The subtitle of this excellent book, by Daniel Koretz, is What Educational Testing Really Tells Us. Here is one excerpt:
The distressingly large achievement differences among racial/ethnic groups and socioeconomic groups in the United States lead many people to assume that American students must vary more in educational performance than others. Some observers have even said that the horse race — simple comparisons of mean scores among countries — is misleading for this reason. The international studies address this question, albeit with one caveat: the estimation of variability in the international surveys is much weaker than the estimation of averages.
…We are limited to more general conclusions, along the lines of "the standard deviations in the United States and Japan are quite similar." Which they are. In fact, the variability of student performance is fairly similar across most countries, regardless of size, culture, economic development, and average student performance.
I was shocked to read this but the book is highly reputable and persuasive.