Equity · United States · CFDs, ETFs, funds, futures · Monthly · United States Free

Text-Based Recession Detection Strategy

Academic paper

AuthorsSalim Baz; Lara Cathcart; Alexander Michaelides

Institute

?Imperial College Business School
Centre for Economic Policy Research
?Centre for Economic Policy Research (CEPR)

Paper linkhttps://papers.ssrn.com/sol3/papers.cfm?abstract_id=4251414

Strategy in a nutshell

Switch between SPY and BIL based on media-based recession signals. Hold stocks when recession probability <50% and cash when >50%, with dynamic rebalancing.

Economic rationale

Uses NLP to track “recession” mentions in financial news. Media signals predict NBER recessions six months ahead, enabling timely market-timing decisions to reduce risk.

Backtest performance

Annualised return8.78%

Volatility13.52%

Beta0.45

Sharpe ratio0.65

Sortino ratio0.338

Maximum drawdown-18.15%

Win rate61%

Full Python code

1from AlgorithmImports import *
2from typing import List
3from data_tools import FREDData, Data, IndexMRI, IndexNBER, MultipleLinearRegression
4# endregion
5
6class TextBasedRecessionDetectionStrategy(QCAlgorithm):
7
8def Initialize(self):
9self.SetStartDate(2000, 1, 1)
10self.SetCash(100_000)
11
12self.leverage: int = 5
13self.recent_month: int = -1
14self.regression_period: int = 6
15self.max_missing_days: int = 35
16self.min_values: int = 15
17self.threshold: float = 0.5
18
19self.data: Data = Data(self.regression_period)
20
21security: Equity = self.AddEquity('SPY', Resolution.Daily)
22security.SetLeverage(self.leverage)
23
24self.spy: Symbol = security.Symbol
25
26security: Equity = self.AddEquity('BIL', Resolution.Daily)
27security.SetLeverage(self.leverage)
28
29self.bil: Symbol = security.Symbol
30
31self.baa10ym: Symbol = self.AddData(FREDData, 'BAA10YM', Resolution.Daily).Symbol
32self.t10y3m: Symbol = self.AddData(FREDData, 'T10Y3M', Resolution.Daily).Symbol
33
34self.mri: Symbol = self.AddData(IndexMRI, 'MRI', Resolution.Daily).Symbol
35self.nber: Symbol = self.AddData(IndexNBER, 'NBER', Resolution.Daily).Symbol
36
37def OnData(self, slice: Slice) -> None:
38if self.baa10ym in slice and slice[self.baa10ym]:
39    # monthly data
40    curr_date:datetime.date = self.Time.date()
41
42    # make sure data still coming
43    if not self.data.baa10ym_data_still_coming(curr_date, self.max_missing_days):
44        self.data.reset_baa10ym()
45
46    self.data.update_baa10ym(curr_date, slice[self.baa10ym].Value)
47
48if self.t10y3m in slice and slice[self.t10y3m]:
49    # daily data
50    self.data.update_t10y3m(slice[self.t10y3m].Value)
51
52if self.mri in slice and slice[self.mri]:
53    # daily data
54    self.data.update_mri(slice[self.mri].Value)
55
56if self.nber in slice and slice[self.nber]:
57    # monthly data
58    curr_date:datetime.date = self.Time.date()
59
60    # make sure data still coming
61    if not self.data.nber_data_still_coming(curr_date, self.max_missing_days):
62        self.data.reset_nber()
63
64    self.data.update_nber(self.Time.date(), slice[self.nber].Value)
65
66# rebalance monthly
67if self.Time.month == self.recent_month:
68    return
69self.recent_month = self.Time.month
70
71# if there aren't enough daily data, monthly data for regresion will be reset
72self.data.update_monthly_values(self.min_values)
73self.data.reset_daily_values()
74
75if not self.data.regression_data_ready():
76    self.Liquidate()
77else:
78    train_y: List[float] = self.data.get_train_regression_y()
79    train_x: List[List[float]] = self.data.get_train_regression_x()
80
81    regression_model = MultipleLinearRegression(train_x, train_y)
82    
83    test_x: List[List[float]] = self.data.get_test_regression_x()
84    predicted_value: float = regression_model.predict(test_x)[0]
85
86    if predicted_value > self.threshold:
87        self.Liquidate(self.spy)
88        if self.bil in slice and slice[self.bil]:
89            self.SetHoldings(self.bil, 1)
90
91    else:
92        self.Liquidate(self.bil)
93        if self.spy in slice and slice[self.spy]:
94            self.SetHoldings(self.spy, 1)