@techreport{NBERw6869,
 title = "The Effects of Class Size and Composition on Student Achievement:  New  Evidence from Natural Population Variation",
 author = "Caroline M. Hoxby",
 institution = "National Bureau of Economic Research",
 type = "Working Paper",
 series = "Working Paper Series",
 number = "6869",
 year = "1998",
 month = "December",
 URL = "http://www.nber.org/papers/w6869",
 abstract = {I use natural population variation to identify the effects of class size and composition on student achievement.  I isolate the credibly random component of population variation in each grade and school district and use this component to generate instrumental variables for class size and composition.  I also exploit the discontinuous changes in class size that occur when natural population variation triggers a change in the number of classes in a grade in a school.  Discontinuity-based results are both consistent and precise only when applied to within-district changes in class size and population.  I find that reductions in class size from a base of 15 to 30 students have no effect on student achievement.  The estimates are precise enough to identify improvements in math, reading, or writing achievement of just 3/100ths of a standard deviation.  I find that the presence of black students in a class, in an of itself, has no effect on achievement.  I demonstrate that estimates of the effect of racial composition that rely on between-district comparisons suffer from substantial bias.  Finally, I show that more female classes perform significantly better in writing in the 4th through 8th grades and in math in the 4th grade.  Comparison of the effects to average male-female differences in test scores suggest that gender composition alters classroom conduct.},
}