Introduction

Spatial conditionally autoregressive (CAR) models are powerful tools for analyzing state politics. This article explains how hierarchical CAR models can be specified and estimated, focusing on data structures common to multilevel analysis of the U.S. states.

Methodological Approach

Using Monte Carlo simulations, we demonstrate that accounting for geographic correlation improves model efficiency in certain scenarios. We show this by comparing standard approaches with spatial autocorrelation methods.

Empirical Applications

1. Cross-sectional Analysis: A hierarchical CAR model was applied to the classic Statehouse Democracy framework (Erikson et al., 1993).

2. Panel Data Example: The approach successfully replicated Margalit's (2013) study of social welfare policy preferences.

Key Findings

• The CAR model provides better fit for the data compared to standard models
• Some inferences differ significantly when geographic correlation is accounted for

Conclusion

This research highlights the importance of incorporating spatial dimensions into state-level political analysis.