Confidence Intervals for Elasticities and Flexibilities: Reevaluating the Ratios of Normals Case

Many important hypotheses in applied economics depend upon the magnitude of estimated elasticities or flexibilities. However, their statistical properties are unknown for many popular models, making standard statistical inference impossible. This problem is addressed in the present paper which analyzes and evaluates alternative methods of constructing confidence intervals for elasticities and flexibilities. The methods studied include three bootstrap‐based approaches, an approximation based on a Taylor's series expansion, and approaches proposed by Fieller and Scheffé. Results show that all method's except Scheffé's worked reasonably well, but the simpler Fieller and Taylor's series methods modestly outperformed the various bootstrapped‐generated intervals.