Norristown Area School District.

Four decades of enrollment

Where Norristown has been, and where it is headed.

Norristown's enrollment is not a straight line: down through the late 1980s, a sharp early-1990s recovery, a long climb through the 2010s to a 2022 peak near 7,800, and a drift downward since. Our model and the Pennsylvania Department of Education's official projection — built independently, on different datasets — both see that drift continuing: -11% by 2034 on our read, -8% on PDE's, with the two paths never more than 3.7% apart across the decade.

The permits panel explains a stretch of the enrollment story that a trend line cannot: the district's municipalities permitted only a trickle of homes through the 1990s, then delivered a burst in 2006–2007 — several hundred single-family homes followed by a wave of apartments — right before the national housing crash. As those units filled and turned over, enrollment climbed nearly 800 students over the following decade, and more than a thousand by the 2022 peak. That is the housing-to-enrollment lag our model measures statewide and applies here.

The test that matters

Rewind to 2008. What would each forecast have said?

Fall 2008: the housing boom has just ended, the crash is underway, and a planner needs a ten-year outlook. Enrollment sits at 6,727 after a flat half-decade, so using only data available that fall, the same-enrollment baseline stays at 6,727 and the trend line drifts to just 6,952 — both blind to the growth about to arrive. The District Foresight model, reading the cohort structure and the just-permitted housing pipeline, projects 7,510 for fall 2018. Actual: 7,491 — within 0.2%.

This window also shows exactly what the housing signal buys. A cohort-survival model without the permits term — the same family Pennsylvania uses officially — would have projected 7,115 for 2018, still 7,491 − 7,115 ≈ 380 students short. The 2006–2007 pipeline, visible in the permits data at projection time, closes that gap.

One window proves little on its own, so here is the honest record: we ran the same exercise from every base year. Across all 24 ten-year windows since 1991 the model's advantage in Norristown is real but not uniform — in a district whose enrollment swings and mean-reverts, the same-enrollment baseline wins some windows, particularly in the data-sparse early 1990s. Since 2000, the model averages 7.2% ten-year error in Norristown against 8.3% for same-enrollment and 9.5% for the trend line. The statewide evidence — 499 districts, 141,134 scored projections — is in the methodology report.

Method & data, side by side

How this compares with Pennsylvania's official projection.

The Pennsylvania Department of Education publishes ten-year projections for every district, and they are good ones — the comparison here is a cross-check between two implementations of the same core idea, not a takedown. The differences are in what gets added on top, and in how accuracy is demonstrated.

The two datasets agree where they overlap — Norristown's fall-2020 kindergarten count is 520 in both the federal and state records, and the fall-2018 district total (7,491) is identical in both — which is exactly what you want from an audit trail. One honest limitation of the comparison: PDE publishes only its current projection vintage, so its historical accuracy cannot be scored retrospectively. Our backtest's cohort-survival-only results stand in for that model family, and the 2008 hindcast above shows what the housing term adds beyond it.

The decade ahead

What the model says now.

From fall 2024, the model projects an 11% decline by 2034 — to 6,842 students — as smaller entering kindergarten cohorts work their way up through the grades and the current housing pipeline runs too small to offset them. PDE's independent projection reads the same direction, a touch shallower (7,067 by 2034). For a district office, the practical takeaways are concrete: elementary seats free up first; secondary enrollment holds up through the late 2020s before the smaller cohorts arrive; and the measured planning range at ten years (±7.9% median, per the statewide calibration) comfortably brackets both projections.