Accruals Seasonality
Log in to collectAcademic paper
The Rise of Accruals Seasonality Spread
Siu Kai Choy; Gerald J. Lobo; Yongxian Tan
- King's College London
- ?affiliation not provided to SSRN
- University of Houston
- ?University of Houston - C.T. Bauer College of Business
- NZUniversity of Otago
Strategy in a nutshell
Universe: NYSE, AMEX, NASDAQ non-financial stocks from CRSP.
Exclusions: Share price < $5 and missing market cap (end of previous month).
Sorting Measure:
Accruals Seasonality (AS) over past 5 years.
For quarter q, take last 20 quarters (q-23 to q-4).
Rank accruals in ascending order (higher accruals → higher rank).
AS = average of ranks from quarters q-4, q-8, q-12, q-16, q-20.
Portfolio Formation:
Each month, include only stocks with expected accruals that month.
Expected accrual = 12 months after last actual accrual date (from 10-K/Q filing or earlier report).
Sort into quintiles by AS.
Long lowest quintile, short highest quintile.
Value-weighted, capped at 95th percentile of NYSE market cap.
Rebalancing: Monthly (with varying stock sets based on reporting dates).
Economic rationale
Accruals anomaly: Investors underreact to accruals information.
Post-2001 awareness: Investors learned about accruals effect but still don’t forecast it—only react when published.
Implication: Seasonal accrual predictability should be priced in efficient markets, but isn’t.
No risk-based story: Return spread is due to investor inattention / unsophisticated arbitrage, not compensation for risk.
Backtest performance
Annualised return7.28%
Volatility8.26%
Beta-0.038
Sharpe ratio0.88
Sortino ratio-0.429
Win rate49%
Full Python code
from dateutil.relativedelta import relativedelta
from AlgorithmImports import *
class AccrualsSeasonality(QCAlgorithm):
def Initialize(self):
self.SetStartDate(2010, 1, 1) # earnings dates starts in 2010
self.SetCash(100000)
self.leverage:int = 5
self.quantile:int = 5
self.period:int = 5 # need n years of stock seasonal accruals values
self.weight:Dict[Symbol, float] = {}
self.accrual_data:Dict[Symbol, AccrualsData] = {} # last accruals values keyed by stocks symbols
self.tickers:Set(str) = set()
self.earnings_data:Dict[int, Dict[int, List[str]]] = {}
self.accruals_values = {}
self.min_share_price:float = 5.
earnings_set:Set(str) = set()
earnings_data:str = self.Download('data.quantpedia.com/backtesting_data/economic/earnings_dates_eps.json')
earnings_data_json:List[Dict] = json.loads(earnings_data)
for obj in earnings_data_json:
date:datetime.date = datetime.strptime(obj['date'], "%Y-%m-%d").date()
year:int = date.year
month:int = date.month
if year not in self.earnings_data:
self.earnings_data[year] = {}
if month not in self.earnings_data[year]:
self.earnings_data[year][month] = []
for stock_data in obj['stocks']:
ticker:str = stock_data['ticker']
self.earnings_data[year][month].append(ticker)
self.tickers.add(ticker)
market:Symbol = self.AddEquity('SPY', Resolution.Daily).Symbol
self.selection_flag:bool = False
self.UniverseSettings.Resolution = Resolution.Daily
self.AddUniverse(self.FundamentalSelectionFunction)
self.Settings.MinimumOrderMarginPortfolioPercentage = 0.
