Original paper
Abstract
We document a new empirical phenomenon in which the positions of money managers (MM), who are sophisticated speculators in the commodity futures market, as disclosed by the CFTC Disaggregated Commitments of Traders (DCOT) reports, can predict the cross-section of commodity producers' stock returns in the subsequent week. We employ cross-sectional methodologies including single-sort, Jensen's alpha analysis, double-sort, and Fama-Macbeth regressions to confirm the predictability results. The results are more pronounced in firms with higher information asymmetry, proxied by analyst dispersion and historical volatility. We thus provide more empirical evidence to the literature on costly information processing which leads to market segmentation and gradual information diffusion across asset markets, as demonstrated in the lead-lag relationship.
Keywords:Â Costly Information Processing, Commodity, Equity Return Predictability, Market Segmentation, Gradual Information Diffusion
Trading rules
- Investment universe: Ordinary common stocks (CRSP SHRCD 10 or 11) traded on NYSE, AMEX, NASDAQ
- Match commodity-producing firms with corresponding commodities using 4-digit U.S. SIC code
- Examine 11 commodities, which include 2 types of industrial metals, 3 kinds of precious metals, 4 energy-related commodities, and 2 varieties of soft commodities.
- For trader positions and classifications, refer to the CFTC's Commitment of Traders (COT) reports.
- Calculate position indicator: Long proportion growth (growth in (MMlong) / (MMshort + MMlong + 2MMspreading))
- Identify firms with corresponding industry codes for each commodity
- Create 11 commodity portfolios
- Use DCOT reports tabulated weekly, starting from Wednesday to Tuesday’s close
- Construct Long-Short portfolio by sorting commodity-stock baskets based on lagged signal variables
- Adopt a long position on stocks showcasing positive signal growth and short those with declining signal growth.
- Strategy presented uses no-weight; other possibilities include Value-Weight, Degree-Weight, or a combination of both
Python code
Backtrader
import backtrader as bt
import datetime
import pandas as pd
class COTStrategy(bt.Strategy):
def __init__(self):
self.position_indicator = None
def next(self):
if self.position_indicator is None:
return
for i, d in enumerate(self.datas):
sig_growth = self.position_indicator[i]
if sig_growth > 0:
self.order_target_percent(d, target=1.0 / len(self.datas))
else:
self.order_target_percent(d, target=-1.0 / len(self.datas))
def notify_order(self, order):
if order.status in [order.Submitted, order.Accepted]:
return
if order.status in [order.Completed]:
if order.isbuy():
self.log(f"BUY EXECUTED, {order.executed.price}")
elif order.issell():
self.log(f"SELL EXECUTED, {order.executed.price}")
def log(self, txt, dt=None):
dt = dt or self.datas[0].datetime.date(0)
print(f"{dt.isoformat()}, {txt}")
@staticmethod
def get_data_feed(ticker, fromdate, todate):
return bt.feeds.GenericCSVData(
dataname=f"{ticker}.csv",
fromdate=fromdate,
todate=todate,
dtformat=("%Y-%m-%d"),
datetime=0,
high=2,
low=3,
open=1,
close=4,
volume=5,
openinterest=-1,
)
@staticmethod
def get_cot_data():
# Dummy function, replace this with actual COT data loading and processing
return pd.DataFrame()
@staticmethod
def calculate_position_indicator(cot_data):
# Dummy function, replace this with actual position indicator calculation
return [0] * len(cot_data)
@staticmethod
def get_sic_commodity_map():
# Dummy function, replace this with actual SIC code to commodity map
return {}
if __name__ == "__main__":
# Load the COT data and calculate the position indicator
cot_data = COTStrategy.get_cot_data()
position_indicator = COTStrategy.calculate_position_indicator(cot_data)
# Map SIC codes to commodities
sic_commodity_map = COTStrategy.get_sic_commodity_map()
# Initialize Backtrader
cerebro = bt.Cerebro()
# Create data feeds for each stock in the universe
for ticker, sic_code in sic_commodity_map.items():
data = COTStrategy.get_data_feed(ticker, datetime.datetime(2006, 1, 1), datetime.datetime(2017, 12, 31))
cerebro.adddata(data)
# Add the strategy to Backtrader
cerebro.addstrategy(COTStrategy)
# Run the strategy
cerebro.run()Please note that this code is not complete and only serves as a starting point for implementing the trading strategy in Backtrader. You will need to replace the dummy functions with actual implementations for loading COT data, calculating the position indicator, and mapping SIC codes to commodities. You should also provide your own stock data in CSV format for the data feeds.