Hoxby [2000] developed a quasi-experiment in order to address this potential bias. She observed that the timings of births in any given school area fluctuate randomly. Because of these fluctuations, kindergarten classes are larger in some years than in others. Although many factors determine whether a child attends s large or small kindergarten class, the variation in births from year to rear represents a random component of this outcome. Hoxby therefore relied on the instrumental variables method, which takes advantage of the random determinant of class size to identify the effect on test scores. She used random fluctuations in enrollment year-to-year as an instrumental variable. This measure is correlated with class size; but does not directly influence test scores. Hoxby found that class size does not have a discernible effect on test scores.