Sinclair also works on enrollment projections, Garry, but I can honestly say our
 approach is somewhat simpler than what I interpret yours to be.  Over the last
 8+ years I have been using penetration rate information, taking age cohorts at
 Sinclair and devising a rate of penetration into similar age cohorts within our
 service area.  Historically, this information, coupled with environmental
 scanning efforts and an understanding of internal change (new programs,
 outreach efforts, etc.) has been within + or - 3%, and usually closer to 2%
 rate of error.  Lots of art, somewhat less science than what you do.  This
 year, however, we also employed the use of regression modeling, finding that by
 employing local unemployment rate, number of high school grads and time trend
 info we could come up with a model with a multiple R of .98560 and an adjusted
 R2 of .95710.  I am sure your model will do better, but we felt that this was
 fairly reasonable as it seems to underpredict on average 1.3%.  Will it work in
 the future?