Original paper
Abstract
We present evidence consistent with markets failing to properly price information in seasonal earnings patterns. Firms with historically larger earnings in one quarter of the year (āpositive seasonality quartersā) have higher returns when those earnings are usually announced. Analysts have more positive forecast errors in positive seasonality quarters, consistent with the returns being driven by mistaken earnings estimates. We show that investors appear to overweight recent lower earnings following positive seasonality quarters, leading to pessimistic forecasts in the subsequent positive seasonality quarter. The returns are not explained by risk-based explanations, firm-specific information, increased volume, or idiosyncratic volatility.
Keywords:Ā Earnings, Seasonality, Mispricing, Analysts, Anomalies
Trading rules
- Stocks in the perimeter: Stocks listed on NASDAQ, NYSE, or AMEX
- Exclude stocks: Valued at less than $5, lacking market capitalization, or without earnings information from any of the last 20 quarters
- Method to determine seasonality (earnrank): Use five consecutive years of earnings data (from quarter t-23 through t-4)
- Evaluate company earnings per share (accounting for stock split adjustments)
- Rank 20 quarters of earnings data from the largest to the smallest
- For quarter t, derive earnrank by taking the average rank from quarters t-4, t-8, t-12, t-16, and t-20
- Adopt a bullish position: On the top 5% companies displaying the highest earnrank
- Adopt a bearish position: On the bottom 5% companies displaying the lowest earnrank
- Seasonality condition: Each firm appears in each portfolio only once per year
- Composition of portfolio: Ensure a minimum of 10 companies for both bullish and bearish positions
- Trading timeline: Initiate positions on day t-1 and maintain them up to day t+1 (where day t is the announcement day)
- Portfolio management: Distribute weight equally and adjust the balance daily.
Python code
Backtrader
import backtrader as bt
class EarningsSeasonalityStrategy(bt.Strategy):
def __init__(self):
self.earnings_data = self.get_earnings_data()
def get_earnings_data(self):
# Load and preprocess earnings data here
pass
def compute_earnrank(self, data):
# Compute earnrank for given stock data
pass
def next(self):
long_stocks = []
short_stocks = []
for data in self.datas:
if data.close[0] < 5 or data.missing_data:
continue
earnrank = self.compute_earnrank(data)
if earnrank is None:
continue
if earnrank >= 0.95:
long_stocks.append(data)
elif earnrank <= 0.05:
short_stocks.append(data)
if len(long_stocks) >= 10 and len(short_stocks) >= 10:
long_weight = 1.0 / len(long_stocks)
short_weight = -1.0 / len(short_stocks)
for data in long_stocks:
self.order_target_percent(data, target=long_weight)
for data in short_stocks:
self.order_target_percent(data, target=short_weight)
else:
self.close_positions()
def close_positions(self):
for data in self.datas:
self.order_target_percent(data, target=0)
if __name__ == '__main__':
cerebro = bt.Cerebro()
# Add data feeds to cerebro here
cerebro.addstrategy(EarningsSeasonalityStrategy)
cerebro.run()Please note that this code is a starting point and some parts of the code like get_earnings_data and compute_earnrank should be implemented according to your specific data sources and requirements. You would also need to add your data feeds to the cerebro instance.