Illiquidity, as a measure of trading cost, can predict stock market returns and real economic activity. The illiquidity measure consists of two components—volatility and a residual. Both components show predictive power in the US stock market, industrial output, and unemployment rate, but the two measures are qualitatively and quantitatively different.
How Is This Research Useful to Practitioners?
Liquidity refers to the ease of converting an asset into its value without loss. The market is generally less liquid when the market fluctuates, the bid–ask spread widens, and the trade size shrinks—all of which mean it costs more in time and money to liquidate an asset. Thus, illiquidity also implies higher transaction costs. The authors find empirical evidence that illiquidity, as a microstructure measure, is significant in predicting stock market returns and predicting macroeconomic activities, including GDP, unemployment rate, consumption, and industrial production level.
The findings are novel in that the illiquidity measure is decomposed into different components, which clarify the predictive content in the aggregate liquidity measure. In particular, stock market volatility is negatively related to liquidity and thus is inevitably embedded as a component of illiquidity. It is unclear which component—volatility or the residual illiquidity—predicts the future stock returns. The authors construct a volatility-adjusted illiquidity measure—that is, separate volatility from illiquidity—to improve the return-forecasting performance. An increase in volatility-adjusted illiquidity is associated with higher expected returns, whereas volatility is negatively related to excess return.
In addition, the illiquidity measure may be affected by structural shifts, such as tick-size reductions in the NYSE in 1997 and 2001. Tick-size reduction can improve liquidity, making the unadjusted illiquidity measure biased in measuring the true liquidity. Therefore, the authors construct a break-adjusted illiquidity measure to isolate the effects of tick reduction, making the illiquidity forecast more robust.
After these adjustments are made, illiquidity is applied to one-month-ahead forecasting for various measures of economic activity: Higher illiquidity levels predict lower future output growth and higher unemployment. In these macroeconomic forecasts, a significant portion of the forecasting power of illiquidity arises from the volatility component.
How Did the Authors Conduct This Research?
The authors construct three categories of variables: illiquidity measures, stock returns and related financial variables, and measures of macroeconomic activity. The illiquidity measures follow a list of illiquidity measures proposed by previous researchers. The stock return measure is the log excess return on the NYSE Index, with the three-month T-bill rate serving as the reference return. Macroeconomic variables include industrial production, unemployment rate, investment, and consumption. The data generally cover the period 1926–2015 monthly and quarterly; some of the data are only available after 1948. Several control variables are included—for example, term spread, default spread, and net equity issuance. In this study, the illiquidity and stock returns are all measured at the aggregate level.
To extract the volatility component from the illiquidity measure, the authors adopt two approaches: analytical-based and regression-based decompositions. The results are similar, but the authors primarily rely on the analytical approach to avoid look-ahead bias.
Several illiquidity measures appear to exhibit structural shifts around tick-size reductions in 1997 and 2001. The authors construct a break-adjusted illiquidity measure to remove the structural shifts in order to estimate the magnitude of level shifts; they compare the illiquidity levels on an ex post basis.
After constructing the volatility-adjusted and break-adjusted illiquidity measures, the authors use regression models to test the predictive ability of illiquidity and its components on stock excess returns. Both in-sample and out-of-sample models are tested, and the results are generally similar. Finally, the predictive power for economic activity is explored. The authors perform a robustness check using different constructs of measures and different model specifications.
The authors shed light on how separate components—namely, volatility and residuals—affect stock returns and economic activity. They discover that the residual component is positively related to excess return, but what factors drive the residual component? Finding out may further improve our knowledge of investment management.
Moreover, the residual component is probably large when the actual bid–ask spread is wide and when the trade size or trade volume is low. In theory, investors require a higher rate of return to compensate for the higher liquidity risk (i.e., the liquidity premium). The premium could also serve as a buffer for higher transaction costs because of low liquidity. Investors should be aware that the excess return in this research is not yet risk adjusted; the excess return may not justify the excess liquidity risk taken. Further research could explore whether it is economically beneficial to bet on this risk.