Oil-Driven Equity Allocation via Monthly Regression
Log in to collectStrategy in a nutshell
Several types of oil can be used (Brent, WTI, Dubai, etc.) without big differences in results. The
Economic rationale
Equity predictability is explained by the underreaction hypothesis. It seems that it takes time before information about oil price changes become fully reflected in stock market prices. Underreaction can occur due to a possible difficulty for investors to assess the impact of information on the value of stocks, or when investors react to information at different points in time.
III. SOURCE PAPER
Striking Oil: Another Puzzle? [Click to Open PDF]
Gerben Driesprong, Erasmus University Rotterdam - Rotterdam School of Management ; Ben Jacobsen, Tilburg University - TIAS School for Business and Society; Massey University ; Benjamin Maat, APG Asset Management
Changes in oil prices predict stock market returns worldwide. In our thirty year sample of monthly returns for developed stock markets, we find statistically significant predictability for twelve out of eighteen countries as well as for the world market index. Results are similar for our shorter time series of emerging markets. We find no evidence that our results can be explained by time varying risk premia. Even though oil price shocks increase risk, investors seem to underreact to information in the price of oil: a rise in oil prices does not lead to higher stock market returns, but drastically lowers returns. For instance, an oil price shock of one standard deviation (around 10 percent) predictably lowers world market returns by one percent. Oil price changes also significantly predict negative excess returns. Our findings are consistent with the hypothesis of a delayed reaction by investors to oil price changes. In line with this hypothesis the relation between monthly stock returns and lagged monthly oil price changes becomes substantially stronger once we introduce lags of several trading days between monthly stock returns and lagged monthly oil price changes.
Backtest performance
Full Python code
from data_tools import QuantpediaFutures, QuandlValue, CustomFeeModel, InterestRate3M
from AlgoLib import *
from typing import List, Dict
import numpy as np
from scipy import stats
from collections import deque
class CrudeOilPredictsEquityReturns(XXX):
def Initialize(self):
self.SetStartDate(2000, 1, 1)
self.SetCash(100000)
self.min_period:int = 13
self.leverage:int = 2
self.data:Dict[Symbol, deque] = {}
self.symbols:List[str] = [
"CME_ES1", # E-mini S&P 500 Futures, Continuous Contract #1
"CME_CL1" # Crude Oil Futures, Continuous Contract #1
]
self.cash:Symbol = self.AddEquity('SHY', Resolution.Daily).Symbol
self.risk_free_rate:Symbol = self.AddData(InterestRate3M, 'IR3TIB01USM156N', Resolution.Daily).Symbol
for symbol in self.symbols:
data = self.AddData(QuantpediaFutures, symbol, Resolution.Daily)
data.SetLeverage(self.leverage)
data.SetFeeModel(CustomFeeModel())
self.data[symbol] = deque()
self.Settings.MinimumOrderMarginPortfolioPercentage = 0.
self.recent_month:int = -1
def OnData(self, data:Slice) -> None:
rebalance_flag:bool = False
for symbol in self.symbols:
if symbol in data:
if self.recent_month != self.Time.month:
rebalance_flag = True
if data[symbol]:
price:float = data[symbol].Value
self.data[symbol].append(price)
if rebalance_flag:
self.recent_month = self.Time.month
ir_last_update_date:Dict[str, datetime.date] = InterestRate3M.get_last_update_date()
last_update_date:Dict[str, datetime.date] = QuantpediaFutures.get_last_update_date()
rf_rate:float = .0
# check if data is still coming
if self.Securities[self.risk_free_rate].GetLastData() and ir_last_update_date[self.risk_free_rate.Value] > self.Time.date():
rf_rate = self.Securities[self.risk_free_rate].Price / 100
else:
return
if not all(last_update_date[x] > self.Time.date() for x in self.symbols):
self.Liquidate()
return
market_prices:np.ndarray = np.array(self.data[self.symbols[0]])
oil_prices:np.ndarray = np.array(self.data[self.symbols[1]])
# At least one year of data is ready.
if len(market_prices) < self.min_period or len(oil_prices) < self.min_period:
return
# Trim price series lenghts.
min_size:float = min(len(market_prices), len(oil_prices))
market_prices = market_prices[-min_size:]
oil_prices = oil_prices[-min_size:]
market_returns = market_prices[1:] / market_prices[:-1] - 1
oil_returns = oil_prices[1:] / oil_prices[:-1] - 1
# Simple Linear Regression
# Y = C + (M * X)
# Y = α + (β ∗ X)
# Y = Dependent variable (output/outcome/prediction/estimation)
# C/α = Constant (Y-Intercept)
# M/β = Slope of the regression line (the effect that X has on Y)
# X = Independent variable (input variable used in the prediction of Y)
slope, intercept, r_value, p_value, std_err = stats.linregress(oil_returns[:-1], market_returns[1:])
expected_market_return = intercept + (slope * oil_returns[-1])
if expected_market_return > rf_rate:
if self.Portfolio[self.cash].Invested:
self.Liquidate(self.cash)
self.SetHoldings(self.symbols[0], 1)
else:
if self.Portfolio[self.symbols[0]].Invested:
self.Liquidate(self.symbols[0])
if self.cash in data and data[self.cash]:
self.SetHoldings(self.cash, 1)