Unweighted results of a recent telephone survey of Mehlville School District residents show about 45 percent of respondents would favor — and about 51 percent would oppose — a possible 94-cent tax-rate increase referendum.

The numbers differ from the survey results presented recently to the Board of Education which show about 51 percent of respondents would support the referendum and about 45 percent would oppose it. But the district’s consulting firm says the unweighted results are not accurate because they are not representative of Mehlville’s population.

The survey was conducted in July by communications firm UNICOM•ARC under an $18,250 contract with the district. It included 501 random telephone interviews.

Board members are scheduled to decide today — Aug. 19 — whether to place a 94-cent tax-rate increase proposal on the Nov. 2 ballot. The increase would fund COMPASS II — Charting the Oakville-Mehlville Path to Advance Successful Schools — a long-range improvement plan that contains more than $107 million worth of proposals designed to make Mehlville a high-performing school district.

The weighted survey results were presented July 29 to board members. During that presentation, UNICOM•ARC President Rod Wright noted that reaching younger families while conducting this survey was difficult because they likely were on vacation. Surveyors, then, performed a “little statistical adjustment” to make the demographics of the survey match those of previous surveys UNICOM•ARC has conducted for Mehlville, Wright said, noting the move is “not uncommon” and that it resulted in a “single percent” difference.

However, board Vice President Venki Palamand said at the board’s annual retreat Saturday that the unweighted results of the initial and reconsideration questions regarding the possible referendum are roughly 5 percentage points different than the weighted results presented.

The unweighted results show that, initially, 40 percent of respondents would support the 94-cent tax-rate increase and 55.2 percent would oppose it. When asked again about the possible referendum toward the end of the survey, 45.1 percent of respondents said they would favor the proposal, and 50.9 percent said they would oppose it.

The weighted results presented to the board July 29 showed that, initially, only 45.8 percent of respondents would favor the possible 94-cent tax-rate increase.

When asked later in the survey about the possible 94-cent tax-rate increase after answering several other questions regarding that increase, 51.3 percent were in favor and 44.8 percent were opposed.

In both sets of results, the margin of error is plus or minus 4.4 percent.

In the unweighted results, 75.6 of respondents said they were 50 years of age or older, while 23.2 percent said they were 18 to 49 years of age. Once weighted, the results presented to the board showed 60.3 percent of respondents said they were 50 years of age or older, while 38.6 percent said they were 18 to 49 years of age.

At last Saturday’s board retreat, Palamand acknowledged UNICOM•ARC felt the initial results did not accurately portray the population of the district, but asked, “Isn’t that what the whole idea of a random sample is, to get an idea of 65,000 registered voters are thinking? You can’t ask 65,000 registered voters, so you do a sample. So you do a random sample. And yes, some people are going to be home and some people aren’t going to be home, but isn’t that all randomized?

“Because it makes a difference whether your results are 51 percent … I mean, I realize you have a margin of error and you could pass, you could fail. But at 45 percent, trying to pass this becomes harder.”

Andy Duttlinger of UNICOM•ARC said, “You’re correct to say that it’s a random sample and to a certain level, you get what you get. But one of the reasons we track demographics when we do a survey is to see whether or not certain segments of the population are being under- or over-represented. And it was really clear, and if you saw those unweighted numbers, the sample skewed very old … We wanted to adjust them to make them more representative of the population to give you a better, more accurate snapshot of where the community’s at.

“The first thing we did was go back and look at the voter registration we had, but the birth date information … is pretty limited to say the least. We weren’t able to use voter registration data the way we normally would to weight that. So what we did is we went back and looked at research we had done for you all in the past, and the most recent survey we did in (2008) is what we used as a basis for weighting the age demographics.”

UNICOM•ARC previously conducted a community survey for the district in July 2008 to gauge residents’ opinions about a 37-cent tax-rate increase proposal and a no-tax-increase transfer proposal.

In that survey, 68.6 percent of respondents said they were 50 years of age or older, while 31.3 percent said they were 18 to 49 years of age.

“So what we did was we weighted this data to make the age demographics more closely match what we had done in that survey,” Duttlinger said. “Even having done that, I still think it’s pretty conservative in terms of the weight that we assigned to that.”

But Palamand said, “It moved five points. Your numbers moved five points. So that’s pretty significant, between 45 and 51 percent.”

Board President Tom Diehl said, “But it’s a common way that you do surveys. When Gallup is trying to determine whether (Robin) Carnahan or (Roy) Blunt is ahead, they know about how many registered Democrats they have and how many registered Republicans they have. Then if they look and see: ‘Gosh, we have 200 more Republicans than we should have for a random survey, they’ll weight the Democratic ones accordingly and make that adjustment. So this is very common. Gosh, probably every opinion survey does that …”

“Which is why UNICOM•ARC is always so accurate,” board member Karl Frank Jr. said.

Board member Erin Weber asked, “Is there a chance though that the non-weighted survey could be more representative of who actually will show up at the polls in our community?”

“I think it was skewed much older than we would expect to see …,” Duttlinger said. “If you look at the age breakdown in the weighted data now, something like 38 percent of the population in the weighted sample is below the age of 50, and only 31 percent are showing as having kids under 18 in the household. So in terms of the kinds of people we expect would support a referendum — and in fact, do — they’re not over-represented in the sample. So it’s still a conservative approach to looking at these numbers … It’s striking a balance between trying to give us an accurate picture and trying not to overcompensate.”