Trading Costs and Returns for US Equities: Estimating high-frequency measures of liquidity from low-frequency data

This study examines various measures of trading costs estimated from highfrequency (TAQ) data and the extent to which these measures can be estimated from daily data (CRSP). The high-frequency measures tend to be positively correlated. Posted spreads and effective costs are very highly correlated. Price impact measures and other statistics derived from dynamic models, however, are only modestly correlated. Among the set of proxies constructed from daily data, a Gibbs estimate of the effective cost based on Roll (1984) stands out, achieving a correlation of 0.944 with the corresponding TAQ estimate. A variant of the illiquidity measure suggested by Amihud (2002) is the best proxy for the price impact coefficient (a correlation of 0.721 with the TAQ measure). Both the Gibbs estimate of effective cost and the illiquidity ratio covary positively with risk-adjusted returns, but the relations exhibit marked seasonality and are not robust to the use of alternative measures of correlation. JEL classification codes: C15, G12, G20