Measuring the cost of liquidity in agricultural futures markets: Conventional and Bayesian approaches

Estimating the cost of liquidity in agricultural futures markets is challenging because bid‐ask spreads are usually not observed. Based on an ability to reflect simulated data from Roll's spread model, we assess the effectiveness of conventional and Bayesian bid‐ask spread estimators under different market conditions. Conventional serial covariance and absolute price change spread estimators appear to be biased. Hasbrouck's Bayesian estimator generates small costs of liquidity whose values depend on the correlation and noise in the data. The absolute value Bayesian estimator is precise and works well under conditions of high levels of noise and correlation usually found in agricultural futures markets. Using data from live cattle (LC) and lean hog (LH) contracts, we find similar patterns of performance that produce economically meaningful cost of liquidity differences.