Performance Of Various Asset Classes During Fed Policy Cycles

Table of Contents

Introduction #

In this post, we will look into the Fed Funds cycles and evaluate asset class performance during tightening and easing of monetary policy.

Python Imports #

# Standard Library
import datetime
import os
import sys
import warnings

from datetime import datetime
from pathlib import Path

# Data Handling
import numpy as np
import pandas as pd

# Data Sources
import pandas_datareader.data as web

# Statistical Analysis
import statsmodels.api as sm

# Machine Learning
from sklearn.decomposition import PCA
from sklearn.preprocessing import StandardScaler

# Suppress warnings
warnings.filterwarnings("ignore")

Add Directories To Path #

# Add the source subdirectory to the system path to allow import config from settings.py
current_directory = Path(os.getcwd())
website_base_directory = current_directory.parent.parent.parent
src_directory = website_base_directory / "src"
sys.path.append(str(src_directory)) if str(src_directory) not in sys.path else None

# Import settings.py
from settings import config

# Add configured directories from config to path
SOURCE_DIR = config("SOURCE_DIR")
sys.path.append(str(Path(SOURCE_DIR))) if str(Path(SOURCE_DIR)) not in sys.path else None

# Add other configured directories
BASE_DIR = config("BASE_DIR")
CONTENT_DIR = config("CONTENT_DIR")
POSTS_DIR = config("POSTS_DIR")
PAGES_DIR = config("PAGES_DIR")
PUBLIC_DIR = config("PUBLIC_DIR")
SOURCE_DIR = config("SOURCE_DIR")
DATA_DIR = config("DATA_DIR")
DATA_MANUAL_DIR = config("DATA_MANUAL_DIR")

Python Functions #

Here are the functions needed for this project:

bb_clean_data: Takes an Excel export from Bloomberg, removes the miscellaneous headings/rows, and returns a DataFrame.
calc_fed_cycle_asset_performance: Calculates metrics for an asset based on a specified Fed tightening/loosening cycle.
load_data: Load data from a CSV, Excel, or Pickle file into a pandas DataFrame.
pandas_set_decimal_places: Set the number of decimal places displayed for floating-point numbers in pandas.
plot_bar_returns_ffr_change: Plot the bar chart of the cumulative or annualized returns for the asset class along with the change in the Fed Funds Rate.
plot_scatter_regression_ffr_vs_returns: Plot the scatter plot and regression of the annualized return for the asset class along with the annualized change in the Fed Funds Rate.
plot_timeseries: Plot the timeseries data from a DataFrame for a specified date range and columns.
summary_stats: Generate summary statistics for a series of returns.

from bb_clean_data import bb_clean_data
from calc_fed_cycle_asset_performance import calc_fed_cycle_asset_performance
from load_data import load_data
from pandas_set_decimal_places import pandas_set_decimal_places
from plot_bar_returns_ffr_change import plot_bar_returns_ffr_change
from plot_scatter_regression_ffr_vs_returns import plot_scatter_regression_ffr_vs_returns
from plot_timeseries import plot_timeseries
from summary_stats import summary_stats

Data Overview #

Acquire & Plot Fed Funds Data #

First, let’s get the data for the Fed Funds target rate (FFR). This data is found in 3 different datasets from FRED.

# Set decimal places
pandas_set_decimal_places(5)

# Pull Federal Funds Target Rate (DISCONTINUED) (DFEDTAR)
fedfunds_target_old = web.DataReader("DFEDTAR", "fred", start="1900-01-01", end=datetime.today())

# Pull Federal Funds Target Range - Upper Limit (DFEDTARU)
fedfunds_target_new_upper = web.DataReader("DFEDTARU", "fred", start="1900-01-01", end=datetime.today())

# Pull Federal Funds Target Range - Lower Limit (DFEDTARL)
fedfunds_target_new_lower = web.DataReader("DFEDTARL", "fred", start="1900-01-01", end=datetime.today())

# Merge the datasets together
fedfunds_combined = pd.concat([fedfunds_target_old, fedfunds_target_new_upper, fedfunds_target_new_lower], axis=1)

# Divide all values by 100 to get decimal rates
for col in fedfunds_combined.columns:
    fedfunds_combined[col] = fedfunds_combined[col] / 100

# Resample to month-end (as if we know the rate at the end of the month)
fedfunds_monthly = fedfunds_combined.resample("ME").last()

display(fedfunds_monthly)

	DFEDTAR	DFEDTARU	DFEDTARL
DATE
1982-09-30	0.10250	NaN	NaN
1982-10-31	0.09500	NaN	NaN
1982-11-30	0.09000	NaN	NaN
1982-12-31	0.08500	NaN	NaN
1983-01-31	0.08500	NaN	NaN
...	...	...	...
2025-10-31	NaN	0.04000	0.03750
2025-11-30	NaN	0.04000	0.03750
2025-12-31	NaN	0.03750	0.03500
2026-01-31	NaN	0.03750	0.03500
2026-02-28	NaN	0.03750	0.03500

