Original paper
Abstract
The Critical Finance Review commissioned Li, Novy-Marx, and Velikov (2017) and Pontiff and Singla (2019) to replicate the results in Pastor and Stambaugh (2003). Both studies successfully replicate our market-wide liquidity measure and find similar estimates of the liquidity risk premium. In the sample period after our study, the liquidity risk premium estimates are even larger, and the liquidity measure displays sharp drops during the 2008 financial crisis. We respond to both replication studies and offer some related thoughts, such as when to use our traded versus non-traded liquidity factors and how to improve the precision of liquidity beta estimates.
Keywords:Â liquidity, liquidity risk, liquidity factor, liquidity beta
Trading rules
- Choose the largest 3,500 firms on NYSE, AMEX, and NASDAQ by market capitalization.
- Ensure the stock price per share is no less than $2.
- Maintain a minimum market capitalization threshold of $10 million for inclusion.
- Exclude REITs, ADRs, ETFs, warrants, Americus Trust Components, and closed-end funds.
- Segment these stocks into four groups based on their market capitalization.
- Focus on stocks within the smallest market capitalization group.
- Further categorize these stocks into quartiles using their past year's turnover.
- Initiate long positions in the least traded stocks.
- Begin short positions in the most frequently traded stocks.
- Annually adjust the portfolio composition (every December).
- Weight stocks equally in the portfolio.
Python code
Backtrader
import backtrader as bt
class LowLiquidityStrategy(bt.Strategy):
def __init__(self):
self.rank_market_cap = None
self.rank_turnover = None
def next(self):
if self.datetime.date().month == 12 and self.datetime.date().day == 31:
# Filter stocks by market cap and per-share price
filtered_stocks = [d for d in self.datas if d.close[0] >= 2 and d.market_cap >= 10000000 and not d.is_excluded]
# Sort by market cap and select top 3500
sorted_by_market_cap = sorted(filtered_stocks, key=lambda d: d.market_cap, reverse=True)[:3500]
# Divide stocks into quartiles by market cap
quartile_size = len(sorted_by_market_cap) // 4
lowest_market_cap_quartile = sorted_by_market_cap[-quartile_size:]
# Divide lowest market cap quartile stocks into quartiles based on turnover
sorted_by_turnover = sorted(lowest_market_cap_quartile, key=lambda d: d.turnover)
turnover_quartile_size = len(sorted_by_turnover) // 4
lowest_turnover_quartile = sorted_by_turnover[:turnover_quartile_size]
highest_turnover_quartile = sorted_by_turnover[-turnover_quartile_size:]
# Go long on stocks with the lowest turnover
for data in lowest_turnover_quartile:
self.order_target_percent(data, target=1.0 / len(lowest_turnover_quartile))
# Go short on stocks with the highest turnover
for data in highest_turnover_quartile:
self.order_target_percent(data, target=-1.0 / len(highest_turnover_quartile))
# Close positions for the stocks not in the lowest or highest turnover quartile
for data in self.datas:
if data not in lowest_turnover_quartile and data not in highest_turnover_quartile:
self.close(data)
cerebro = bt.Cerebro()
cerebro.addstrategy(LowLiquidityStrategy)
# Add your data feeds to cerebro here
cerebro.run()Note that this code is a starting point for implementing the trading rules you provided. You will need to add your data feeds to cerebro and ensure that the data feeds have the necessary information, such as market capitalization, turnover, and any exclusion criteria (REITs, warrants, ADRs, ETFs, Americus Trust Components, and closed-end funds). Additionally, the code assumes the data feeds are daily. You may need to adjust it if you are using a different frequency.