Use this Skill for comparative political analysis: QoG, Polity 5, Freedom House merge, cross-sectional OLS with FE, multilevel models, and regime classification.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Une commande directe contourne le prompt de vérification. Examinez la source avant de l'exécuter.
Use this Skill for comparative political analysis: QoG, Polity 5, Freedom House merge, cross-sectional OLS with FE, multilevel models, and regime classification.
Download and merge the Quality of Government (QoG) standard dataset, Polity 5 scores, and
Freedom House annual ratings into a unified country-year panel
Classify countries into regime types (autocracy, hybrid/anocracy, democracy) using Polity 5
thresholds, and track transitions over time
Run panel OLS with country and year fixed effects using the linearmodels package (handles
the Frisch-Waugh-Lovell within transformation efficiently)
Perform Hausman test to choose between fixed effects and random effects models
Detect democratic backsliding as a rolling 3-year change in Polity score
Produce Lipset-style scatter plots (GDP per capita vs. democracy score) with regional coloring
Estimate multilevel models where countries are nested within regions
This skill covers exploratory analysis, static regression, and dynamic models. For time-series
specific methods (cointegration, error correction), use a dedicated time-series skill.
Background
QoG (Quality of Government) is maintained by the University of Gothenburg. The standard
dataset (qog_std_cs_jan23.csv or time-series qog_std_ts_jan23.csv) aggregates ~2,000 variables
from over 100 sources. Key governance variables:
QoG Variable
Description
Source
wbgi_cce
Control of Corruption Estimate
World Bank
wbgi_rle
Rule of Law Estimate
World Bank
wbgi_gee
Government Effectiveness Estimate
World Bank
undp_hdi
Human Development Index
UNDP
wdi_gdpcapcon2015
GDP per capita (constant 2015 USD)
World Bank
Polity 5 (Marshall & Gurr, 2020) codes democracy and autocracy on -10 to +10 scale:
≤ -6: full autocracy
-5 to +5: hybrid (anocracy)
≥ +6: democracy
Special codes: -66 (interruption), -77 (interregnum), -88 (transition) — replace with NA.
Freedom House provides Political Rights (PR) and Civil Liberties (CL) scores (1-7 each,
lower = more free). Combined score (14 = least free, 2 = most free). Status: Free/Partly Free/
Not Free.
Panel regression with FE: The within transformation demeans all variables by entity (country)
mean, removing all time-invariant country characteristics (culture, geography, history). linearmodels
PanelOLS with entity_effects=True implements this. Standard errors should be clustered at the
country level.
Hausman test: Compares FE and RE estimates. Under H0 (RE consistent), both estimators agree.
Rejection of H0 → use FE. linearmodels provides the compare function for this test.
Democratic backsliding: A 3-year rolling decline in Polity score below a threshold (e.g., -3
points). Distinguished from transitions that start from democratic levels (±6 threshold).
Teorell, J. et al. (2023). The Quality of Government Standard Dataset. University of Gothenburg.
Marshall, M.G. & Gurr, T.R. (2020). Polity 5: Political Regime Characteristics and Transitions, 1800–2018. Center for Systemic Peace.
Lipset, S.M. (1959). Some social requisites of democracy. APSR, 53(1), 69-105.
Examples
Example 1: Merge QoG + Polity + FH and Classify Regimes
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
rng = np.random.default_rng(42)
countries_list = [f"Country_{i:02d}"for i inrange(1, 51)]
years_list = list(range(2000, 2023))
records_qog, records_polity = [], []
for ctr in countries_list:
base_polity = rng.integers(-10, 11)
base_gdp = rng.lognormal(9, 1.5)
for yr in years_list:
drift = rng.integers(-1, 2)
polity_val = int(np.clip(base_polity + drift + rng.integers(-1, 2), -10, 10))
records_qog.append({
"country": ctr, "iso3": ctr[:3].upper(),
"year": yr,
"wbgi_cce": rng.normal(0, 1),
"wbgi_rle": rng.normal(0, 1),
"wdi_gdpcapcon2015": base_gdp * rng.lognormal(0, 0.05),
"wdi_pop": rng.lognormal(15, 1),
})
records_polity.append({
"country": ctr, "year": yr, "polity2": polity_val,
"regime_type": classify_regime(polity_val),
})
df_qog_sim = pd.DataFrame(records_qog)
df_polity_sim = pd.DataFrame(records_polity)
panel = build_comparative_panel(df_qog_sim, df_polity_sim)
# Regime distribution over time
regime_counts = panel.groupby(["year", "regime_type"]).size().unstack(fill_value=0)
fig, ax = plt.subplots(figsize=(12, 5))
regime_counts.plot(kind="area", stacked=True, ax=ax, colormap="RdYlGn", alpha=0.8)
ax.set_title("Global Regime Distribution Over Time (Polity 5 Classification)")
ax.set_ylabel("Number of Countries")
ax.set_xlabel("Year")
ax.legend(title="Regime Type")
plt.tight_layout()
plt.savefig("regime_distribution.png", dpi=150)
plt.show()
print("\nRegime counts (latest year):")
print(panel[panel["year"] == panel["year"].max()]["regime_type"].value_counts())
Example 2: Within-Country FE Regression — Corruption and GDP
import numpy as np
import pandas as pd
# Using the simulated panel from Example 1
panel_reg = panel.copy()
panel_reg["log_gdppc"] = np.log(panel_reg["wdi_gdpcapcon2015"].clip(lower=1))
result_fe = panel_fe_regression(
panel_reg,
outcome="wbgi_cce",
predictors=["log_gdppc"],
entity_effects=True,
time_effects=True,
)
print("=== Country+Year FE: GDP per capita → Corruption Control ===")
print(result_fe.summary.tables[1])
print(f"\nWithin R²: {result_fe.rsquared:.4f}")
Example 3: Democratic Backsliding Detection and Trend Plot
