Benefit Transfer Equivalence Tests with Non-normal Distributions

Equivalence testing reverses traditional null and alternative hypotheses—welfare estimates are assumed different unless tests demonstrate that the difference is smaller than a specified tolerance limit. Within benefit transfer, researchers have universally used the “two one-sided t-test” (TOST) equivalence test, an approach that is invalid for non-normal welfare distributions. This paper proposes an alternative based on the difference between independent empirical distributions, denoted the “two one-sided convolutions” (TOSC) test. The TOSC permits valid inference for non-normal distributions. Empirical assessments show large divergences between TOST and TOSC p-values when distributions are non-normal—demonstrating the likelihood of erroneous inference under the TOST.