Flawed Enrollment Projections

To the Superintendent and Members of the Board (sent 3/28):

At last Thursday's school board work session, Ms. Henning asked about the accuracy of the PSU demographic projections. The staff indicated that these projections were generally accurate; the only school identified as having had flawed projections was Forest Park.

A quick review of the information available on the PPS web site demonstrates that this is by no means the case, at least not when one is talking about long-term trends. While PSU's predictions may be reasonably accurate in the short term, that's likely because PSU re-normalizes predictions based on the prior year's enrollment. If one goes back to 2000, however, and compares the predictions for 2005 that were relied on then, to the actual enrollment for those schools in 2005 (taking into account, as best I can, changes due to boundary fluctuations and closures), there are a number of schools you see where the 5-year predictions for schools other than Forest Park proved to be off as much as 71%, with an average (absolute) error of 26% (well more than one full class of students), and a median (absolute) error of 19%. That's a long way from being "pretty good," as was suggested last Thursday. An Excel spreadsheet, excerpted below, summarizes my findings about the quality of 5-year projections.

The spreadsheet looks only at west side schools other than Forest Park. Because I've had to make some adjustments in 2005 enrollment to take into account boundary and other program changes over the last 5 years, I've stuck to the west side schools with which I'm most familiar. The adjustments are explained in the spreadsheet; for each, I've made an effort to estimate the number of students curently enrolled who are from the 2000 school year boundaries; in essence, it subtracts the effect of the Smith Closure to three schools and the Odyssey addition to Hayhurst. I'm sure my estimates aren't 100% correct, but it should be close enough to give pause. Carried over to the east side, I expect you'll see many of the same problems.

I should note that the point about the validity of these projections was discussed around the time of the 2003 proposed closures, and it's very unfortunate that it's not clearer at this point. (I'd point out that the same thing is true regarding portables, which were discussed on Thursday as well; at issue in 2003 was the fact that capacity calculations included (probably improperly) portables for some of the schools (including, for instance, Maplewood). If the issue of portables isn't perfectly clear by now, I really worry about the quality of the data on which the superintendent is relying. That certainly isn't the only issue that's being missed.)

This analysis, though admittedly limited in scope to the west side, suggests that the Board should use great care in relying on these numbers for anything but one- or two-year trends at individual schools. This is precisely one of those areas in which, if you engage parent involvement before making decisions, you can learn a lot of information that staff otherwise misses because they have so much else to do. I'd suggest a careful comparison of the 2000 predictions to 2005, and a carryover of those lessons to any predictions used for 2010.

Bridlemile underestimated by 22%
Capitol Hill underestimated by 71%
Chapman underestimated by 8%
Hayhurst overestimated by 42%
Maplewood underestimated by 35%
Markham overestimated by 8%
Rieke underestimated by 19%
Skyline overestimated by 17%
Stephenson overestimated by 9%

Jeff Dobbins

Email to sjdprods at yahoo dot com for the full spreadsheet.