The authors test a liquidity-adjusted version of the capital asset pricing model (LCAPM) over a long period (1927–2010), allowing parameters to change over time. The risk premium (RP) charged for illiquidity can be viewed as having three components based on covariances: covariance of individual asset illiquidity and marketwide illiquidity (commonality illiquidity); covariance of individual asset return and marketwide illiquidity (which, if positive, identifies a stock that is a hedge against an illiquid market); and covariance of individual asset illiquidity and market return (which, if positive, identifies a stock that is liquid when the market is performing poorly). In addition, a premium for the level of illiquidity (LP) is estimated as the ratio of the expected cost of illiquidity to the expected holding period.

Both measures, LP and RP, vary over time. LP varies between 1.25% and 1.28% annually, and RP varies between 0.46% and 0.83% annually. These estimates are roughly one-third of the values estimated in previous research, perhaps because the measure of illiquidity used in this study is assumed to be more accurate. The components of RP all contribute positively to RP, but it is the covariance between individual asset illiquidity and the market return that contributes the most (0.38–0.68% annually). LP tends to dominate RP over time by a wide margin, but this margin narrows during times of financial distress; the two measures are positively correlated.

This research will help practitioners recognize the effects of illiquidity on prices (and return) over time and identify when different components of illiquidity are more important. The authors’ empirical demonstration of an illiquidity premium estimate that is approximately one-third as large as previous estimates is also valuable.

From a hedging perspective, separating the RP into components might be an effective way to select assets depending on the type of illiquidity a portfolio needs protection against.

Regressions are performed using measures for LP and the components of RP: asset illiquidity, market illiquidity, asset return, and market return. The authors perform these regressions using monthly data from NYSE/AMEX stocks (1927–2010). The market is defined as an equal-weighted portfolio of eligible stocks (minimum price of $5 and at least 100 days of price data).

Such attributes of illiquidity as cost and expected holding period do not readily transform into frequencies associated with traditional data collection. Illiquidity is measured by using the daily effective spread (a measure based on the quote and the trading price) that is accumulated into an annual measure. The illiquidity measure for the market is also an annual measure and is an equal-weighted combination of the individual security illiquidity measures.

The expected holding period or average holding period is the reciprocal of annual NYSE turnover rates. This measure is calculated during the associated current year for the given data, which assumes that investors know their investment horizon at the point of stock purchase.

The empirical work is impressive and provides interesting insights with regard to the size and variability of illiquidity premiums over time. But I am curious about the requirement of 100 days of price data each year. I understand the necessity for trading activity data, but I suspect that very illiquid securities may have been excluded from the sample. If so, the illiquidity premiums estimated may be biased downward.