This article critiques the common practice in political science of using hazard ratios to interpret Cox proportional hazards models. While these are standard statistical tools, they don't clearly address researchers' focus on time-to-event outcomes.
The authors propose methods that instead calculate expected durations and marginal changes in duration for specific covariate shifts within this model framework. These new metrics directly align with political science theoretical interests regarding how long events or processes last under changing conditions.
Using three distinct articles from different subfields, they demonstrate how these duration-based quantities improve interpretation clarity while retaining the core statistical capabilities of standard hazard ratios.
Key Findings:
- Duration-based quantities are more intuitive for understanding time-to-event outcomes in political science contexts
- Hazard ratios remain useful but should be complemented with duration estimates when possible
- The approach significantly enhances communication between researchers and their audiences
Methods Used:
- Reanalysis of existing studies using standard political science datasets
- Emphasis on translating statistical outputs into more accessible language for substantive interpretation
