Statistics and Computational Methods Seminar Series - Fall 2026

Speaker: Giulia Livieri

Title: Testing for endogeneity of irregular sampling schemes

Abstract

In high-frequency financial data, sampling times are often assumed to be independent of the sampled price process. We derive two statistical tests to verify this hypothesis: one suitable for moderate sampling frequencies and one applicable to tick-by-tick data. Using a large dataset of NYSE transaction prices, we document that time endogeneity is frequency and liquidity dependent. Tick-by-tick sampling reveals that trade times are broadly close to exogenous, although a statistically significant excess rejection appears at the lower end of the liquidity distribution. This is consistent with the empirical evidence that information-based trading and information risk become more relevant as trading activity declines. At moderate frequencies business-time sampling produces strong evidence of endogeneity across the liquidity distribution. Monte Carlo simulations provide a micro-founded interpretation of the empirical results, showing how the timing of informed trades can generate endogeneity while uninformed trading may attenuate it in aggregate transaction data.

Link Teams: https://teams.microsoft.com/meet/3173443044780938?p=XH4OsfUVto114RH8zY