Controlling for Thin Trading in Return Regressions
Summary
The discussion distinguishes stock size from trading liquidity when adjusting return regressions. The SMB factor captures exposure associated with small firms, but it is not a direct measure of infrequent trading: a large company can still have a thinly traded stock. Adding SMB alone therefore may not control for liquidity risk.
For a Fama–French-style analysis, the answer points to a liquidity factor based on order-flow information, following Pastor and Stambaugh. It describes constructing portfolios of more and less liquid stocks and using their return difference to represent liquidity exposure. A time-series dummy for illiquid periods may be confounded by other causes of those periods. The discussion also mentions measuring the share of illiquid stocks or using continuous liquidity data in cross-sectional work. It offers methodological suggestions rather than empirical comparisons, and does not specify a single best procedure for every dataset or regression.
Key ideas
- SMB controls for size exposure, which is distinct from thin trading.
- A stock can be large while still trading infrequently.
- A Pastor–Stambaugh-style factor uses order-flow information to measure liquidity risk.
- Illiquidity indicators may coincide with other conditions that affect returns.
- Continuous liquidity measures can preserve information that discrete portfolio buckets discard.
Tags
Full text
# Adjust regression for thin trading # Adjust regression for thin trading What procedures can I apply to control in a regression on company returns for thinly traded stocks? Is the inclusion of the SMB-factor a potential approach? Or just a dummy variable indicating if a stock is traded infrequently? Thanks for your help! ## Answer by John (score 3, accepted) https://quant.stackexchange.com/a/20899 SMB is controlling for small stocks. Small and thinly traded are not equivalent. For instance, for most of its history, Berkshire Hathaway was a large stock, but thinly traded (b/c of its high price). There are a number of ways to handle liquidity risk. If you're looking to supplement a Fama-French regression, Pastor and Stambaugh (2003) uses order flow information to construct a liquidity factor. Both authors link to the factor data on their websites, so it may just be easier to do that. Your point about using a dummy variable may not work well for a time series regression. This is because if you're identifying the periods where the stock is illiquid there might be other reasons why it is illiquid. I did see a paper by Bakaert et al that looked at something like the % of stocks that are illiquid, which might be similar to your dummy approach. The benefit of the Pastor-Stambaugh approach is that it is in line with Fama-French in constructing portfolios from the most liquid and least liquid stocks and looking at the returns. You could also do something similar on a cross-sectional basis, but you may as well use the continuous data instead of converting it into discrete buckets.
Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.