On the political and socioeconomic geography of violence: Spatial heterogeneity and scale effects in Brazil
Abstract
Using a geographically weighted regression (GWR), Ingram and Marchesini da Costa (World Development, 2019) studied the territorially uneven effects of political and socioeconomic factors on violence in Brazil. This article builds on their work by confirming and extending their main findings using a cloud-based Python environment. Through the lens of a multiscale geographically weighted regression (MGWR) and an updated inference framework, we assess the spatial scale at which political and socioeconomic factors affect violence. We also identify geographical clusters of violence that remain statistically significant, even after considering the effects of political and socioeconomic factors.
🤖 AI Podcast Summary
💻 Replication Notebooks
All analyses are fully reproducible in cloud-based Jupyter notebooks via Google Colab:
📈 2. Ordinary Least Squares (OLS)
🗺️ 3. Geographically Weighted Regression (GWR)
🔄 5. Comparing GWR vs MGWR Coefficients
🌍 Introduction
- Territorial violence in Brazil exhibits strong spatial patterns.
- Builds on Ingram & Marchesini da Costa (2019) using MGWR.
- Goal: Explore spatial heterogeneity & scale effects on lethal violence.
🧪 Methodological Innovations
- Used cloud-based computational notebooks for full replication and open science.
- Adopted Multiscale Geographically Weighted Regression (MGWR).
- Identified persistent geographical violence clusters.
- Applied a multiple testing correction to ensure robustness.
📊 Data Overview
- Unit of analysis: 5,562 municipalities (2007–2012)
- Dependent variable: Change in homicide rate (Δ 2011–2012 vs. 2007–2008)
Political Variables
- Margin of victory (%)
- Party alignment with state governor
- Vote abstention (%)
- Mayor’s party identification:
- Brazilian Democratic Movement Party (PMDB)
- Brazilian Social Democracy Party (PSDB)
- Workers' Party (PT)
Socioeconomic Variables
- Population density
- Young male population (%)
- Gini index (income inequality)
- Human Development Index (HDI)
- Households headed by single mothers (%)
- Adult employment rate (%)
- Bolsa Família eligibility (%)
🧭 Modeling Frameworks
- OLS: Global effect estimates.
- GWR: Localized effect estimates with a single spatial scale.
- MGWR: Localized effect estimates with multiple spatial scales.
🔍 OLS Results
- PMDB mayors → Increased violence.
- PT & PSDB → No consistent effect.
- Vote abstention → Strongly linked to higher homicide rates.
- Unexpected: GINI index showed negative correlation.
🗺️ GWR Results: Political Variables
- PMDB: Positive correlation in the northeast.
- PT: Violence-reducing in many regions.
- PSDB: Mixed effects—north (↑), south (↓).
- Abstention: Violence-increasing in several regions.
🧮 GWR Results: Socioeconomic Variables
- Population density & Bolsa Família → Heterogeneous effects.
- Young male % & single mothers → Generally increased violence.
- Effect direction & significance vary spatially.
🗺️ MGWR Results: Political Variables
- PMDB still ↑ violence in northeast.
- PSDB now only ↓ violence in southern Brazil.
- PT effect mostly disappears after statistical corrections.
- Abstention: Consistent with GWR in some regions.
🌆 MGWR Results: Socioeconomic Variables
- Bolsa Família: No significant impact.
- Young male %: Significant across more areas due to large spatial scale.
- Population density & single mothers: Small-scale, heterogeneous influence.
📌 Mapping Residual Clusters
- Intercept mapping (MGWR): Reveals unexplained clusters.
- Central Brazil: Positive residuals → unobserved structural factors?
- South & Central-East: Negative residuals.
🧩 Conclusion
- MGWR provides nuanced spatial insights.
- Confirms heterogeneity in violence determinants.
- Suggests need for regional, tailored policy interventions.
- Further investigation needed into residual violence clusters.
Carlos Mendez
Associate Professor of Development Economics
My research interests focus on the integration of development economics, spatial data science, and econometrics to understand and inform the process of sustainable development across regions.