Original paper
Abstract
We analyze a novel alpha momentum strategy that invests in stocks based on three-factor alphas which we estimate using daily returns. The empirical analysis for the U.S. and for Europe shows that i) past alpha has power in predicting the cross-section of stock returns, ii) alpha momentum exhibits less dynamic factor exposures than price momentum and iii) alpha momentum dominates price momentum only in the U.S. Connecting both strategies to behavioral explanations, alpha momentum is more related to an underreaction to firm-specific news while price momentum is primarily driven by price overshooting due to momentum trading.
Keywords:Â Alpha momentum; price momentum; stock-specific return; price overshooting; slow information diffusion; reversal
Trading rules
- Targeted investments: Concentrate on primary domestic stocks from NYSE, AMEX, and NASDAQ with a price exceeding $1.
- Exclude closed-end funds, REITs, unit trusts, ADRs, and foreign stocks.
- Prioritize by size: Keep 10% largest stocks by market capitalization.
- Approach definition: Design the strategy using a top-minus-bottom decile portfolio, based on monthly ranking of past 12-month residual returns (excluding most recent month), standardized by standard deviation of residual returns in the same period.
- Monthly residual returns: Estimate residual returns for every stock on a monthly basis via a 3-year regression model.
- Regression specifics: Incorporate Fama and French's three factors as independent variables in the regression analysis for qualifying stocks.
- Portfolio management: Assign equal weight to stocks and adjust the portfolio every month.
Python code
Backtrader
import backtrader as bt
import pandas as pd
from sklearn.linear_model import LinearRegression
class ResidualMomentum(bt.Strategy):
params = (
('lookback', 12),
('exclude_recent', 1),
('rebalance_period', 21),
('num_stocks', 0.1),
)
def __init__(self):
self.counter = 0
self.ff_factors = get_fama_french_factors() # Implement function to get Fama-French factors data
def next(self):
if self.counter % self.params.rebalance_period == 0:
eligible_stocks = self.get_eligible_stocks()
residual_returns = self.calculate_residual_returns(eligible_stocks)
selected_stocks = self.select_stocks(residual_returns)
self.adjust_portfolio(selected_stocks)
self.counter += 1
def get_eligible_stocks(self):
pass # Implement function to get eligible stocks from NYSE, AMEX, and NASDAQ, filtering out excluded types and priced over $1
def calculate_residual_returns(self, eligible_stocks):
residual_returns = {}
for stock in eligible_stocks:
stock_data = self.get_stock_data(stock)
stock_returns = self.calculate_stock_returns(stock_data)
residual_return = self.calculate_stock_residual_return(stock_returns)
residual_returns[stock] = residual_return
return residual_returns
def select_stocks(self, residual_returns):
num_stocks = int(len(residual_returns) * self.params.num_stocks)
sorted_stocks = sorted(residual_returns, key=residual_returns.get, reverse=True)
return sorted_stocks[:num_stocks]
def adjust_portfolio(self, selected_stocks):
weight = 1.0 / len(selected_stocks)
for stock in selected_stocks:
self.order_target_percent(stock, target=weight)
def get_stock_data(self, stock):
pass # Implement function to get stock data needed for regression analysis and stock returns calculation
def calculate_stock_returns(self, stock_data):
stock_data['return'] = stock_data['close'].pct_change()
return stock_data
def calculate_stock_residual_return(self, stock_returns):
lookback_range = self.params.lookback - self.params.exclude_recent
X = self.ff_factors.iloc[-lookback_range:]
y = stock_returns['return'].iloc[-lookback_range:]
model = LinearRegression()
model.fit(X, y)
residuals = y - model.predict(X)
return residuals.mean() / residuals.std()This code outlines the main structure of the Residual Momentum strategy using Backtrader. You will need to implement additional functions such as get_fama_french_factors(), get_eligible_stocks(), and get_stock_data(), which are specific to your data sources for the Fama-French factors and stock data. Make sure to handle the data and date alignments correctly when implementing these functions.