Proposition 99 and cigarette sales in 39 US states (synthetic control 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_sc101/data/"
use "${BASE}smoking_sc.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_sc101/data/"
df = pd.read_stata(BASE + "smoking_sc.dta")

# load every dataset at once
files = ["smoking_sc"]
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 + "smoking_sc.dta", "smoking_sc.dta")
df, meta = pyreadstat.read_dta("smoking_sc.dta")

Copy and paste this snippet into an empty Google Colab notebook: 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_sc101/data/"
df <- read_dta(paste0(BASE, "smoking_sc.dta"))

Overview & sources

Proposition 99 raised the cigarette tax in California by 25 cents per pack from January 1989 and funded anti-smoking education. Abadie, Diamond, and Hainmueller (2010), hereafter ADH (2010), evaluate it with a panel of 39 US states from 1970 to 2000. Each row is one state in one year, so the panel has 1,209 rows. California is the treated state, and the other 38 states form the donor pool. The outcome cigsale measures cigarette sales per capita, in packs. Four covariates describe each state: log GDP per capita, the share of the population aged 15–24, the retail price of cigarettes, and beer consumption per capita. The Python tutorial builds a synthetic California from these variables. It estimates that the program reduced sales by 18.98 packs per capita per year over 1989–2000.

Data sources

Cite this data

Please cite this dataset as follows.

APA

Mendez, C. (2026). Proposition 99 and cigarette sales in 39 US states (synthetic control tutorial) [Data set]. https://carlos-mendez.org/post/python_sc101/

Abadie, A., Diamond, A., & Hainmueller, J. (2010). Synthetic control methods for comparative case studies: Estimating the effect of California's tobacco control program. Journal of the American Statistical Association, 105(490), 493–505. https://doi.org/10.1198/jasa.2009.ap08746

BibTeX

@misc{mendez2026pythonsc101,
  author       = {Mendez, Carlos},
  title        = {Proposition 99 and cigarette sales in 39 US states (synthetic control tutorial)},
  year         = {2026},
  howpublished = {\url{https://carlos-mendez.org/post/python_sc101/}},
  note         = {Data set}
}

@article{abadie2010synthetic,
  author  = {Abadie, Alberto and Diamond, Alexis and Hainmueller, Jens},
  title   = {Synthetic Control Methods for Comparative Case Studies: Estimating the Effect of {California's} Tobacco Control Program},
  journal = {Journal of the American Statistical Association},
  volume  = {105}, number = {490}, pages = {493--505}, year = {2010},
  doi     = {10.1198/jasa.2009.ap08746}
}

Variable explorer search & filter all 7 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

How the post builds the predictors

• Covariate averages. For each state, the post averages lnincome, age15to24, retprice, and beer over 1980–1988. Missing years drop out of each average, so the beer average rests on 1984–1988 alone.
• Lagged sales. The columns cigsale_1988, cigsale_1980, and cigsale_1975 repeat the sales of each state in that year in every row of the state. Each one enters the fit as a predictor with its own one-year window.
• Treatment indicator. The column treated equals 1 for California in 1989–2000, which gives 12 rows, and 0 otherwise.

What the post estimates from this file

• Synthetic California. It is a weighted average of donor sales, with nonnegative weights that sum to one. Five states receive positive weight: Utah (0.335), Nevada (0.236), Montana (0.202), Colorado (0.160), and Connecticut (0.068).
• Gap and ATT. The gap in a year is actual sales in California minus synthetic sales. The average treatment effect on the treated (ATT) is the mean gap over 1989–2000, −18.98 packs per capita per year.
• Pre-treatment fit. The root mean squared error of the gaps over 1970–1988 is 1.754 packs per capita.

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

• The variable lnincome measures GDP, not income. Despite its name, lnincome is the log of GDP per capita, according to the original label and Table 1 of ADH (2010). The text and Appendix A of the same paper call it per capita state personal income (logged), so the source itself uses both descriptions.
• The variable age15to24 is a share, not a percent. Its values lie between 0.129 and 0.204, although the original label reads percent. Multiply it by 100 to compare it with the percentages in Table 1 of ADH (2010).
• Covariate coverage is incomplete. The variable lnincome covers 1972–1997, beer covers 1984–1997, and age15to24 covers 1970–1990. The beer average over 1980–1988 therefore rests on 1984–1988 alone. A fake start in 1984 leaves no beer data in its window of 1980–1983, so that in-time placebo fails in the post.
• State coding differs across files. The original .dta file stores state as a numeric code with value labels, and the CSV and the .dta file generated here store the state name as text. In Stata, encode state, generate(state_id) restores the original codes, because encode numbers the names in alphabetical order. California then has code 3, as in the original file.
• Read the CSV at full precision. The original file stores every number in single precision, and the CSV writes these exact values as doubles. Some parsers miss the last digits. The default parser of pandas 3.0.1 changes 1,422 of the 4,797 numeric values, and readr::read_csv 2.1.6 changes 771, each by less than 1e-13. The pandas option float_precision="round_trip" recovers the exact values, and load_data() in the post uses it. The functions read.csv in base R and import delimited in Stata 19 also read them exactly.
• The generated .dta file stores doubles. The renderer writes every number through pyreadstat as a double, including year, whereas the original file stores floats. The values themselves are identical to those of the original file. In Stata, recast float cigsale lnincome beer age15to24 retprice restores the original storage type without changing any value.
• Sales come from tax records. Cigarette sales per capita rest on state tax revenues rather than on surveys of smoking. ADH (2010) note that smuggling across tax jurisdictions affects such data (Section 3.2).
• A second tax increase in 1999. Proposition 10 raised the cigarette tax in California by a further 50 cents per pack in January 1999. Sales in California in 1999 and 2000 therefore reflect both measures.