Ανάρτηση Ερευνητικού Δοκιμίου no 10/26

Ερευνητικό Δοκίμιο no 10/26 με τίτλο " Empirical Likelihood Tests for Pairwise Stochastic Dominance”

των Στέλιου Αρβανίτη, Richard McGee, and Thierry Post

Περίληψη

We develop an empirical likelihood test for pairwise stochastic dominance in weakly dependent time series. The procedure localizes candidate binding constraints as local minima of the empirical integrated distribution gap, then computes a multivariate empirical likelihood ratio over the resulting contact set. Critical values use re-centered pseudo-samples, implemented blockwise for dependence. The statistic is self-normalizing and covariance-adaptive, aggregating joint evidence from localized contacts rather than relying on a supremum or integrated support-wide distance. We prove First order asymptotic validity and show that the null limit is a Gaussian-cone, or chi-bar square, projection governed by the covariance of binding moments. Local-power results show equivalence between empirical likelihood and benchmark distance tests with one binding constraint, but possible power gains with multiple bindings because empirical likelihood exploits contact-set covariance geometry. Monte Carlo designs calibrated to option-enhanced indexation show accurate, stable size, robustness to the localization threshold, feasible calibration, and high power relative to Linton–Maasoumi–Whang and Linton–Song–Whang tests, especially when local violations align with covariance features captured by the statistic. An application to option-strategy indices rejects more economically implausible pairwise dominance relations, consistent with strategies trading off upside participation, option income, and downside protection rather than
dominating each other.

O Στέλιος Αρβανίτης ειναι Καθηγητής του τμήματος Οικονομικής Επιστήμης του Οικονομικό Πανεπιστήμιο Αθηνών,  o Dr. Richard McGee είναι Επίκουρος Καθηγητής στο University College Dublin και ο Τhierry Post είναι Καθηγητής στο Nazarbayev University.