Original paper
Abstract
Using combinations of weekdays and times of day (before, during, and after trading hours) of earnings announcements, we examine whether managers attempt to strategically time these announcements. We document that the worst earnings news is announced on Friday evening and find robust evidence that only Friday evening announcements represent managers’ rational opportunistic behavior. Friday evening announcements are followed by insider trading in the direction of earnings news and the largest post-earnings announcement drift. Managers also attempt to reduce interaction with investors and hide more than just earnings news by announcing on Friday evening. We find that Friday evening announcements occur later in the evening than announcements on other evenings, firms have a reduced propensity to hold conference calls, and major firm restructuring events are relatively more likely to occur after Friday evening announcements.
Keywords:Â earnings announcements, timing, Friday evening, inattention, hiding news, post-earnings announcement drift
Trading rules
- Targeted assets: US equities with a valuation exceeding $100 million, a stock price over $2, and earnings announcements on Fridays.
- Create portfolios based on earnings surprise magnitude.
- Use empirical SUE distribution from year t-1 to calculate medians of positive and negative SUE values.
- Construct portfolios each month in year t:
- Go long on companies with SUE values surpassing the positive median.
- Go short on companies with SUE values falling below the negative median.
- If a company has a positive (negative) SUE greater (lesser) than the median, incorporate it into the long (short) portfolio at the end of the month or the following month if the announcement coincides with the month's last trading day.
- Maintain stock in portfolio for 12 months, from month t+1 to t+12 (inclusive).
- Equally weight stocks and rebalance portfolio monthly.
Python code
Backtrader
import backtrader as bt
import datetime
class EarningsSurpriseStrategy(bt.Strategy):
def __init__(self):
self.positive_sue_median = None
self.negative_sue_median = None
def next(self):
date = self.data.datetime.date()
if date.month != self.prev_month:
# Calculate medians of positive and negative SUE values
positive_sue_list = [d.sue[0] for d in self.datas if d.sue[0] > 0]
negative_sue_list = [d.sue[0] for d in self.datas if d.sue[0] < 0]
self.positive_sue_median = self.calc_median(positive_sue_list)
self.negative_sue_median = self.calc_median(negative_sue_list)
# Rebalance portfolio
long_portfolio = [d for d in self.datas if d.sue[0] > self.positive_sue_median]
short_portfolio = [d for d in self.datas if d.sue[0] < self.negative_sue_median]
for data in self.datas:
position = self.getposition(data).size
if data in long_portfolio:
if position <= 0:
self.buy(data, exectype=bt.Order.Market, size=1)
elif data in short_portfolio:
if position >= 0:
self.sell(data, exectype=bt.Order.Market, size=1)
elif position != 0:
self.close(data, exectype=bt.Order.Market)
self.prev_month = date.month
@staticmethod
def calc_median(values_list):
sorted_list = sorted(values_list)
length = len(sorted_list)
if length % 2 == 0:
median = (sorted_list[length // 2 - 1] + sorted_list[length // 2]) / 2
else:
median = sorted_list[length // 2]
return median
cerebro = bt.Cerebro()
# Add stock data
for stock_data in your_stock_data_list:
if stock_data.market_value > 100e6 and stock_data.share_price > 2 and stock_data.earnings_announcement_on_friday():
cerebro.adddata(stock_data)
# Add strategy
cerebro.addstrategy(EarningsSurpriseStrategy)
# Run backtest
results = cerebro.run()Please note that this code assumes that you have stock data with the required attributes (market_value, share_price, earnings_announcement_on_friday()) and the SUE value available in your custom DataFeed. You’ll need to adjust the code to work with your specific data source.