Sunday, December 3, 2006

Balanced Calendar II - The New (Erroneous) Math

Posted by on Sun, Dec 3, 2006 at 10:21 AM

There is an apparent and potentially serious flaw in the data being reported by MNPS regarding the balanced calendar survey. As explained here in a previous post, parents in the school system who responded to the survey were evenly divided on the proposed new calendar, 45-44 percent. When school board members, who will decide the matter on December 12, have addressed the issue, they have tended to signal intentions to vote in line with constituent preferences. The MNPS survey gives each board member some specific data on his or her own constituents by breaking down the survey results by high school cluster.

The problem is that the cluster data do not match the overall data. The results released Friday by MNPS show the balanced calendar prevailing in nine of eleven clusters, in all but one case by a margin of at least five percentage points. This might give most board members the impression that their constituents favor the change by a non-trivial margin. But: How can there be such cluster-by-cluster support for the balanced calendar when the overall household poll showed an evenly divided survey result?

(UPDATE: More on this question in a subsequent post.) The answer: There can't; the numbers don't add up. Here's a quick spreadsheet snapshot of the survey data reported by MNPS. It points either to a calculation error by MNPS or to a calculated effort to distort the results.

click to enlarge surveyspsht.jpg

The overall results from MNPS indicate that 18,207 households responded to the survey, but the cluster breakdown totals to 22,156 (the only number in the spreadsheet above that I calculated; all others appear in the Metro release). That's 3,949 households, or 22 percent, too many. Where did the extra 3,949 households come from? You could, if you wanted to, make the cluster results look more pro-balanced-calendar by adding in the faculty data, but there were 3,061 faculty respondents, so that doesn't add up. Maybe the cluster results include everyone--households, faculty, and staff? No, that makes the cluster total off by 844 people. Maybe the numbers don't add up because some surveyed staff and faculty also responded as parents: the total of households, faculty, and staff respondents would then be greater than the total number of actual respondents. But that doesn't fly because MNPS on the first page of its release is reporting 23,000 total responses, which is the precise total of the household, faculty, and staff numbers. If they are double-counting faculty who are also parents by putting them in both categories, then that inflates support for the balanced calendar in the total sample, given marked faculty support.

Bottom line: the citywide household split of 45-44 percent and the cluster breakdowns are mathematically irreconcilable. Assuming the total household respondent figure of 18,207 is correct, the cluster-by-cluster results significantly inflate support for the balanced calendar among student households within the clusters.

Before relying on any of this survey data, school board members should ask administrators to explain the discrepancies. While they're at it, board members should insist on having analyses for survey sample response bias performed (easily done with data in hand) to ensure that the results can be trusted as representative of the school household population.

Comments (11)

Showing 1-11 of 11

Add a comment

Subscribe to this thread:
Showing 1-11 of 11

Add a comment

Top Topics in
Pith in the Wind

Media (64)

Phillips (60)

Politics (47)

Around Town (27)

Education (20)

Legislature (19)

Law and Order (13)

40 Jealous Whores (9)

Sports (9)

Crazy Crap (7)

All contents © 1995-2014 City Press LLC, 210 12th Ave. S., Ste. 100, Nashville, TN 37203. (615) 244-7989.
All rights reserved. No part of this service may be reproduced in any form without the express written permission of City Press LLC,
except that an individual may download and/or forward articles via email to a reasonable number of recipients for personal, non-commercial purposes.
Powered by Foundation