Micro(structure) before Macro?

Yong Chen is Associate Professor of Finance at Mays Business School of Texas A&M University, Gregory W. Eaton is Assistant Professor of Finance at Spears School of Business of Oklahoma State University, and Bradley S. Paye is Assistant Professor of Finance at Pamplin College of Business of Virginia Tech. This post is based on a their recent article, forthcoming in the Journal of Financial Economics.

In our article, Micro(structure) before Macro? The Predictive Power of Aggregate Illiquidity for Stock Returns and Economic Activity (Journal of Financial Economics, 2018, 130 (1), pp. 48-73), we provide new, relatively comprehensive empirical evidence concerning the predictive content of aggregate illiquidity for stock returns and macroeconomic activity. Liquidity conditions in securities markets fluctuate over time, which has important implications for equity markets and the broader economy. Theoretical models link aggregate (or market-wide) liquidity with time-variation in the equity premium. In the model of Acharya and Pedersen (2005), for example, persistent variation in trading costs implies that liquidity forecasts future stock returns. A related theoretical literature emphasizes connections between financial frictions and macroeconomic activity. Illiquidity, coupled with other financial frictions, can generate nonlinear amplification effects and exacerbate economic downturns (Brunnermeier and Pedersen, 2009). This theoretical literature suggests that empirical measures of aggregate illiquidity should contain predictive signals for future stock returns and economic activity.

We construct a variety of measures of market frictions at monthly and quarterly frequencies using Center for Research in Security Prices (CRSP) daily data for stocks listed on the NYSE. These time series cover a 90-year period from 1926–2015 and include multiple proxies for the effective spread, several alternative measures of price impact, and a measure of aggregate price delay. Since our measures relate positively to transaction costs, we refer to them as “illiquidity” measures.

Aggregate illiquidity contains a cyclical component. Liquidity tends to erode during economic recessions, and markets are considerably more illiquid during both the Great Depression and the 2007–2009 financial crisis. We propose two key adjustments to conventional aggregate illiquidity measures. Since we demonstrate that many measures of liquidity are log-linear in stock return volatility, our first adjustment decomposes the illiquidity measures into a component reflecting aggregate volatility and a residual. Removing the embedded volatility component ensures that effects driven by market volatility are not falsely attributed to market illiquidity. We find that a simple measure of illiquidity based on the frequency of zero return days, proposed by Lesmond, Ogden, and Trzcinka (1999), can be interpreted as a liquidity measure that is already volatility-adjusted. After extracting a volatility component, other common spread measures are highly correlated with the zeros measure in the aggregate.

The second adjustment we make corrects illiquidity measures for sharp downward structural shifts related to NYSE minimum tick-size reductions in 1997 and 2001. Existing evidence regarding the impact of tick-size reductions on market liquidity is mixed. This evidence indicates that tick-size reductions lead to decreases in quoted and effective spreads, but also to associated reductions in depth. Consequently, the net effect on transaction costs is unclear and potentially varies across alternative classes of investors. For example, Jones and Lipson (2001) find that total trading costs for institutional investors executing large trades increased following the 1997 reduction. Stark decreases in aggregate spread proxies around tick-size reductions likely overstate any true liquidity improvements. Consequently, we adjust illiquidity measures for breaks using several alternative approaches, including “real time” approaches that mitigate concerns regarding spurious forecasting power attributable to look-ahead bias. We illustrate the empirical importance of our two adjustments by examining the predictive content of adjusted versus unadjusted illiquidity measures for excess stock market returns and various measures of economic activity.

We find little evidence that unadjusted illiquidity measures predict stock market returns. However, weak forecasting performance is potentially attributable to level shifts in illiquidity. Indeed, we obtain stronger evidence of stock return forecasting power for break-adjusted measures. The zeros measure of transaction costs (Lesmond, Ogden, and Trzcinka, 1999) significantly forecasts stock market returns and performs best among break-adjusted measures. The zeros proxy is essentially uncorrelated with market volatility. This suggests that more consistent evidence of stock return forecasting power will obtain for volatility-adjusted illiquidity measures. Consistent with this intuition, we find that many break- and volatility-adjusted illiquidity measures significantly forecast returns.

The predictive power associated with several break- and volatility-adjusted trading costs measures is economically significant. One-month-ahead predictive regressions for excess returns based on adjusted trading cost measures yield in-sample R-squared values of approximately 0.5–1.5%. This compares favorably to R-squared values for many traditional predictors over our sample period, including the short-term interest rate, term spread, default spread, and popular financial ratios such as the book-to-market ratio and earnings-to-price ratio. Break- and volatility-adjusted illiquidity measures remain statistically significant predictors of stock returns in multivariate predictive regressions that include an array of traditional return forecasting variables. Out-of-sample return forecasting results confirm in-sample evidence of predictive power, and the adjusted illiquidity measures generally outperform both the historical average and other common return forecasting variables in out-of-sample settings.

Turning to measures of economic activity, we find strong evidence that aggregate liquidity predicts economic activity, controlling for past activity, corroborating Naes, Skjeltorp, and Odegaard (2011). Consistent with intuition, more illiquid markets predict lower future output growth and higher future unemployment. Adjusting for structural shifts generally improves forecasting performance. Most series that exhibit a downward break following tick-size reductions only predict economic activity after break adjustment. Upon decomposing illiquidity measures into a volatility component and a residual, we find that a portion of illiquidity measures’ predictive power for economic activity derives from the volatility component. Although the residual component of illiquidity possesses weak predictive power in univariate models, conditioning on a wider information set produces stronger evidence of marginal predictive power for volatility-adjusted measures. Consequently, we conclude that both components of illiquidity are informative concerning future economic activity.

In summary, we find strong and robust evidence that volatility- and break-adjusted illiquidity measures forecast stock market returns. In addition, we find that both the volatility and residual components of illiquidity contain information regarding future real economic activity. Our research provides new and important information to the investment community, who can incorporate our forecasting results into the investment decision-making process, and regulators who monitor and shape the information environment and well-being of financial markets.

The complete article is available here.

References

Acharya, V.V., Pedersen, L.H., 2005. Asset pricing with liquidity risk. Journal Financial Economics 77(2), 375–410.

Brunnermeier, M.K., Pedersen, L.H., 2009. Market liquidity and funding liquidity. Review of Financial Studies 22(6), 2201–2238.

Jones, C.M., Lipson, M.L., 2001. Sixteenths: direct evidence on institutional execution costs. Journal of Financial Economics 59(2), 253–278.

Lesmond, D.A., Ogden, J.P., Trzcinka, C.A., 1999. A new estimate of transaction costs. Review of Financial Studies 12(5), 1113–1141.

Næs, R., Skjeltorp, J.A., Ødegaard, B.A., 2011. Stock market liquidity and the business cycle. Journal of Finance 66(1), 139–176.

Both comments and trackbacks are currently closed.