Bootstrap Confidence Intervals and Hypothesis Testing for Market Information Shares

Abstract

Abstract Market information shares are widely used in empirical finance to measure one market’s contributions to price discovery. In contrast to common factor components, the literature on market information shares only provides rudimentary tools to test general hypotheses. Using Monte Carlo simulations, we show that bootstrap confidence bands proposed by Sapp (2002) perform well if markets have similar information shares but are too narrow if one market dominates price discovery. We design a new bootstrap-based method to test the “one-central-market” hypothesis and show that our tests have correct size and substantial power against the null hypothesis. Empirical results in the context of CDS and bonds markets complement the theoretical analysis.