Conditional Average Treatment Effects (CATE) with Stata 19

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/stata_cate/data/"
use "${BASE}assets3_raw.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/stata_cate/data/"
df = pd.read_stata(BASE + "assets3_raw.dta")

# load every dataset at once
files = ["assets3_raw"]
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 + "assets3_raw.dta", "assets3_raw.dta")
df, meta = pyreadstat.read_dta("assets3_raw.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/stata_cate/data/"
df <- read_dta(paste0(BASE, "assets3_raw.dta"))

Overview & sources

Companion data for a hands-on Stata tutorial on estimating Conditional Average Treatment Effects (CATE) with Stata 19's new cate command. The dataset is the canonical assets3 excerpt — an extract from Chernozhukov & Hansen (2004) that ships with Stata 19 (webuse assets3) — covering 9,913 U.S. households. The outcome is net total financial assets (US$); the treatment is employer-offered 401(k) e401k eligibility (not actual participation); the remaining columns are demographic and financial covariates that describe the heterogeneity of interest. The post contrasts partialing-out (PO) and augmented inverse-probability weighting (AIPW) estimators against a parametric teffects aipw benchmark, then probes heterogeneity with estat heterogeneity, estat projection, GATE on prespecified income groups, GATES on data-driven quartiles, estat classification, and a nonparametric estat series fit. The raw eligible-versus-ineligible gap of $19,557 shrinks to a doubly robust ATE near $8,000, and income emerges as the dominant moderator.

Data sources

Cite this data

Please cite this dataset as follows.

APA

Mendez, C. (2026). Conditional Average Treatment Effects (CATE) with Stata 19 [Data set]. https://carlos-mendez.org/post/stata_cate/

Chernozhukov, V., & Hansen, C. (2004). The effects of 401(k) participation on the wealth distribution: an instrumental quantile regression analysis. Review of Economics and Statistics, 86(3), 735–751. https://doi.org/10.1162/0034653041811734

BibTeX

@misc{mendez2026statacate,
  author       = {Mendez, Carlos},
  title        = {Conditional Average Treatment Effects (CATE) with Stata 19},
  year         = {2026},
  howpublished = {\url{https://carlos-mendez.org/post/stata_cate/}},
  note         = {Data set}
}

@article{chernozhukov2004effects,
  author  = {Chernozhukov, Victor and Hansen, Christian},
  title   = {The effects of 401(k) participation on the wealth distribution: an instrumental quantile regression analysis},
  journal = {Review of Economics and Statistics},
  volume  = {86}, number = {3}, pages = {735--751}, year = {2004},
  doi     = {10.1162/0034653041811734}
}

Variable explorer search & filter all 11 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

The estimand is the Conditional Average Treatment Effect — the average effect of 401(k) eligibility d on net assets y for households with covariate profile x:

• CATE: τ(x) = E{ y(1) − y(0) | x = x } — a function of x, not a single number.
• ATE: ATE = E{ τ(X) } — the CATE averaged over the sample.
• GATE (prespecified group g): τ(g) = E{ Γ_i | G_i = g }, the doubly robust AIPW score Γ_i averaged over households in group g.
• GATES: the same average, but groups are data-driven quartiles of the predicted effect τ̂_i (out-of-sample, cross-fit).
• AIPW score: Γ_i = [ ŷ(1)_i + d_i{y_i − ŷ(1)_i}/f̂_i ] − [ ŷ(0)_i + (1−d_i){y_i − ŷ(0)_i}/(1−f̂_i) ] — doubly robust (consistent if either the outcome model or the propensity model is correct).

Stata 19's cate command estimates τ(x) with cross-fitted lasso nuisance models (the learners for the outcome and treatment functions) plus a generalized random forest for the individual-effect function, and an honest-tree bootstrap for inference. The PO (partialing-out, partial-linear) and AIPW (fully interactive) routes make different model assumptions but both return a per-household τ̂(x_i). Note: the columns in assets3_raw.csv are the raw inputs only; the IATE predictions τ̂_i are produced by the do-file and not part of this dataset.

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

• Observational, not randomized. 401(k) eligibility is not randomly assigned (it depends on the employer the household head chose), so the raw eligible-versus-ineligible asset gap ($19,557) mixes the causal effect with selection; identification rests on unconfoundedness given the covariates.
• Eligibility, not participation. e401k records whether the employer offers a 401(k), not whether the household contributes — a distinct treatment from actual participation.
• Heavy-tailed outcome. Net assets is extremely right-skewed (mean $18,054 vs. median $1,499; max ≈ $1.5M; min ≈ −$502,302 for negative-net-worth households), so the treatment effect almost certainly varies across the distribution.
• String-coded categories. In this CSV the binary covariates are stored as their Stata value-label text (e.g. e401k = "Eligible"/"Not eligible", pension = "Receives pension"/"No pension") rather than 0/1; incomecat alone is numeric (0–4).
• Excerpt. assets3 is a teaching excerpt of the original Chernozhukov & Hansen (2004) data shipped with Stata; it is intended for illustration and may not match the full source extract variable-for-variable.