Original paper

SSRN-id1998387.pdf663.3KB

Abstract

An accounting-based model has strong out-of-sample power not only to detect fraud, but also to predict cross-sectional returns. Firms with a higher probability of manipulation (MSCORE) earn lower returns in every decile portfolio sorted by: Size, Book-to-Market, Momentum, Accruals, and Short-Interest. We show that the predictive power of MSCORE is related to its ability to forecast the persistence of current-year accruals, and is most pronounced among low-accrual (ostensibly high earnings-quality) stocks. Most of the incremental power derives from measures of firms’ predisposition to manipulate, rather than their level of aggressive accounting. Our evidence supports the investment value of careful fundamental analysis, even among public firms.

Keywords: earnings manipulation detection, accounting fraud, returns prediction, market efficiency, financial statement analysis, market learning, information, behavioral finance, price discovery

Trading rules

Targeted assets: Enterprises valued at over $1 Billion in market capitalization.
Evaluation: yearly in April (after year-end fundamental data is reported).
For every company, derive the O-SCORE, ranging between 0 and 5, based on the following criteria:

Highest quintile of sales growth
Lowest quintile of CFO to total assets
Net stock issuance greater than industry median (current or prior year)
Acquisition of another company within the last five years
Highest quintile of PROBM (probability of manipulation)

Bet against those stocks that achieve an O-SCORE of 5.
Adjust the holdings in the portfolio on an annual basis.
PROBM calculation using Beneish (1999) model with eight variables:

Days Sales in Receivables Index (DSRI)
Gross Margin Index (GMI)
Asset Quality Index (AQI)
Sales Growth Index (SGI)
Depreciation Index (DEPI)
Sales General and Administrative Expenses Index (SGAI)
Leverage Index (LVGI)
Total Accruals to Total Assets (TATA)

Update regression coefficients in PROBM equation annually

Python code

Backtrader

import backtrader as bt
import pandas as pd
import numpy as np
from scipy.stats import mstats

class OvervaluedStocks(bt.Strategy):
    def __init__(self):
        self.order = None

    def next(self):
        if self.order:
            return
        if len(self.datas) > 0:
            self.log('Rebalance')
            self.rebalance_portfolio()
            self.order = self.sell()

    def rebalance_portfolio(self):
        stocks = self.get_overvalued_stocks()
        for stock in stocks:
            self.order_target_percent(stock, target=-1/len(stocks))

    def get_overvalued_stocks(self):
        # Get fundamental data and calculate O-SCORE for each stock
        fundamental_data = self.get_fundamental_data()
        fundamental_data['o_score'] = self.calculate_o_score(fundamental_data)
        # Short stocks with O-SCORE of 5
        overvalued_stocks = fundamental_data[fundamental_data['o_score'] == 5].index.tolist()
        return overvalued_stocks

    def get_fundamental_data(self):
        # TODO: Implement method to get fundamental data for the investment universe
        pass

    def calculate_o_score(self, df):
        # Calculate O-SCORE based on the 5 criteria
        df['sales_growth_quintile'] = pd.qcut(df['sales_growth'], 5, labels=False)
        df['cfo_to_assets_quintile'] = pd.qcut(df['cfo_to_assets'], 5, labels=False)
        df['probm_quintile'] = pd.qcut(df['probm'], 5, labels=False)
        df['o_score'] = (
            (df['sales_growth_quintile'] == 4).astype(int) +            (df['cfo_to_assets_quintile'] == 0).astype(int) +
            (df['net_stock_issuance'] > df['industry_median_issuance']).astype(int) +
            (df['has_acquisition'] == 1).astype(int) +            (df['probm_quintile'] == 4).astype(int)
        )
        return df['o_score']

    def log(self, txt):
        print(f'{self.datetime.date()}, {txt}')

if __name__ == '__main__':
    cerebro = bt.Cerebro()
    cerebro.addstrategy(OvervaluedStocks)
    # Add data feeds for investment universe
    # TODO: Implement method to get data feeds for stocks with market cap > $1 Billion
    cerebro.broker.setcash(100000)
    cerebro.broker.setcommission(commission=0.001)
    cerebro.addsizer(bt.sizers.PercentSizer, percents=95)
    print('Starting Portfolio Value: %.2f' % cerebro.broker.getvalue())
    cerebro.run()
    print('Final Portfolio Value: %.2f' % cerebro.broker.getvalue())

Please note that you need to implement the methods for getting fundamental data and data feeds for the investment universe, as well as the PROBM calculation using the Beneish (1999) model. The provided code is a template that covers the basic structure of the strategy.

Fraud Detection and Expected Returns

Original paper

Abstract

Trading rules

Python code

Backtrader