Original paper
Abstract
Using a unique proprietary data set of positions held by all large traders in the crude oil, gasoline, and heating oil futures markets, we use actual trader profits to test the predictions of various commodity futures pricing models. We find statistically and economically significant evidence that: (a) mean hedger profits are negative while speculator profits are positive, which is consistent with the risk premium hypothesis, (b) traders (whether speculators or hedgers) who hold long (short) positions when likely hedgers in aggregate are net short (long) have higher profits than traders whose net positions are aligned with likely hedgers, which is consistent with the hedging pressure hypothesis, and (c) profits on long positions vary inversely with inventories and directly with price volatility, which is consistent with the modern theory of storage. We establish these associations while controlling for macroeconomic risk factors that potentially affect futures returns and for trader characteristics. Our results indicate also that the momentum in commodity futures markets may be due largely to hedging pressure.
Trading rules
- Targeted assets: Focus on 27 futures from commodities (comprising 12 from agriculture, 5 from energy, 4 livestock, 5 metals, plus lumber).
- Hold the contract closest to maturity up to a month before its expiry; thereafter, shift to the subsequent closest one to avoid physical delivery.
- Exclude 25% of commodities with the lowest average open interest from the past weeks for the nearest trading week.
- Hedge pressure: Calculate by dividing the long positions by the overall positions of commercial traders (sourced from the Commitment of Traders report).
- Portfolio strategy:
- Adopt a long position on the bottom 20% of contracts reflecting the lowest hedge intensity (indicating predominant short positions by hedgers) from the last N weeks.
- Adopt a short position on the top 20% of contracts reflecting the highest hedge intensity (indicating predominant long positions by hedgers) from the past N weeks.
- Keep these positions for the upcoming H weeks; new long-short portfolios formed after that.
- Ranking duration (N) choices: 4, 13, 26, or 52 weeks.
- Holding (H) choices: 4, 13, 26, or 52 weeks.
- Total: 16 long-short portfolios created and rebalanced.
- Investment universe: 27 commodity futures (12 agricultural, 5 energy, 4 livestock, 5 metal, and lumber)
- Hold nearest contract until one month before maturity; roll to second nearest contract to avoid physical delivery
- Exclude 25% of commodities with lowest average open interest over preceding weeks from trading in the nearest week
- Hedging pressure: long positions divided by total positions of commercial traders (using Commitment of Traders report data)
- Portfolio:
- Long positions on 20% of contracts with lowest hedging pressure (hedgers net short) over previous R weeks
- Short positions on 20% of contracts with highest hedging pressure (hedgers net long) over previous R weeks
- Positions held for next H weeks; new long-short portfolios formed after that
- Ranking period (R) options: 4, 13, 26, and 52 weeks
- Holding period (H) options: 4, 13, 26, and 52 weeks
- Total: 16 long-short portfolios created and rebalanced
Python code
Backtrader
import backtrader as bt
import numpy as np
class HedgersEffect(bt.Strategy):
params = (
('ranking_period', [4, 13, 26, 52]),
('holding_period', [4, 13, 26, 52]),
)
def __init__(self):
self.hedging_pressure = dict()
def next(self):
if self.datas[0].datetime.date(0).isocalendar()[1] % min(self.params.ranking_period) == 0:
self.calculate_hedging_pressure()
if self.datas[0].datetime.date(0).isocalendar()[1] % min(self.params.holding_period) == 0:
self.rebalance_portfolio()
def calculate_hedging_pressure(self):
for data in self.datas:
long_positions = data.positions.long
total_positions = data.positions.total
self.hedging_pressure[data._name] = long_positions / total_positions
def rebalance_portfolio(self):
sorted_hedging_pressure = sorted(self.hedging_pressure.items(), key=lambda x: x[1])
long_symbols = [symbol for symbol, _ in sorted_hedging_pressure[:int(len(sorted_hedging_pressure) * 0.2)]]
short_symbols = [symbol for symbol, _ in sorted_hedging_pressure[-int(len(sorted_hedging_pressure) * 0.2):]]
for data in self.datas:
if data._name in long_symbols:
self.order_target_percent(data, target=1 / len(long_symbols))
elif data._name in short_symbols:
self.order_target_percent(data, target=-1 / len(short_symbols))
else:
self.order_target_percent(data, target=0)
if __name__ == '__main__':
cerebro = bt.Cerebro()
# Add 27 commodity futures data feeds
# ...
cerebro.addstrategy(HedgersEffect)
cerebro.run()Please note that this code is a basic implementation of the strategy and assumes that you have already set up the data feeds for the 27 commodity futures. The actual implementation may require modifications based on the data source and format you are using. Additionally, you may need to add a custom commission scheme, slippage, and additional logic for rolling contracts based on your specific data.