self.Schedule.On(self.DateRules.MonthStart(market), self.TimeRules.BeforeMarketClose(market, 0), self.Selection)
def OnSecuritiesChanged(self, changes: SecurityChanges) -> None:
for security in changes.AddedSecurities:
security.SetFeeModel(CustomFeeModel())
security.SetLeverage(self.leverage)
def FundamentalSelectionFunction(self, fundamental: List[Fundamental]) -> None:
if not self.selection_flag:
return Universe.Unchanged
selected:List[Fundamental] = [x for x in fundamental if x.HasFundamentalData and x.Market == 'usa' and \
x.Symbol.Value in self.tickers and x.MarketCap != 0 and x.AdjustedPrice > self.min_share_price and \
not np.isnan(x.FinancialStatements.BalanceSheet.CurrentAssets.ThreeMonths) and x.FinancialStatements.BalanceSheet.CurrentAssets.ThreeMonths != 0 and \
not np.isnan(x.FinancialStatements.BalanceSheet.CashAndCashEquivalents.ThreeMonths) and x.FinancialStatements.BalanceSheet.CashAndCashEquivalents.ThreeMonths != 0 and \
not np.isnan(x.FinancialStatements.BalanceSheet.CurrentLiabilities.ThreeMonths) and x.FinancialStatements.BalanceSheet.CurrentLiabilities.ThreeMonths != 0 and \
not np.isnan(x.FinancialStatements.BalanceSheet.CurrentDebt.ThreeMonths) and x.FinancialStatements.BalanceSheet.CurrentDebt.ThreeMonths != 0 and \
not np.isnan(x.FinancialStatements.BalanceSheet.IncomeTaxPayable.ThreeMonths) and x.FinancialStatements.BalanceSheet.IncomeTaxPayable.ThreeMonths != 0 and \
not np.isnan(x.FinancialStatements.IncomeStatement.DepreciationAndAmortization.ThreeMonths) and x.FinancialStatements.IncomeStatement.DepreciationAndAmortization.ThreeMonths != 0 and \
not np.isnan(x.FinancialStatements.BalanceSheet.TotalAssets.ThreeMonths) and x.FinancialStatements.BalanceSheet.TotalAssets.ThreeMonths != 0
]
curr_date:datetime.date = self.Time.date()
prev_month:datetime.date = curr_date - relativedelta(months=1)
curr_year, curr_month = curr_date.year, curr_date.month
prev_months_year, prev_months_month = prev_month.year, prev_month.month
seasonal_accruals_values:Dict[FineFundamental, dict[int, float]] = {}
current_accruals_data = {}
for stock in selected:
symbol:Symbol = stock.Symbol
ticker:str = symbol.Value
# accruals calculation
current_accruals_data[symbol] = AccrualsData(stock.FinancialStatements.BalanceSheet.CurrentAssets.ThreeMonths, stock.FinancialStatements.BalanceSheet.CashAndCashEquivalents.ThreeMonths,
stock.FinancialStatements.BalanceSheet.CurrentLiabilities.ThreeMonths, stock.FinancialStatements.BalanceSheet.CurrentDebt.ThreeMonths,
stock.FinancialStatements.BalanceSheet.IncomeTaxPayable.ThreeMonths, stock.FinancialStatements.IncomeStatement.DepreciationAndAmortization.ThreeMonths,
stock.FinancialStatements.BalanceSheet.TotalAssets.ThreeMonths)
# check if stock had earnings in previous month and make sure stock has accruals data from previous selection
if symbol in self.accrual_data and prev_months_year in self.earnings_data and prev_months_month in self.earnings_data[prev_months_year] \
and ticker in self.earnings_data[prev_months_year][prev_months_month]:
accrual_value:float = self.CalculateAccruals(current_accruals_data[symbol], self.accrual_data[symbol])
if symbol not in self.accruals_values:
self.accruals_values[symbol] = {}
if prev_months_year not in self.accruals_values[symbol]:
self.accruals_values[symbol][prev_months_year] = {}
# store stock's accrual keyed by previous date month, year and stock's symbol
self.accruals_values[symbol][prev_months_year][prev_months_month] = accrual_value
# check if stock will have earnings in current month
if symbol in self.accruals_values and curr_year in self.earnings_data \
and curr_month in self.earnings_data[curr_year] and ticker in self.earnings_data[curr_year][curr_month]:
stock_seasonal_accruals:Dict[int, float] = self.GetSeasonalAccruals(symbol, curr_year, curr_month)
# make sure stock has seasonal accruals for 5 years
if len(stock_seasonal_accruals) == self.period:
# store stock's seasonal accruals keyed by stock's object
seasonal_accruals_values[stock] = stock_seasonal_accruals
# set new accruals
self.accrual_data = current_accruals_data
# make sure there are enough data for selection
if len(seasonal_accruals_values) < self.quantile:
return Universe.Unchanged
# perform selection
quantile:int = int(len(seasonal_accruals_values) / self.quantile)
