here over the weekend reported that the MNPS school calendar survey
found support for the balanced-calendar option in most high-school clusters across the city that was higher than seemed possible given the essentially even split among households citywide. MNPS has explained how this can be: faculty and staff responses, not just parent responses, are included in the cluster-by-cluster results.*
Given that faculty in the poll favored the balanced-calendar option by a wide margin, their inclusion in cluster results has the inevitable effect of elevating cluster support beyond what it would be for just parent households. Does this matter?
It does if school board members are interested in gauging sentiment on this issue among parents in their districts. In the overall (citywide) sample, parent opinions were closely split between the balanced (45%) and traditional (44%) calendars, but in nine of eleven high-school clusters the balanced calendar wins, in eight of those by several percentage points. This makes it harder for school board members to draw accurate conclusions about parent sentiment in their districts.
We asked MNPS why they haven't released cluster-by-cluster survey results that include only
parent households. MNPS spokeswoman Diane Long's reply: "No one brought that up. This is meant to be a general guide to the board." But doesn't the inclusion of faculty data make cluster support look higher than it really is for households with parents and students? "I don't know if you can necessarily make that conclusion," said Long, who added that the survey "doesn't need to be sliced and diced."
We also asked MNPS for information that would help us assess whether there is sample response bias in the telephone poll, which drew input from 36 percent of parent households. Ordinarily, a survey with a modest response rate is of questionable value unless you can show that respondents and non-respondents do not differ in meaningful ways. It would be relatively easy to test for some forms of response bias with the data MNPS have in hand, but they aren't interested: "This is not an instrument that lends itself to looking for bias," said Long, "I think you can get an excellent gauge of public opinion without being scientific."*Some faculty and staff, such as those who work in the central office, are not associated with a cluster; hence the cluster response total falls short of the overall response total.