Standardized tests, noise analysis of the tool used to measure it and its effect on public policies. Chilean case

Abstract

Standardized Tests scores are often a main factor in public policy decisions in education. In addition, those countries with voucher systems for the financing, have an influence in the enrollment decisions of the students. However, there are doubts about how large is the effect of the error (noise) considering the signal given by the test. Previous studies have shown that this error has an important effect on the final test scores. This research examines a large data set, using test scores from students in Chile from 2005 to 2008 to understand the presence of noise in testing environments to analyze the presence of error in the results and increase the test-noise discussion to a country with a long history of free school choice policies. This study serves as a contribution, first, because the previous research has been sparse, and because we incorporated previously unused information and a larger, previously unused database. We found that the average signal moves between 45% and 47% for the mathematical test. For the language test, we found smaller values with an average signal between 37.3% and 44.2%.