天美麻豆

Get stories like this delivered straight to your inbox. Sign up for The 74 Newsletter

The best way to predict a school鈥檚 test scores this year is to look at its scores from the previous year. But do 鈥済ood鈥� schools tend to stay high-performing? For how long?

To find out, I looked at 20 years of test results from my home state of Virginia. For simplicity鈥檚 sake, I focused on third grade math results and narrowed my search to schools that had at least 30 test-takers. To control for changes to the underlying tests, I sorted schools into four quartiles, then looked to see if they had moved significantly up or down from their initial category.

As you might expect, schools with high test scores in 2024 also tended to have high scores the year before. Among those in the top 25% of math scores in 2023, 68% remained there in 2024. The same was true at the bottom end, where 76% of schools that fell into the bottom 25% in 2023 placed there again in 2024.

But zoom out a few years, and the results start to become a bit more variable. This should be obvious, but the number of schools that stayed in the exact same category after five years was lower than it was after just one year. The results are the most interesting for schools in either the top or bottom quartile. Among schools that scored in the highest 25% in 2019, 61% were still there in 2024, and 66% of schools in the lowest tier were still there in 2024. But about one-third of the highest-scoring schools had fallen out of the top five years later, and a similar percentage had climbed out of the very bottom. 

What about looking back even further? I ran the same analysis to compare school performance in 2004 versus 2024. It鈥檚 still true that earlier results are predictive of later performance, but the relationship weakens significantly. For example, among schools in the bottom 25% in 2004, 44% were also there in 2024 鈥� but 56% were not. On the other end, 50% of schools in the top quartile in 2004 were still there in 2024 鈥� but that means half were not. In general, you鈥檇 still want to bet on a good school staying good and a bad school staying bad, but you should be much less confident the longer your time horizon.

Virginia started in 1997-98, well before other states were required to do so under the 2002 No Child Left Behind Act. The data for those older years suggest a similar weakening pattern over time. More surprising, there were some schools that looked bad in 1998 that produced some of the highest scores in 2024, and vice versa 鈥� some that looked outstanding in 1998 but had fallen dramatically by 2024.

For example, Virginia’s top elementary school in 1998 was still among the top 10 in 2024. But one that scored at the 99^th percentile in 1998 fell to the 74^th percentile in 2004, the 23^rd in 2019, and all the way to the bottom 10% in 2024. The name on the building stayed the same, but its performance plummeted.

For simplicity鈥檚 sake, I focused this analysis on school-level third grade math scores, and I suspect there would be more consistency if I had looked at district-level results or included more grades.

This discussion is also missing the dynamic effects of school openings and closures. For example, as in my recent analysis of Florida, Virginia had 138 elementary schools that were operating in 1998 but had closed by 2024. These tended to have low test scores. Of the 100 lowest-scoring schools in 1998, 32 were no longer operating in 2024. (Out of the top 100, just three had closed.)

Comparing two years of performance also misses the value that new schools can add. For example, of the 100 highest-scoring schools in 2024, 23 did not exist back in 1998. This combination of closing low-performing schools and opening new ones helped improve Virginia鈥檚 portfolio of public schools.

There鈥檚 also a lesson here for prospective home buyers who might be hoping to purchase a house that will buy them access to a 鈥済ood鈥� local school in perpetuity. That鈥檚 probably a good bet in the short term, but things can also change more than they might expect.聽

Did you use this article in your work?

We鈥檇 love to hear how The 74鈥檚 reporting is helping educators, researchers, and policymakers.