Tutoring and GPA in 35 high schools (DiD tutorial)

Downloads

Each dataset is available as a labeled Stata .dta and its source file.

⇩ Download all data (ZIP)stata_codebook.do

Run stata_codebook.do in Stata once to attach long-form per-variable notes to the .dta files.

Load directly in code

Every file loads straight from GitHub (raw URLs). Swap the file name to load any dataset.

Stata

* Stata 14+ : `use` reads an https URL directly
global BASE "https://raw.githubusercontent.com/cmg777/starter-academic-v501/master/content/post/python_did101/data/"
use "${BASE}tutoring_did.dta", clear
describe
notes

Python

!pip install -q pyreadstat
import pandas as pd
BASE = "https://raw.githubusercontent.com/cmg777/starter-academic-v501/master/content/post/python_did101/data/"
df = pd.read_stata(BASE + "tutoring_did.dta")

# load every dataset at once
files = ["tutoring_did", "tutoring_didevent"]
data = {f: pd.read_stata(BASE + f + ".dta") for f in files}

# pyreadstat (richest metadata) reads LOCAL files -> download first
import pyreadstat, urllib.request
urllib.request.urlretrieve(BASE + "tutoring_did.dta", "tutoring_did.dta")
df, meta = pyreadstat.read_dta("tutoring_did.dta")

Copy and paste this snippet in Google Colab app. https://colab.research.google.com/notebooks/empty.ipynb

R

# R : haven::read_dta auto-downloads an https URL
library(haven)
BASE <- "https://raw.githubusercontent.com/cmg777/starter-academic-v501/master/content/post/python_did101/data/"
df <- read_dta(paste0(BASE, "tutoring_did.dta"))

Overview & sources

Companion data for the tutorial Introduction to Difference-in-Differences (DiD) in Python, which uses the simulated case study of Corral and Yang (2024). In one region, 10 of 35 high schools adopt an after-school tutoring program for low-income students at the same time; the other 25 schools never adopt it. The outcome gpa is the school's average GPA of low-income students. In the 2×2 file, treated schools rise from 60.17 to 96.37 (a naive change of 36.20 points) while comparison schools rise from 71.22 to 82.10 (10.89 points), so the DiD estimate of the average treatment effect on the treated (ATT) is 25.315 GPA points (25.32 from the rounded means). The event-study file follows the same 35 schools for 8 periods with adoption in period 5; its event-time coefficients are 0.34, −0.32 and 0.59 before adoption and 25.03 to 25.70 after.

Data sources

Cite this data

Please cite this dataset as follows.

APA

Mendez, C. (2026). Tutoring and GPA in 35 high schools (DiD tutorial) [Data set]. https://carlos-mendez.org/post/python_did101/

Corral, D., & Yang, M. (2024). An introduction to the difference-in-differences design in education policy research. Asia Pacific Education Review, 25(3), 663–672. https://doi.org/10.1007/s12564-024-09959-0

BibTeX

@misc{mendez2026pythondid101,
  author       = {Mendez, Carlos},
  title        = {Tutoring and GPA in 35 high schools (DiD tutorial)},
  year         = {2026},
  howpublished = {\url{https://carlos-mendez.org/post/python_did101/}},
  note         = {Data set}
}

@article{corral2024did,
  author  = {Corral, Daniel and Yang, Minseok},
  title   = {An introduction to the difference-in-differences design in education policy research},
  journal = {Asia Pacific Education Review},
  volume  = {25}, number = {3}, pages = {663--672}, year = {2024},
  doi     = {10.1007/s12564-024-09959-0}
}

Variable explorer search & filter all 8 variables

Type to filter by name or label, or use the chips to filter by type. Each row shows a mini distribution. Click a header to sort.

Cross-file variable index

Which file each variable appears in (● = present).

Construction & formulas

Design variables

• treated = 1 for the 10 schools that adopt the program (id 26–35), 0 for the 25 comparison schools. Fixed within school.
• post = 1 after adoption: time = 2 in the 2×2 file; time ≥ 5 in the event-study file.
• txp = treated × post: the DiD treatment indicator (1 for a treated school in a treated period).
• timeToTreat = time − 5 for treated schools (−4, …, 3; 0 = first treated period); missing for comparison schools, which are never treated.

Estimates the post computes from these files

• 2×2 DiD: (mean treated post − mean treated pre) − (mean comparison post − mean comparison pre) = 36.20 − 10.886 = 25.315 (25.32 from means rounded to two decimals).
• Two-way fixed effects: gpa ~ txp | id + time, standard errors clustered by school: 25.315 (SE 0.585).
• Event study: gpa ~ i(timeToTreat, ref = -1) | id + time on the 280-row file, with comparison schools' missing timeToTreat replaced by a placeholder so they stay in the sample.

The datasets

Switch datasets with the tabs. Each shows the full variable dictionary plus a sortable statistics table with mini distributions and data coverage.

Expand all datasets expand to search (Ctrl/⌘+F) or print across all datasets

Known limitations & caveats

• Simulated, not observed. The schools come from the teaching example of Corral and Yang (2024), built to illustrate DiD mechanics. The R² near 0.99 and a 25-point effect on GPA are far larger than real tutoring programs produce.
• GPA can exceed 100. gpa is described as a 0–100 score, but the simulated event-study values reach 107.68 (the 2×2 file stays below 100, maximum 99.15). Treat it as a continuous score, not a bounded grade.
• timeToTreat is missing by design for the 25 comparison schools. Software that drops rows with missing values (pyfixest, fixest, Stata's i. operator) silently removes every comparison school; fill it with a placeholder or build the event dummies by hand.
• Two independent simulations. The 2×2 file is not periods 4 and 5 of the event-study file, so estimates from the two files differ (25.315 vs. a collapsed 24.897).
• female_share varies within schools over time. It is a time-varying covariate; the post uses it only as a robustness check (it moves the DiD estimate by 0.013).
• CSV values carry full float precision. The original .dta files store gpa and female_share as single-precision floats; the CSVs write those exact values as doubles (17 significant digits). In Stata, read the CSV with import delimited …, asdouble.