Backtest Period
Markets Traded
Equities
Maximum Drawdown
Period of Rebalancing
Monthly
Return (Annual)
Sharpe Ratio
Standard Deviation (Annual)
Original paper
SSRN-id3309787.pdf977.4KB
Abstract
Theoretical models on herd behavior predict that under different assumptions, herding can bring prices away (or towards) fundamentals and reduce (or enhance) market efficiency. In this article, we study the joint effect of herding and momentum at the industry level. We find that the momentum effect is magnified when there is a low level of investor herding. Herd behavior in investors helps move asset prices towards fundamentals, enhance market efficiency and reduce the momentum effect. A trading strategy taking a long position in winner industries and a short position in loser industries when the herding level is low can generate significant returns.
Keywords:Â Herding, Momentum, Industry
Trading rules
- Organize stocks into the 49 industry groups as identified by Fama and French.
- Estimate cross-sectional standard deviation (CSSD) of single stock returns compared to industry mean returns.
- Normalize CSSD by subtracting the mean of industry and dividing by the standard deviation of CSSD.
- Calculate herding level using previous 1-month returns.
- Identify the top third of industries with low herding levels.
- Calculate momentum as past 6-month returns (7 to 1-month returns, skipping one month) for each industry.
- In industries with low herding, take a long position on the upper half based on momentum and short the lower half.
- Monthly, readjust the portfolio, distributing weight equally across chosen stocks.
Python code
Backtrader
import backtrader as bt
class HerdingMomentumStrategy(bt.Strategy):
params = (
('herding_percentile', 0.3),
('momentum_percentile', 0.5),
)
def __init__(self):
self.industries = {} # Dictionary to store industry data
def prenext(self):
self.next()
def next(self):
# Calculate CSSD and store in dictionary
for industry in self.industries:
industry_returns = [data.close[0] / data.close[-1] - 1 for data in self.industries[industry]]
industry_mean = sum(industry_returns) / len(industry_returns)
cssd = (sum([(ret - industry_mean)**2 for ret in industry_returns]) / len(industry_returns))**0.5
self.industries[industry]['cssd'] = cssd
# Normalize CSSD and calculate herding level
cssd_values = [industry_data['cssd'] for industry_data in self.industries.values()]
cssd_mean = sum(cssd_values) / len(cssd_values)
cssd_std = (sum([(cssd - cssd_mean)**2 for cssd in cssd_values]) / len(cssd_values))**0.5
for industry_data in self.industries.values():
industry_data['herding'] = (industry_data['cssd'] - cssd_mean) / cssd_std
# Select low herding industries (top 30%)
sorted_industries = sorted(self.industries.items(), key=lambda x: x[1]['herding'])
low_herding_industries = sorted_industries[:int(len(sorted_industries) * self.params.herding_percentile)]
# Calculate momentum for each low herding industry
for industry, industry_data in low_herding_industries:
momentum = sum([data.close[0] / data.close[-6 * 30] - 1 for data in industry_data['stocks']])
industry_data['momentum'] = momentum
# Long top 50% and short bottom 50% of low herding industries based on momentum
sorted_low_herding = sorted(low_herding_industries, key=lambda x: x[1]['momentum'])
num_positions = len(sorted_low_herding)
for i, (industry, industry_data) in enumerate(sorted_low_herding):
position_size = 1 / num_positions if i < num_positions * self.params.momentum_percentile else -1 / num_positions
for data in industry_data['stocks']:
self.order_target_percent(data, position_size)
# Add data, cerebro, strategy, and other configurations here.