臺大管理論叢 NTU Management Review VOL.29 NO.1

17 NTU Management Review Vol. 29 No. 1 Apr. 2019 3.2 The Premium for Illiquidity, Estimated Using Cross-Sectional Regressions Our second approach to estimating the illiquidity premium is to run Fama-MacBeth cross-sectional regressions of stock returns on stock illiquidity and other firm characteristics. Specifically, we estimate the following model: R c,j,t = b0 c,t + b1 c,t * ILLIQ c,j,t-2 + b2 c,t * SIZE c,j,t-2 + b3 c,t * B/M c,j,t-n + b4 c,t * SD c,j,t-2 + b5 c,t * R c,j,t-2-4 + b6 c,t * R c,j,t-5-13 + e c,j,t (2) R c,j,t is the (percent) return on stock j in month t for market c ; and ILLIQ c,j,t-2 is the stock j ’s mean-adjusted ILLIQ ratio calculated using the daily return and volume data in the three-month period from month t -4 to t -2. We use the mean value of ILLIQ in each month across all stocks within a market to scale the raw ILLIQ so that stock illiquidity is standardized over time within each market. In equation (2), we control for other firm characteristics that have been shown to predict stock returns including (i) SIZE c,j,t-2 , the logarithm of the stock market capitalization, calculated at the end of month t -2; (ii) B/M c,j,t-n is the lagged book-to-market ratio which is known at the beginning of month t , calculated in accordance with the procedure in Fama and French (1993); (iii) SD c,j,t-2 is the logarithm of the standard deviation of stock returns calculated using daily returns over months t -4 to t -2; (iv) R c,j,t-2-4 ( R c,j,t-5-13 ) are the lagged returns (in decimals) over the preceding three (nine) months from t -4 to t -2 ( t -13 to t -5) to capture the price momentum effect (Jegadeesh and Titman, 1993). We estimate equation (2) in each month within each market using the return-weighted method proposed by Asparouhova et al. (2010) and obtain the time-series average of the estimated coefficients. We have 44 markets in the final sample (excluding Romania), as we require each market to have at least 30 non-missing values of the explanatory variables and at least 36 monthly regressions.