Approaches to modeling market liquidity

Abstract

Intraday data was collected for the U.S. Financial Sector ETF and two of its component stocks: Bank of America and Citigroup. The analysis period includes 32 trading days, ranging from February 4, 2019 to March 20, 2019. From this trade and quote data, we construct 2,496 five-minute aggregate time bars for each security and calculate a series of spread, volume, depth, trade count, and price change liquidity measures. We examine the summary statistics of these liquidity measures before applying principal components analysis to them through which we identify key liquidity dimensions in each security and common liquidity dimensions across them. Vector autoregressive models are applied to these principal component scores in order to gain further insight into their time series structure and the ways in which the measures interact over a 32-day period. Finally, the same methodology of principal components and time series analyses are applied to daily-normalized liquidity measures in order to better understand the intraday, rather than multi-day, dynamics of liquidity.