Appendices to : “ Does Competition among Public Schools Benefit Students and Taxpayers ?

Each issue of the American Economic Review includes a paragraph describing the " Policy on Data Availability: " " …to publish papers only if the data used in the analysis are clearly and precisely documented and are readily available to any researcher for purposes of replication. " Despite repeated requests over several years, Hoxby has not provided the data used to generate the results in her published paper (Hoxby, 2000). She has, however, provided a corrected data set (Hoxby, 2004a) to the National Center of Educational Statistics, which will distribute it to researchers with licenses for the confidential version of the National Educational Longitudinal Survey. I refer to these data throughout as the " Hoxby/NCES data. " Hoxby (2004c) notes several differences between the data used in her paper and the corrected Hoxby/NCES data. First, where she originally used MSA codes from the School District Data Book (SDDB) to assign school districts to metropolitan areas, she now uses the Common Core of Data (CCD), with a few manual corrections, for this purpose. Second, " [t]he published version of the paper showed a version of the first stage regression that was outdated. The paper went through multiple revisions and Table 2 evidently was not updated so that it corresponded exactly with Tables 3+. " As I document in the main text, the actual first stage regression is substantially different than that indicated in the published paper. Hoxby constructs her new district-MSA crosswalk from the Common Core of Data agency files, using a field on these files that indicates the MSA in which each district is located. Unfortunately, there are substantial errors in this field. Many of these come from a small number of MSAs whose boundaries have changed over time; the CCD often contains ii obsolete MSA codes for districts in these areas. 1 One example is the " Kansas City, MO-KS MSA, " which was coded 3760 in 1990 but had been divided into separate Missouri (code 3760) and Kansas (code 3755) components in 1983. Even the 1991-92 CCD, the most recent used for Hoxby's crosswalk, reports the obsolete 3755 code for 19 districts. Hoxby corrects six of these in Johnson County, Kansas, but retains the incorrect code for 13 districts in other counties. 2 Another common occurrence is a metropolitan district for which the CCD is simply missing an MSA code. One such district is the Collier County School …