Executive summary: The cost-effectiveness of charitable interventions appears to follow a power law distribution, as confirmed by fitting global health intervention data to a Pareto distribution with α ≈ 1.11, though it also fits a log-normal distribution; estimation errors may bias the observed α, and further data collection is needed to clarify the true distribution.
Key points:
Cost-effectiveness estimates from the Disease Control Priorities 3 (DCP3) report fit a power law distribution with α = 1.11, as confirmed by a Kolmogorov-Smirnov goodness-of-fit test (p = 0.79).
The data also fits a log-normal distribution, and distinguishing between the two requires more tail-end observations.
Estimation errors introduce bias, generally making the tail appear fatter and underestimating α, but the effect depends on error magnitude.
Simulation tests suggest that the true α may be closer to 1.15 if estimation error is around 50%, or as high as 1.8 with 100% estimation error.
A more comprehensive database covering multiple intervention types (beyond global health) is needed to refine cost-effectiveness distribution analysis.
Future work should focus on obtaining more extreme cost-effectiveness estimates to clarify whether a power law or log-normal distribution best describes the data.
This comment was auto-generated by the EA Forum Team. Feel free to point out issues with this summary by replying to the comment, and contact us if you have feedback.
Executive summary: The cost-effectiveness of charitable interventions appears to follow a power law distribution, as confirmed by fitting global health intervention data to a Pareto distribution with α ≈ 1.11, though it also fits a log-normal distribution; estimation errors may bias the observed α, and further data collection is needed to clarify the true distribution.
Key points:
Cost-effectiveness estimates from the Disease Control Priorities 3 (DCP3) report fit a power law distribution with α = 1.11, as confirmed by a Kolmogorov-Smirnov goodness-of-fit test (p = 0.79).
The data also fits a log-normal distribution, and distinguishing between the two requires more tail-end observations.
Estimation errors introduce bias, generally making the tail appear fatter and underestimating α, but the effect depends on error magnitude.
Simulation tests suggest that the true α may be closer to 1.15 if estimation error is around 50%, or as high as 1.8 with 100% estimation error.
A more comprehensive database covering multiple intervention types (beyond global health) is needed to refine cost-effectiveness distribution analysis.
Future work should focus on obtaining more extreme cost-effectiveness estimates to clarify whether a power law or log-normal distribution best describes the data.
This comment was auto-generated by the EA Forum Team. Feel free to point out issues with this summary by replying to the comment, and contact us if you have feedback.