Original paper

SSRN-id2676633.pdf1632.3KB

Abstract

We find a robust negative relation between skewness/lotter-like features, proxied by maximum return (MAX) over the last month, and future returns for stocks preferred by individual investors. This negative relation is nonexistent for the rest of stocks. We identify stocks preferred by individual investors through bundling 10 stock characteristics associated with their stock preferences. The negative relation between MAX and future return is produced by the stocks preferred by individuals that account for less than 5% of the overall market capitalization. Our results are robust to alternative definitions of MAX and lotter-like features such as total, idiosyncratic, and expected skewness.

Keywords: MAX, lottery-like features, skewness, individual investor preference index, cross-sectional return predictability

Trading rules

Target stocks from NYSE, AMEX, and NASDAQ.
Sort stocks into ten groups based on their highest daily returns over the previous month.
Take long positions in stocks with the smallest maximum daily returns.
Short stocks with the most significant maximum daily gains.
Use a value-weighting approach for portfolio composition.
Adjust portfolios every month.

Python code

Backtrader

import backtrader as bt
import backtrader.feeds as btfeeds
import datetime

class LotteryEffectStrategy(bt.Strategy):
    def __init__(self):
        self.inds = dict()
        for d in self.datas:
            self.inds[d] = dict()
            self.inds[d]['max_return'] = bt.indicators.Highest(d.close / d.close(-1) - 1, period=21)

    def next(self):
        max_returns = []
        for d in self.datas:
            max_returns.append((self.inds[d]['max_return'][0], d))

        max_returns.sort(key=lambda x: x[0])

        long_stocks = max_returns[:len(max_returns) // 10]
        short_stocks = max_returns[-(len(max_returns) // 10):]

        for _, d in long_stocks:
            self.order_target_percent(d, target=1 / len(long_stocks))

        for _, d in short_stocks:
            self.order_target_percent(d, target=-1 / len(short_stocks))

        for _, d in max_returns[len(long_stocks):-len(short_stocks)]:
            self.order_target_percent(d, target=0)

cerebro = bt.Cerebro()

data_path = 'path_to_your_data_directory'
tickers = ['NYSE', 'AMEX', 'NASDAQ']

for ticker in tickers:
    data = btfeeds.GenericCSVData(
        dataname=data_path + ticker + '.csv',
        dtformat=('%Y-%m-%d'),
        datetime=0,
        high=2,
        low=3,
        open=1,
        close=4,
        volume=5,
        openinterest=-1,
        reverse=False
    )
    cerebro.adddata(data)

cerebro.addstrategy(LotteryEffectStrategy)
cerebro.broker.set_cash(100000)
cerebro.addanalyzer(bt.analyzers.SharpeRatio, _name='sharpe_ratio')
cerebro.addanalyzer(bt.analyzers.DrawDown, _name='drawdown')
cerebro.addanalyzer(bt.analyzers.AnnualReturn, _name='annual_return')

results = cerebro.run()

print('Sharpe Ratio:', results[0].analyzers.sharpe_ratio.get_analysis())
print('Drawdown:', results[0].analyzers.drawdown.get_analysis())
print('Annual Return:', results[0].analyzers.annual_return.get_analysis())

cerebro.plot()

Please note that you will need to provide your own data in CSV format for the NYSE, AMEX, and NASDAQ stocks and adjust the data path accordingly.

Skewness, Individual investor preference, and the Cross-Section of Stock Returns

Original paper

Abstract

Trading rules

Python code

Backtrader