be weighted because it is based on a self-selected internet sample. As far as survey research goes, Zogby is crap, not Riley.
The required sample size all depends on the size of the subgroup that you wish to examine and the variability of the phenomenon under examination. For instance, a "sample" of 10 might be deemed sufficient to determine whether a coin is fair or not; a sample of 100,000 would be insufficient to study the political attitudes of black catholic women.
157 may be enough to gauge in general terms a yes/no or candidate 1/candidate 2 race. The gap is so much wider than the standard error (23% vs. 8%) that this does merit coverage.