The Thin Reed: Accommodating Weak Evidence for Critical Parameters in Cost-Benefit Analysis

Area of Science:

• Quantitative analysis
• Policy evaluation
• Econometrics

Background:

• Cost-benefit analysis requires quantifying policy impacts.
• Monte Carlo simulation is used to assess prediction certainty.
• Hypothesis testing of parameters can lead to suboptimal prediction.

Purpose of the Study:

• Compare methods for handling parameter estimates in policy analysis.
• Evaluate prediction accuracy in cost-benefit analysis.
• Identify optimal strategies for using regression coefficients.

Main Methods:

• Monte Carlo simulation to compare prediction errors.
• Three methods evaluated: significant coefficients only, all coefficients, shrunk coefficients.
• Analysis focused on minimizing mean squared error of prediction.

Main Results:

• Discarding statistically insignificant coefficients rarely minimizes prediction error.
• Using all estimates improves prediction over using only significant ones.
• Shrinking estimates provides a robust minimization of mean squared error.

Conclusions:

• Shrinking parameter estimates is a robust approach for policy analysis.
• This method is particularly effective when true parameter values may be near zero.
• Routinely shrinking estimates enhances the reliability of cost-benefit predictions.