Original paper
Abstract
An examination of the Shiller cyclically adjusted pricing-earnings (CAPE) ratio reveals its forecasting power for 12-month CRSP equally weighted (EW) excess returns and value weighted (VW) excess returns. The 12-month EW excess returns following low CAPE ratios are, on average, 20.7% higher than those following high CAPE ratios for the period of 1927-2016. This dichotomy in the Shiller CAPE ratio has a more reliable predictability than the January barometer. Previous studies report that the Halloween indicator was weak or negative in the US stock market prior to the 1950s. We find that the Halloween effect is strongly present following high CAPE ratios, even for the period of 1926-1971. Our results recommend a practical investment strategy. More specifically, if the CAPE ratio in September is lower than the 36-month median of the CAPE ratio, invest in stock markets from November to October of the following year; otherwise, invest for six months from November to April and sell in May and go away.
Keywords:Â Cyclically Adjusted Price-Earnings (CAPE) Ratio, January Barometer, Halloween Effect
Trading rules
- Target stocks: Those in the CRSP value-weighted index (available through ETFs, CFDs, futures, or index funds).
- Determine CAPE status for each month:
- Compare the CAPE ratio of September with its 36-month median.
- High CAPE month: If September's CAPE ratio surpasses the median.
- Low CAPE month: If September's CAPE ratio is below or equal to the median.
- Strategy during low CAPE months:
- Stay invested in stocks from November to the subsequent October.
- Strategy during high CAPE months:
- Invest in stocks from November through April.
- Transition to cash from May to October.
- Rebalance strategy every 6 months.
Python code
Backtrader
import backtrader as bt
import pandas as pd
import numpy as np
class CAPEStrategy(bt.Strategy):
params = (
('CAPE_period', 36),
('rebalance_months', 6),
)
def __init__(self):
self.CAPE_data = None # Placeholder for CAPE data
def next(self):
current_month = self.data.datetime.date(0).month
current_year = self.data.datetime.date(0).year
if current_month == 9: # September
CAPE_data = self.get_CAPE_data(current_year)
median_CAPE = np.median(CAPE_data[-self.params.CAPE_period:])
current_CAPE = CAPE_data[-1]
high_CAPE = current_CAPE > median_CAPE
if high_CAPE:
self.high_CAPE = True
else:
self.high_CAPE = False
if current_month == 11: # November
if self.high_CAPE:
self.order_target_percent(target=1.0)
else:
self.order_target_percent(target=1.0)
self.rebalance_date = self.data.datetime.date(0) + pd.DateOffset(months=self.params.rebalance_months)
if self.high_CAPE and current_month == 4: # April
self.order_target_percent(target=0.0)
if not self.high_CAPE and self.data.datetime.date(0) == self.rebalance_date:
self.order_target_percent(target=1.0)
self.rebalance_date = self.data.datetime.date(0) + pd.DateOffset(months=self.params.rebalance_months)
def get_CAPE_data(self, current_year):
# Load or calculate CAPE data
if self.CAPE_data is None:
# Load CAPE data for the available years
self.CAPE_data = pd.read_csv('CAPE_data.csv', index_col=0, parse_dates=True)
return self.CAPE_data.loc[:str(current_year)]
if __name__ == '__main__':
cerebro = bt.Cerebro()
cerebro.addstrategy(CAPEStrategy)
data = bt.feeds.YahooFinanceData(
dataname='^GSPC',
fromdate=pd.Timestamp('1926-01-01'),
todate=pd.Timestamp('2016-12-31'),
timeframe=bt.TimeFrame.Months,
compression=1,
)
cerebro.adddata(data)
cerebro.broker.setcash(100000.0)
cerebro.run()Please note that you will need to provide your own CAPE data in a CSV file named ‘CAPE_data.csv’ with a datetime index and a single column containing the CAPE ratios. This code assumes that the CAPE data is monthly and aligned with the CRSP value-weighted stocks data.