social-science-economics

Social Science and Economics

Overview

This skill covers quantitative social science and economics methods: from survey instrument validation and regression diagnostics to causal program evaluation and administrative data linkage. For causal identification strategies (IV, DiD, RDD), also see the causal-inference skill.

When to Use This Skill

• Analyzing survey data (Likert scales, factor analysis, reliability)
• Running econometric regressions with proper diagnostic tests
• Working with panel/longitudinal data
• Evaluating policy interventions or field experiments
• Merging administrative datasets with fuzzy matching

---

1. Survey Data Analysis

Likert Scale Handling

import pandas as pd
import numpy as np
import pingouin as pg

# Likert scales are ORDINAL — treat carefully
# 5-point scale: 1=Strongly Disagree, 5=Strongly Agree

# ❌ Wrong: treating Likert as continuous without justification
# ✅ Better: report median and IQR; use non-parametric tests

likert_data = pd.DataFrame({
    "Q1": [3, 4, 5, 2, 4, 3, 5, 4, 3, 2],
    "Q2": [4, 4, 5, 3, 5, 4, 4, 5, 3, 3],
    "Q3": [2, 3, 4, 2, 3, 3, 4, 3, 2, 2],
    "Q4": [3, 4, 4, 2, 4, 3, 5, 4, 3, 2],
})

# Summary statistics
print(likert_data.describe())
print("\nMedians:")
print(likert_data.median())

# Compare two groups: Mann-Whitney U (non-parametric)
from scipy.stats import mannwhitneyu
group_A = likert_data["Q1"][:5]
group_B = likert_data["Q1"][5:]
stat, p = mannwhitneyu(group_A, group_B, alternative="two-sided")
print(f"Mann-Whitney: U={stat}, p={p:.4f}")

Internal Consistency: Cronbach's Alpha

# Cronbach's alpha: measures how consistently items measure the same construct
# α ≥ 0.90: excellent, 0.80-0.89: good, 0.70-0.79: acceptable, < 0.70: questionable

alpha_result = pg.cronbach_alpha(data=likert_data)
print(f"Cronbach's α = {alpha_result[0]:.3f} (95% CI: {alpha_result[1]})")

# Item-total correlations: identify items that don't fit
for col in likert_data.columns:
    rest = likert_data.drop(columns=[col])
    r = likert_data[col].corr(rest.sum(axis=1))
    print(f"{col}: item-total r = {r:.3f}  {'⚠️ low' if r < 0.3 else '✅'}")