Measuring the uncertainty of shadow economy estimates using Bayesian and frequentist model averaging

Abstract

Economic literature provides little discussion on the uncertainty around the macroeconometric shadow economy estimates. We fill this gap by deriving the measurement error of the shadow economy estimates stemming from the model uncertainty by using frequentist and Bayesian model averaging techniques. This allows us to make useful insights into the optimal selection of regressors within the Currency Demand Analysis (CDA) framework, basing on the marginal probabilities that the selected variables are included in the ''true'' model. Hence, we provide the CDA researchers with an additional guidance with respect to the selection of shadow economy determinants that makes CDA-based shadow economy measurements less arbitrary. Our results show that the selection of regressors can have a material and highly country-specific impact on the estimated level of the shadow economy. In consequence, one cannot attribute the same level of uncertainty to every country across the panel. We use our results to demonstrate the average shadow economy estimates as of 2014 for 64 countries, along with the confidence intervals.