Crombie cruising to an easy win? Five reasons maybe not.

I speculated earlier in the week that Steve Mahoney might do great in southwest Mississauga, but Crombie could expect to do well in the other three quadrants. The last Forum Poll suggests that Mahoney may be trailing Bonnie Crombie even in his "home" quadrant in the southwest. Mahoney is also trailing with voters over age 65.

That is bad news for Mahoney.

Municipal elections are often decided by differential turnout. The big challenge for pollsters is not determining who citizens prefer, but rather who is likely to actually vote. Older voters who have lived in an area for a long time are one of the groups most likely to vote. In this election, I expect that is a big advantage for Mahoney, who last held elected office in 2004. People who have long supported Mahoney and his wife, City Councillor Katie Mahoney, in southwest Mississauga should be expected to be Mahoney's core base of support. And many of those who remember voting for Mahoney at the federal or provincial level are among the most likely to turn out to vote. Advantage Mahoney.

Surprisingly, the latest (and potentially last) poll, by Forum Research (see http://poll.forumresearch.com/post/179/crombie-maintains-wide-lead-in-mississauga and full results here: http://poll.forumresearch.com/data/Mississauga%20Issues%20News%20Release%20(2014%2010%2024)%20Forum%20Research.pdf) show Mahoney behind in both his home "quadrant," 49% - 41% and trailing among older voters. Actually, to be blunt, Mahoney is trailing in pretty much in every demographic category: young voters, old voters, men, women, PCs, and Liberals. This does not portend well for Mahoney.

However, be cautious when reviewing these results. This poll may be an accurate gauge of public opinion in Mississauga, but much can be wrong about these results.

This Forum Research poll is an automated telephone call conducted on a Friday night. Only 308 respondents answered the call and told Forum Research who they intended to vote for (or had already voted for). Forum Research included 286 of these responses in a proprietary weighting scheme based on the demographic information the respondents provided to give estimates that should reflect the voting population in Mississauga. There are five reasons to be especially cautious when reviewing this polls' findings:

First, automated polls are good at sampling households or phone numbers, but not household members. If certain demographics are more likely to answer the phone in a house - like they are in my house- that could bias the results. For example, many young people assume that no one they know will be calling their parents' home phone, so many do not answer that line. This can be minimized by the pollster's weighting algorithm, but only if one assumes that those who answered the phone had similar views to those with similar demographics who did not. In other words, a 22 year old living with her parents who did not answer her phone would be expected to have similar political views to a 22 year old living in an apartment (with a number that identifies her as a Mississauga resident).

Second, the poll was completed over just one evening. A Friday. Were you home all evening on Friday? Many people go out on Friday nights. Especially if they are celebrating Diwali, or honoring a soldier on the Highway of Heroes, as many did on Friday night, October 24, 2014. They didn't answer their phones. Neither did some who were enjoying a family dinner. That could bias the sample. Most phone surveys go out in the field for more than one evening to avoid biases caused by certain demographic groups being out one night compared to other groups. I am not surprised that about 40% of the sample were older than 65.

Third, most surveys are larger than 300 people. I suspect that Forum intended this survey to be larger; a recent poll of Brampton voters was a more typical 1000 voters. This may have been intentional, as Forum may have been working with a small sample to mute criticism that their previous poll was biased by a robocall sent out shortly beforehand by the Crombie campaign. Or, it may reflect a paltry response rate (it was a Friday night!). Forum didn't disclose their response rate (unless I'm misreading the difference between 308 calls and 286 respondents with demographic info). Response rate biases are not covered in the published margin of error (6 percentage points in sample of this size), but can be minimized by Forum's weighting algorithm IF there is nothing remarkable about the people with similar demographics who did answer the phone.

Fourth, Forum Research did not disclose the number of undecided voters. Forum's reported figures include those who have decided whom they are voting for, and those who are leaning towards one candidate or the other. Since the percentages sum to 100, Forum Research must have excluded all those who said they were so undecided that they were not even leaning one way or another. Or, maybe that is the gap between 308 voters they said they sampled and the 286 voters whose demographics they report - 10% of the sample, a likely proportion. If many voters are undecided, or are leaners, we could potentially see a big shift between this poll and election day - if those people vote. Undecideds and leaners are less likely voters than those who have already decided whom they are supporting. If they turnout, and vote in ways different than many of their neighbors, the results could be very different than this poll.

Fifth, all the usual problems with telephone polls in the early 21st Century still apply: many citizens do not have land line phones, some are not comfortable with speaking English on the phone, many simply do not respond to polls.

Remember this poll does not substitute for voting on election day. There is little or nothing promising in this poll for Steve Mahoney, but this poll is even more suspect than many other polls conducted by Forum Research. Treat these results with caution.