Panel Analysis of U.S. Cigarette Consumption

Estimating Price Elasticity with Fixed Effects Models

Econometrics
Panel Data
Statistics
Panel regression modeling of U.S. cigarette demand.
Published

June 5, 2025

Estimating how cigarette prices influence consumption using panel data econometrics

Using a panel dataset covering 46 U.S. states between 1963 and 1992, I estimated cigarette demand while accounting for unobserved state-level differences. The project compared pooled OLS, fixed effects, and random effects models to identify the most appropriate specification for estimating price elasticity.

View Source Code View Full Report ← Back to Projects

Role
Econometrics Analyst

Domain
Public Policy · Economics

Stack
R · plm · Econometrics

Methods
Panel Data · Fixed Effects · Hausman Test

Project Snapshot

Question

How sensitive is cigarette consumption to changes in cigarette prices?

Approach

Compared pooled OLS, fixed effects, and random effects panel models before selecting the appropriate specification through formal statistical testing.

Finding

Statistical tests supported the fixed effects model, suggesting that controlling for unobserved state-specific characteristics produced more reliable estimates of price elasticity.

Overview

Estimating demand elasticity using observational data presents several challenges because states differ in ways that cannot always be directly measured. Factors such as culture, public health attitudes, and smoking regulations may influence cigarette consumption independently of price.

This project applied panel data econometrics to account for these differences by exploiting repeated observations across states and years. By comparing multiple model specifications, the analysis identified the model that best isolated the relationship between cigarette prices and cigarette demand.

The Challenge

Unobserved heterogeneity

States differ in persistent characteristics that may influence smoking behavior but cannot easily be measured directly.

Longitudinal data

Observations were collected repeatedly across both states and years, requiring panel data techniques instead of standard regression.

Model selection

The project compared competing econometric models and formally tested which specification was most appropriate.

Policy interpretation

Results needed to provide economically meaningful interpretations of cigarette price elasticity.

Analysis Pipeline

1
📥

Data Collection

Loaded cigarette consumption data across 46 U.S. states from 1963–1992.

2
🧹

Data Preparation

Prepared panel structure and economic variables for longitudinal analysis.

3
📈

Model Estimation

Estimated pooled OLS, fixed effects, and random effects models.

4
⚖️

Model Comparison

Performed F-tests and Hausman tests to compare competing specifications.

5
📊

Elasticity Estimation

Estimated the relationship between cigarette prices and cigarette demand.

6

Interpretation

Translated econometric results into policy-relevant conclusions.

Panel econometrics workflow for estimating cigarette demand elasticity.

Technical Stack

  • R
  • Panel Data
  • Econometrics
  • plm
  • Fixed Effects
  • Random Effects
  • Hausman Test
  • F Test
  • Elasticity Estimation

What I Built

Panel Data Pipeline

Prepared a longitudinal dataset covering multiple decades of state-level cigarette consumption.

Econometric Models

Estimated pooled OLS, fixed effects, and random effects models to compare alternative specifications.

Model Selection Framework

Applied Hausman and F-tests to determine whether unobserved state-specific effects should be explicitly modeled.

Economic Interpretation

Estimated cigarette price elasticity and discussed the implications for taxation and public health policy.

Key Results

🏛️
46

U.S. states analyzed

📅
30

Years of panel observations

📊
FE

Fixed effects selected through Hausman testing

What I Learned

This project strengthened my understanding of panel data econometrics and the importance of selecting an appropriate model specification. Rather than assuming observations are independent, panel models explicitly account for repeated measurements and unobserved heterogeneity.

I also learned how statistical model selection tests such as the Hausman test provide objective evidence for choosing between competing econometric models instead of relying solely on model fit.

Research Deliverable

The complete report includes the econometric methodology, statistical tests, regression results, and policy discussion.

View Source Code

View Full Report (PDF)

← Back to Projects