vdem-analysis

name	vdem-analysis
description	Analyze V-Dem (Varieties of Democracy) cross-national panel data to measure democratic indices, detect backsliding, and run panel regressions linking institutional quality to economic outcomes.
tags	["political-science","democracy","panel-data","v-dem","comparative-politics","regression"]
version	1.0.0
authors	[{"name":"awesome-rosetta-skills contributors","github":"@awesome-rosetta-skills"}]
license	MIT
platforms	["claude-code","codex","gemini-cli","cursor"]
dependencies	["pandas>=2.0.0","numpy>=1.24.0","matplotlib>=3.7.0","seaborn>=0.12.0","linearmodels>=5.3.0","scipy>=1.10.0","statsmodels>=0.14.0","requests>=2.31.0"]
last_updated	2026-03-17

V-Dem Analysis — Varieties of Democracy Dataset

Overview

The Varieties of Democracy (V-Dem) project provides the most comprehensive cross-national dataset on democratic institutions, covering over 200 countries from 1789 to the present. This skill covers the full analytical pipeline: loading the large CSV dataset efficiently, computing democratic indices, detecting backsliding episodes, visualizing trends, and running panel regressions.

Key Democratic Indices

Column	Description
v2x_polyarchy	Electoral democracy index
v2x_libdem	Liberal democracy index
v2x_partipdem	Participatory democracy index
v2x_delibdem	Deliberative democracy index
v2x_egaldem	Egalitarian democracy index

All indices are on a 0–1 scale where higher values indicate more democracy.

---

Setup

pip install pandas numpy matplotlib seaborn linearmodels scipy statsmodels requests

Download the V-Dem dataset from https://www.v-dem.net/data/the-v-dem-dataset/ (V-Dem Country-Year Core dataset, CSV format, ~200 MB). Store the path as an environment variable or pass it directly.

---

Core Functions

import os
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import matplotlib.cm as cm
import seaborn as sns
from scipy import stats
from linearmodels.panel import PanelOLS, PooledOLS, RandomEffects
import warnings
warnings.filterwarnings("ignore", category=FutureWarning)

# ---------------------------------------------------------------------------



VDEM_INDICES = [
    ,
    ,
    ,
    ,
    ,
]

ALWAYS_KEEP = [
    ,
    ,
    ,
    ,
    ,          
    ,            
    ,   
]

REGION_LABELS = {
    : ,
    : ,
    : ,
    : ,
    : ,
    : ,
}


 () -> pd.DataFrame:
    
    keep_cols = ALWAYS_KEEP + VDEM_INDICES + (extra_cols  [])

    
    header = pd.read_csv(filepath, nrows=)
    available = [c  c  keep_cols  c  header.columns]

    
    chunks = []
     chunk  pd.read_csv(filepath, usecols=available, chunksize=, low_memory=):
         years:
            chunk = chunk[(chunk[] >= years[]) & (chunk[] <= years[])]
         countries:
            mask = chunk[].isin(countries) | chunk[].isin(countries)
            chunk = chunk[mask]
        chunks.append(chunk)

    df = pd.concat(chunks, ignore_index=)
    df[] = df[].(REGION_LABELS)
    df = df.sort_values([, ]).reset_index(drop=)
     df







 () -> pd.DataFrame:
    
    df = df.sort_values([, ]).copy()
    df[] = df.groupby()[index_col].shift(window)
    df[] = df[index_col] - df[]
    episodes = df[df[] <= -threshold].copy()
    episodes[] = episodes[].()
     (
        episodes[[, , index_col, , , ]]
        .sort_values(, ascending=)
        .reset_index(drop=)
    )







 () -> plt.Figure:
    
    subset = df[df[].isin(countries)].copy()
    fig, ax = plt.subplots(figsize=figsize)

    palette = cm.get_cmap(, (countries))
     i, country  (countries):
        cdata = subset[subset[] == country].sort_values()
        ax.plot(cdata[], cdata[index_col], label=country, color=palette(i), linewidth=)

    ax.set_xlabel(, fontsize=)
    ax.set_ylabel(index_col, fontsize=)
    ax.set_title(title  , fontsize=)
    ax.legend(loc=, fontsize=)
    ax.set_ylim(, )
    ax.grid(, alpha=)
    fig.tight_layout()

     save_path:
        fig.savefig(save_path, dpi=)
     fig


 () -> plt.Figure:
    
    subset = df[df[] == year].dropna(subset=[index_col, ])
    fig, ax = plt.subplots(figsize=figsize)
    order = (
        subset.groupby()[index_col]
        .median()
        .sort_values()
        .index.tolist()
    )
    sns.boxplot(
        data=subset,
        x=index_col,
        y=,
        order=order,
        palette=,
        ax=ax,
    )
    ax.set_title(, fontsize=)
    ax.set_xlabel(index_col)
    ax.set_ylabel()
    fig.tight_layout()
     save_path:
        fig.savefig(save_path, dpi=)
     fig







 () -> :
    
    controls = controls  []
    req_cols = [outcome, treatment] + controls + [, ]
    panel_df = df[req_cols].dropna().copy()

     log_transform_treatment  panel_df[treatment].gt().():
        panel_df[treatment] = np.log(panel_df[treatment])
        panel_df = panel_df.rename(columns={treatment: })
        treatment = 

    panel_df = panel_df.set_index([, ])
    exog_vars = [treatment] + controls
    exog = panel_df[exog_vars]
    endog = panel_df[outcome]

     model_type == :
        model = PanelOLS(endog, exog, entity_effects=, time_effects=)
     model_type == :
        model = RandomEffects(endog, exog)
    :
        model = PooledOLS(endog, exog)

    result = model.fit(cov_type=, cluster_entity=)
     {
        : result.summary,
        : result,
        : result.rsquared,
        : result.params,
        : result.pvalues,
    }







 () -> pd.DataFrame:
    
    indices = indices  VDEM_INDICES
    subset = df  year    df[df[] == year]
    corr_matrix = subset[indices].corr(method=method)

    fig, ax = plt.subplots(figsize=figsize)
    mask = np.triu(np.ones_like(corr_matrix, dtype=), k=)
    sns.heatmap(
        corr_matrix,
        annot=,
        fmt=,
        cmap=,
        vmin=-,
        vmax=,
        mask=mask,
        ax=ax,
        square=,
        linewidths=,
    )
    year_label = (year)  year  
    ax.set_title(, fontsize=)
    fig.tight_layout()
     save_path:
        fig.savefig(save_path, dpi=)
    plt.show()
     corr_matrix