https://doi.org/10.5061/dryad.qfttdz0r8

This readme file describes the various attached .do files and .csv files, and how to replicate the analysis in our paper. All .do files were run using Stata 17. Of course, all the analyses can also be run in the open-source software R, after converting the code in the do files. All the variables in the datasets included with the replication files are labeled to provide a description of the data. If you have any questions about these data or .do files, please contact eric.brunner@uconn.edu.

Restricted Use Data

Our research relies on data protected by the Family Educational Rights and Privacy Act (FERPA), and as a result this data cannot be posted on-line or made freely available. Rather, an individual who wishes to replicate or extend this work would need to contact officials at the Connecticut State Department of Education, Labor Department and the state’s P20Win program to obtain permission to use the data for a well-defined purpose. That purpose, like our study, must represent an evaluation of an education program (the only research exception to FERPA restrictions on use of administrative student data without consent of participating students). Further, any analysis of the data must be conducted as an official employee of the state. Luckily, for these purposes, any faculty in part-time residence at the university even as an unpaid visiting faculty is viewed by the state as an employee, and therefore can be approved to work with such data. In the interest of replication, the authors are willing to work with individuals interested in replication or extension providing them with the appropriate contact information within the state, and will facilitate both visiting rights, office space and computer resources during such a visit through our academic departments. In accordance with federal funding guidelines, the state has agreed to archive this data for 10 years following the completion of the project.

Public Use Data Files Included in Data Archive

All datasets are saved as comma separated values (.csv) files. When an observation is missing in a given dataset, a field will contain the value “n/a”. Missing values may arise for several reasons including: 1) a school or district not reporting the relevant information (e.g. test scores or spending) or the number of students that make up the value (e.g. a test score) is too small to report due to confidentiality. 

cutoff_scores.csv: Contains the cut-off scores for all CTHSS schools and all years in our sample. These are merged in with the main data using the program called dataset.do. The fields are: 1) CTECS school ID (numeric), 2) Year Student Applied to CTECS (numeric), 3) cutoff admission score of CTECS school (numeric, points).

district_cmt_gr6.csv: Contains average district-level test scores in math, reading and writing, for all school in Connecticut. This data is used in our balancing tests. The fields are: 1) district id code (numeric), 2) district name (text), 3) district math scores (numeric, points), 4) district reading scores (numberic, points), 5) district writing scores (numeric, points).

school_vars.csv: Contains data on spending per pupil and pupil-teacher ratios for all schools in Connecticut. The fields are: 1) 8th grade school district ID (numeric), 2) 8th grade school ID (numeric), 3) Year Student Applied to CTECS (numeric), 4) school enrollment (numeric, count), 5) school pupil-teacher ratio (numeric, pupils per teacher), 6) school spending per pupil (numeric, dollars).

cmt_scores.csv: Contains 6th grade proficiency rates in math and reading by town. The fields are: 1) application year (numeric), 2) town FIPS id (numeric), 3) CT state town id (numeric), 4) town name (text), 5) percent of students proficient in math (numeric, percentage), 6) percent of students proficient in reading (numeric, percentage). 

cpi.csv: Contains annual and quarterly data on the Consumer Price Index (CPI). The fields are: 1) year (numeric), 2) quarter of year (numeric), 3) CPI (numeric, index).

Do Files

cutoff_scores.do: Program to estimate the cut-off application scores for each CTECS school and year. Specifically, the program looks for discontinuities in the probability of receiving an offer letter to one of the CTHSS schools. The algorithm is based on the work in Porter, J., & Yu, P. (2015). Regression discontinuity designs with unknown discontinuity points: Testing and estimation. Journal of Econometrics, 189(1), 132-147. The program calls the file cthss_scores which was created by running the file dataset.do which creates the clean dataset from the raw administrative data.  */

dataset.do: Program to create main dataset and summary statistics for the analysis in the paper.

estimate.do: This program creates outcomes reported in the paper along with the reduced form and first-stage graphs. The program lists what table is associated with the various analyses at each step.

 create_figures.do: This program creates the figures reported in the paper.

 create_histogram.do: This program creates the histogram of admission scores for students with disabilities and students without a disability.