The central problem confronting American education is not that we lack models of excellence; it is our inability to routinely replicate those models. Build a slicker cell phone or brew a tastier cup of coffee and the world beats a path to your door. Find a better way to teach kids everything from calculus to Cantonese and… crickets.

Our failure to replicate educational excellence has been recognized for years. In an attempt to overcome it, philanthropists have poured hundreds of millions of dollars into scaling-up networks of charter schools — hoping to grow the great ones and crowd-out the laggards. Regrettably, this strategy isn’t working.

Over the past half-year I collected data on the donations made to networks of charter schools in California — any group of two or more charters that share the same management, methods, or founders. I then studied the academic performance of these networks on the California Standards Tests, controlling for individual student characteristics, school-wide peer effects, and addressing a concern known as selection bias. As a check on those results, I also looked at charter networks’ AP test performance. If philanthropists were consistently directing their generosity to the highest-performing networks, there would be a strong correlation between grant dollars and academic performance.

There isn’t. The three top-performing charter school networks rank 21st, 27th, and 39th in terms of the grant funding they’ve received, out of 68 total networks. In fact, the correlation between CST scores and grant funding is so tiny as to be negligible — it’s a 1 on a scale from 0 to 10. That’s smaller than some correlations we see due to random chance. For instance, the correlation between the length of the networks’ names and their CST scores is twice as strong as the correlation between grant funding and scores.

The AP results are worse; higher grant funding is associated with lower AP performance, though the correlations are negligible in magnitude.

This disappointing overall result is not due to a lack of excellent charter school networks. For example, low-income black and Hispanic students at the top-ranked American Indian Public Charter Schools outperform the statewide averages for middle- and upper-income whites and Asians. AIPCS resoundingly outscores famous charter school networks like KIPP (which also does well), and even beats two of the most prestigious and academically selective public schools in the nation: Lowell High in San Francisco and Gretchen Whitney outside of Los Angeles. Lowell and Whitney receive thousands of applications each year, of which they accept only a small fraction — and they consider the applicants’ test scores in their admissions decisions. AIPCS consistently outscores them despite accepting all applicants or using a random lottery when oversubscribed.

Nor is the problem that growth inevitably breeds mediocrity. There is no significant relationship between enrollment growth and academic performance. So we have models of excellence, and there is no intrinsic reason why they can’t be replicated, but our philanthropy-plus-charter-schools model isn’t managing to do it. Why?

Rather than merely speculate about the cause of our failure and immediately hop on another education reform bandwagon, perhaps it’s time we pull over and look at a map. As it happens, there already are a number of places around the globe where educational excellence is scaling up. Where top teachers use the Web to reach not hundreds or thousands of students but hundreds of thousands. And where they are rewarded for doing so with salaries in the millions of dollars. There are successful networks of schools that have grown not merely to a few dozen schools in a few dozen states, but to tens of thousands of schools in scores of countries.

Why do top teachers in Korea’s for-profit tutoring sector become celebrities who earn more than the nation’s professional baseball players? Why has the Japanese tutoring chain, Kumon, expanded to serve over four million students worldwide? Could it be because the tutoring sector operates within the same free enterprise system that has resulted in the massive scale-up of excellence in every other field? Is it an accident that when we reward education entrepreneurs for their success, their success grows? Could it be that philanthropists have failed to consistently fund the best charter schools because they do not expect a return on their investment, as hard-nosed venture capitalists do?

These questions have obvious, if inflammatory, answers. Until we let education participate in the same free enterprise system that drives the scale-up of everything from iPhones to Facebook, excellent schools and teachers will remain floating candles—beautifully illuminating their immediate vicinity, but doomed never to touch off a wider blaze.