Tackling boundary effects in nonparametric estimation of intra-day liquidity measures (Q1855636)

The notion of liquidity is a crucial one in economics and finance. The volume, price and time to order execution are the principal factors that contribute to the liquidity of an asset. In this paper the intensity function of the pure trading process as well as liquidity measures based on weighted duration are considered. By simulation study the authors show that standard methods produce estimates of the intra-day liquidity measures that are several affected by a boundary bias. The straightforward adaptations of the variable kernel and the local linear estimation methods that alleviate this problem are proposed. Using financial transactions data it is shown that the proposed estimators are able to detect the striking asymmetry between the open and close of the New York stock exchange trading process. This asymmetry of the trading process is not detectable by standard methods.