Why Hasn’t the Liquidity Risk Factor Caught On More?

On this episode of The Long View, economist Lubos Pastor, whose research focuses on financial markets and asset management, discusses measuring liquidity at the individual stock level.

Here are a few excerpts from Pastor’s conversation with Morningstar’s Christine Benz and Jeff Ptak.

Benz: You referenced factors at the outset of our conversation, and your most cited paper is on liquidity as a factor. Can we start out by discussing what a factor is, how you would define it, and also on the flip side, what’s something that someone might think of as a factor but really isn’t?

Pastor: I think of a factor as an economic variable that moves over time and captures common variations in returns. So, it’s a variable that is correlated with the performance of many assets. Or put differently, maybe slightly more technically, there are many assets that have significant betas, or significant loading, or significant exposures to this variable. So, that’s what I think of as a factor.

You mentioned liquidity as a factor. When liquidity evaporates from the market, many assets lose market value. So, liquidity is highly correlated with asset returns, and we find that that is the case above and beyond any correlation with the stock market as a whole because the stock market also falls when liquidity evaporates. But many assets have a beta with respect to liquidity even after controlling for their exposure to the stock market. So, that’s a factor. I think it’s also very important to distinguish just a factor from a priced factor because not every factor is priced. And a priced factor is a factor that has a risk premium attached to it. So, a priced factor is one where assets that have high betas with respect to this factor have higher average returns going forward. Assets that have low betas have lower returns going forward. And that is precisely what we find about our liquidity factor, that assets that have higher liquidity betas have higher risk-adjusted returns going forward.

Ptak: That’s helpful. Maybe at this point it makes sense to talk about liquidity and how you defined it for purposes of the paper that you wrote and also, maybe how sensitive the findings were to one’s definition of liquidity.

Pastor: Rob Stambaugh and I designed this liquidity factor many years ago, I think 20 years ago. And we measured liquidity at the individual stock level by essentially trying to measure price impact. Because we believe that that’s the dimension of liquidity that matters the most to institutional investors. How much do you move the price when you trade a given amount? Specifically, we designed a regression model where we asked, suppose you trade $1 million worth of a stock, and you move the price. How big is the reversal after that price impact the following day? So, the idea is if you sell $1 million worth of the stock, you depress the price because some market maker has to provide liquidity. They have to get compensated. So, the price drops temporarily. But then there will be a reversal going forward so that the market maker can actually get paid in expectation for providing liquidity.

We measure that temporary reversal in price that results from a $1 million trade. And then, we average this quantity across many stocks, across all stocks that are trading in the marketplace to get a marketwide measure of liquidity. And when you plot it, it looks like liquidity. Most of the time, it’s just fine. It’s not doing much. But now and then, it dries up. Now and then, there’s a spike in liquidity where liquidity evaporates from the market. These are moments like October ‘87, ‘98, 2008. So, that’s what we do.

You also asked about sensitivity to other definitions of liquidity. We haven’t really tried many in the paper. We’ve tried a few to show that they don’t work. It’s very important what you use as a measure of liquidity. Some obvious candidates are not very good when you’re trying to construct the liquidity factor. For example, people like to use trading volume as a metric for liquidity. The idea being when there’s a lot of trading, liquidity is high. And I think that’s a good idea if you’re looking across assets. But it’s not a good idea to do this when you look across time. Trading volume is not a good measure of liquidity over time. A perfect counter example is October 1987 when we had a total evaporation of liquidity, and yet we had a record high trading volume in the market. So, price impact was very high; at the same time, the trading volume was very high. It was basically one-sided volume. Long story short, it actually matters a lot how you define liquidity, and if you do it the way we do it by looking at price impact, then you find that liquidity is priced.

Benz: Why hasn’t liquidity caught on more? And if it did, is there a reason to believe that investors would love it to death and arbitrage away whatever premium might have formerly been available?

Pastor: Good question. Why hasn’t it caught on more? I don’t know. I am in the business of academic research. So, I’d love to know why it hasn’t caught on more. But as far as it being arbitraged away, there are signs that it has not been arbitraged away. We wrote our paper 20 years ago. We published it in 2003. Our dataset ended in December 1999. We found that there was a liquidity risk premium, a significant one, in that period, 20th century essentially. Since then, we and others have shown that the liquidity risk premium has persisted out of sample. So, going forward, if you add almost 20 years of data, you find that the result continues. In fact, the liquidity risk premium has been slightly larger out of sample. So, high-liquidity beta starts continue having higher average returns out of sample. So, it may have been arbitraged away, but it wasn’t. It was a possibility, but that possibility was not realized.

I will also mention that even though we designed our liquidity measure 20 years ago before the financial crisis of 2008, it was good and comforting to see that this measure, the same measure, actually successfully captures the large drops in liquidity in 2008. So, when you plotted our measure in 2008, which is after it was designed, you see big drops in liquidity. It’s doing what it’s supposed to be doing.