“Remarkably, we find no association between the estimated gains in achievement and the level scores of countries. At all levels of test scores we observe countries with high achievement gains and countries with low gains in achievement. The lack of association has been found for both tests (math, science) and for both age groups. Hence, assessments of the performance of education systems based on the estimated gains in achievement often are inconsistent with performance assessments based on level scores. This inconsistency might be explained by limitations of both measures. Level scores do not distinguish between the contribution of time in school and the contribution of time out of school. The gain scores only refer to the gain in achievement in the year before the test. The inconsistency of the two measures implies that the benchmarking of education systems based on level scores might yield misleading information about the performance of education systems. Low levels of test scores, or declining trends in test scores, might not be the result of low performing education systems. High levels of test scores could mask low performing education systems. Using gain scores as an additional instrument for the assessment of the performance of education systems is likely to reduce the risk of providing misleading policy information.”