Original paper

SSRN-id2620085.pdf446.4KB

Abstract

Given that extreme industry returns may herald long-term structural changes in the industries involved that may eventually lead to reversals in industry fortunes, we investigate the evidence for long-term return reversal in industry returns. Our study employs both pure contrarian strategies and late-stage contrarian strategies, and includes extra-long strategy formation periods (up to 132 months) to allow sufficient time for structural changes to begin. We find strong evidence of reversals in the long-term returns of industries. These reversals continue for many years (with valuation effects observed up to ten years after commencement) and are difficult to reconcile with overreaction.

Keywords: late-stage contrarian, reversal, mergers, acquisitions

Trading rules

Investment scope: ETFs mirroring key stock sectors (48 primary sectors examined in the research)
Periodicity: Every quarter with a holding duration of 3 months
Initial Step: Organize sectors into four groups based on the previous 10-year returns
Secondary Step: Sub-segment the top (outperformers) and bottom (underperformers) quartiles according to the past year's returns
Strategy: Buy long-term losers with strong recent short-term performance and sell long-term winners with weak recent short-term performance
Portfolio Composition: Equally weighted and rebalanced every 3 months.

Python code

Backtrader

import backtrader as bt
import pandas as pd
import numpy as np

class IndustryMomentumReversal(bt.Strategy):
    params = (
        ('lookback_long', 120),
        ('lookback_short', 12),
        ('holding_period', 3),
    )

    def __init__(self):
        self.month_counter = 0

    def next(self):
        self.month_counter += 1
        if self.month_counter % self.params.holding_period != 0:
            return

        performance_long = {}
        performance_short = {}

        for d in self.getdatanames():
            data = self.getdatabyname(d)
            if len(data) < self.params.lookback_long:
                continue

            performance_long[d] = (data.close[0] / data.close[-self.params.lookback_long]) - 1
            performance_short[d] = (data.close[0] / data.close[-self.params.lookback_short]) - 1

        sorted_by_long = sorted(performance_long.items(), key=lambda x: x[1])
        losers = sorted_by_long[:len(sorted_by_long) // 4]
        winners = sorted_by_long[-len(sorted_by_long) // 4:]

        losers_sorted_by_short = sorted(losers, key=lambda x: x[1], reverse=True)
        winners_sorted_by_short = sorted(winners, key=lambda x: x[1])

        selected_losers = [x[0] for x in losers_sorted_by_short[:len(losers_sorted_by_short) // 4]]
        selected_winners = [x[0] for x in winners_sorted_by_short[:len(winners_sorted_by_short) // 4]]

        positions = self.broker.get_value()

        for d in self.getdatanames():
            data = self.getdatabyname(d)
            position_size = positions / (len(selected_losers) + len(selected_winners))

            if d in selected_losers:
                self.order_target_value(data, target=position_size)
            elif d in selected_winners:
                self.order_target_value(data, target=-position_size)
            else:
                self.order_target_value(data, target=0)

cerebro = bt.Cerebro()

# Add data feeds here with appropriate names
# Example:
# data = bt.feeds.YahooFinanceData(dataname='SPY', fromdate=start_date, todate=end_date)
# cerebro.adddata(data, name='SPY')

cerebro.addstrategy(IndustryMomentumReversal)
cerebro.broker.setcash(1000000)
cerebro.broker.setcommission(commission=0.001)

print('Starting Portfolio Value:', cerebro.broker.getvalue())
cerebro.run()
print('Final Portfolio Value:', cerebro.broker.getvalue())

Industry Long-Term Return Reversal

Original paper

Abstract

Trading rules

Python code

Backtrader