mean_seasonal_ranks:Dict[Symbol, float] = self.CalcSeasonalMeanRank(seasonal_accruals_values)
sorted_by_mean_rank:List[Symbol] = [x[0] for x in sorted(mean_seasonal_ranks.items(), key=lambda item: item[1])]
# long lowest
long_leg:List[Symbol] = sorted_by_mean_rank[:quantile]
# short highest
short_leg:List[Symbol] = sorted_by_mean_rank[-quantile:]
for i, portfolio in enumerate([long_leg, short_leg]):
mc_sum:float = sum([x.MarketCap for x in portfolio])
for stock in portfolio:
self.weight[symbol] = ((-1) ** i) * stock.MarketCap / mc_sum
return list(self.weight.keys())
def OnData(self, data: Slice) -> None:
# rebalance monthly
if not self.selection_flag:
return
self.selection_flag = False
# trade execution
portfolio:List[PortfolioTarget] = [PortfolioTarget(symbol, w) for symbol, w in self.weight.items() if symbol in data and data[symbol]]
self.SetHoldings(portfolio, True)
self.weight.clear()
def CalculateAccruals(self, current_accural_data, prev_accural_data) -> float:
delta_assets:float = current_accural_data.CurrentAssets - prev_accural_data.CurrentAssets
delta_cash:float = current_accural_data.CashAndCashEquivalents - prev_accural_data.CashAndCashEquivalents
delta_liabilities:float = current_accural_data.CurrentLiabilities - prev_accural_data.CurrentLiabilities
delta_debt:float = current_accural_data.CurrentDebt - prev_accural_data.CurrentDebt
delta_tax:float = current_accural_data.IncomeTaxPayable - prev_accural_data.IncomeTaxPayable
dep:float = current_accural_data.DepreciationAndAmortization
avg_total:float = (current_accural_data.TotalAssets + prev_accural_data.TotalAssets) / 2
bs_acc:float = ((delta_assets - delta_cash) - (delta_liabilities - delta_debt-delta_tax) - dep) / avg_total
return bs_acc
def GetSeasonalAccruals(self, symbol:Symbol, year:int, month:int) -> Dict:
stock_accruals = self.accruals_values[symbol]
seasonal_accruals_values:Dict[int, float] = {}
for index in range(self.period):
look_up_year:int = year - index - 1
# make sure stock has accrual value in looking year and month
if look_up_year in stock_accruals and month in stock_accruals[look_up_year]:
stock_accrual_value = stock_accruals[look_up_year][month]
# store stock's accrual value under looking year
seasonal_accruals_values[look_up_year] = stock_accrual_value
else:
# stock doesn't have all seasonal accruals value, so return blank dictionary
return {}
return seasonal_accruals_values
def CalcSeasonalMeanRank(self, seasonal_accruals_values) -> Dict:
acc_values:Dict[int, List[List[FineFundamental, float]]] = {}
# firstly create data structure, which lists with tuples (stock, stock acc) are keyed by year
for stock in seasonal_accruals_values:
stock_seasonal_acc = seasonal_accruals_values[stock]
for year in stock_seasonal_acc:
if year not in acc_values:
# initialize new list for year
acc_values[year] = []
stock_acc_value = stock_seasonal_acc[year]
# store tuple (stock, stock's acc value) keyed by year
acc_values[year].append((stock, stock_acc_value))
seasonal_ranks = {}
# get stocks ranks for each year
for year in acc_values:
# sort stocks by accruals values in curr year
sorted_by_acc = sorted(acc_values[year], key=lambda item: item[1])
for index in range(len(sorted_by_acc)):
stock_object = sorted_by_acc[index][0]
if stock_object not in seasonal_ranks:
seasonal_ranks[stock_object] = []
# append stock's rank in list keyed by stock's object
seasonal_ranks[stock_object].append(index)
mean_seasonal_ranks = dict(map(lambda kv: (kv[0], np.mean(kv[1])), seasonal_ranks.items()))
return mean_seasonal_ranks
def Selection(self):
self.selection_flag = True
class AccrualsData():
def __init__(self, current_assets, cash_and_cash_equivalents, current_liabilities, current_debt, income_tax_payable, depreciation_and_amortization, total_assets):
self.CurrentAssets = current_assets
self.CashAndCashEquivalents = cash_and_cash_equivalents
self.CurrentLiabilities = current_liabilities
self.CurrentDebt = current_debt
self.IncomeTaxPayable = income_tax_payable
self.DepreciationAndAmortization = depreciation_and_amortization
self.TotalAssets = total_assets
# Custom fee model
class CustomFeeModel(FeeModel):
def GetOrderFee(self, parameters):
fee = parameters.Security.Price * parameters.Order.AbsoluteQuantity * 0.00005
return OrderFee(CashAmount(fee, "USD"))