A two-stage random-effects estimator for meta-analyses of the value per statistical life

Area of Science:

• Econometrics
• Risk Analysis
• Statistical Modeling

Background:

• Synthesizing value per statistical life (VSL) estimates is crucial for policy-making.
• Existing meta-analysis methods may not adequately address complex error structures in VSL data.
• Unbalanced panels and diverse error sources (sampling and non-sampling) pose challenges.

Purpose of the Study:

• To develop and evaluate a novel two-stage random-effects meta-analysis estimator for VSL.
• To assess the estimator's performance using Monte Carlo simulations.
• To compare the proposed estimator against existing meta-analysis techniques.

Main Methods:

• Developed a two-stage random-effects meta-analysis model.
• Employed Monte Carlo simulations with constructed datasets to test performance.
• Differentiated between within-group and between-group sampling and non-sampling errors.
• Applied the estimator to a meta-dataset of 113 VSL estimates.

Main Results:

• The proposed meta-estimator performs best with homogeneous within-group non-sampling error variances in modest-sized datasets.
• This homogeneity assumption enhances statistical efficiency by preserving degrees of freedom.
• The estimator demonstrates favorable performance compared to other common meta-analysis approaches, including other two-stage methods.
• The method was successfully applied to a large VSL meta-dataset.

Conclusions:

• The developed two-stage random-effects meta-analysis estimator is a robust tool for synthesizing VSL estimates.
• It effectively handles unbalanced data and complex error structures.
• The approach offers improved statistical efficiency and accuracy in meta-estimations of VSL.