📊 Tobacco Consumption and Household Budget Analysis

🧠 Overview

This project explores the socio-economic and health-related factors influencing household tobacco consumption in the context of a national budget survey. Through a series of descriptive statistics, regression analyses, and diagnostic tests, it identifies key relationships and evaluates model performance while accounting for potential specification errors.

---

📁 Data Sources

• data1.dta: Primary dataset containing household budget data
• health.dta: Health expenditure records to proxy tobacco-related harm

---

📚 Project Goals

• Analyze tobacco consumption across households using economic and demographic indicators
• Determine the statistical significance and magnitude of those indicators using regression modeling
• Address modeling limitations such as normality, heteroscedasticity, and omitted variable bias
• Include health-related proxies to enrich the model

---

🔍 Analysis Workflow

1️⃣ Data Preprocessing

• Loaded .dta datasets using haven
• Identified non-binary variables for statistical analysis
• Merged health data with primary household dataset using hhid

2️⃣ Descriptive Statistics & Correlations

• Summarized income, age, education, unit price, and tobacco consumption
• Notable correlations:
  • Age ↔️ Education: −0.3486
  • Education ↔️ Income: 0.4200
  • Education ↔️ Unit Price: 0.5353
  • Unit Price ↔️ Income: 0.5028

3️⃣ Visual Explorations

• Household Income vs Tobacco Consumption: No strong linear pattern; higher variance for low-income households
• Unit Price vs Tobacco Consumption: Clear inverse relationship, price sensitivity evident

4️⃣ Normality Checks

Used Shapiro-Wilk and histograms for:

• income: Strongly non-normal (p ≈ 0.444)
• unitvalue: Also non-normal distribution

5️⃣ Baseline Regression

weight ~ income + unitvalue + age + female + leduc + own + child_less_14