522 rows × 3 columns

We can then generate several useful plots. First, the Fed Funds target rate:

plot_timeseries(
    price_df=fedfunds_monthly,
    plot_start_date=None,
    plot_end_date=None,
    plot_columns=["DFEDTAR", "DFEDTARU", "DFEDTARL"],
    title="Fed Funds Target Rate",
    x_label="Date",
    x_format="Year",
    x_tick_spacing=2,
    x_tick_rotation=45,
    y_label="Rate (%)",
    y_format="Percentage",
    y_format_decimal_places=0,
    y_tick_spacing=0.01,
    grid=True,
    legend=True,
    export_plot=False,
    plot_file_name=None,
)

png

# First drop the lower column and merge
fedfunds_monthly["fed_funds"] = fedfunds_monthly["DFEDTAR"].combine_first(fedfunds_monthly["DFEDTARU"])
fedfunds_monthly = fedfunds_monthly.drop(columns=["DFEDTARL", "DFEDTARU", "DFEDTAR"])

# Compute change in rate from either column
fedfunds_monthly["fed_funds_change"] = fedfunds_monthly["fed_funds"].diff()

display(fedfunds_monthly)

	fed_funds	fed_funds_change
DATE
1982-09-30	0.10250	NaN
1982-10-31	0.09500	-0.00750
1982-11-30	0.09000	-0.00500
1982-12-31	0.08500	-0.00500
1983-01-31	0.08500	0.00000
...	...	...
2025-10-31	0.04000	-0.00250
2025-11-30	0.04000	0.00000
2025-12-31	0.03750	-0.00250
2026-01-31	0.03750	0.00000
2026-02-28	0.03750	0.00000

522 rows × 2 columns

And then the change in FFR from month-to-month:

plot_timeseries(
    price_df=fedfunds_monthly,
    plot_start_date=None,
    plot_end_date=None,
    plot_columns=["fed_funds_change"],
    title="Fed Funds Change In Rate",
    x_label="Date",
    x_format="Year",
    x_tick_spacing=2,
    x_tick_rotation=45,
    y_label="Rate Change (%)",
    y_format="Percentage",
    y_format_decimal_places=2,
    y_tick_spacing=0.0025,
    grid=True,
    legend=False,
    export_plot=False,
    plot_file_name=None,
)

png

This plot, in particular, makes it easy to show the monthly increase and decrease in the FFR, as well as the magnitude of the change (i.e. slow, drawn-out increases or decreases or abrupt large increases or decreases).

Define Fed Policy Cycles #

Next, we will define the Fed policy tightening and easing cycles:

# Set timeframe
start_date = "1989-12-15"
end_date = "2026-01-31"

# Copy DataFrame to avoid modifying original
fedfunds_cycles = fedfunds_monthly.copy()
fedfunds_cycles = fedfunds_cycles.loc[start_date:end_date]

# Reset index
fedfunds_cycles = fedfunds_cycles.reset_index()

# Drop the "fed_funds_change" column as we will use it to determine the cycle type later, but we don't need it in the final dataframe
fedfunds_cycles = fedfunds_cycles.drop(columns=["fed_funds_change"])

# Rename the date column to "Start Date", and the "fed_funds" column to "Fed Funds Start"
fedfunds_cycles = fedfunds_cycles.rename(columns={"DATE": "start_date", "fed_funds": "fed_funds_start"})

# Copy the "Start Date" column to a new column called "End Date" and shift it up by one row
fedfunds_cycles["end_date"] = fedfunds_cycles["start_date"].shift(-1)

# Copy the "Fed Funds Start" column to a new column called "Fed Funds End" and shift it up by one row
fedfunds_cycles["fed_funds_end"] = fedfunds_cycles["fed_funds_start"].shift(-1)

# Calculate the change in Fed Funds rate for each cycle
fedfunds_cycles["fed_funds_change"] = fedfunds_cycles["fed_funds_end"] - fedfunds_cycles["fed_funds_start"]

# Based on "Fed Funds Change" column, determine if the rate is increasing, decreasing, or unchanged
fedfunds_cycles["cycle"] = fedfunds_cycles["fed_funds_change"].apply(
    lambda x: "Tightening" if x > 0 else ("Easing" if x < 0 else "Neutral"))

display(fedfunds_cycles)

	start_date	fed_funds_start	end_date	fed_funds_end	fed_funds_change	cycle
0	1989-12-31	0.08250	1990-01-31	0.08250	0.00000	Neutral
1	1990-01-31	0.08250	1990-02-28	0.08250	0.00000	Neutral
2	1990-02-28	0.08250	1990-03-31	0.08250	0.00000	Neutral
3	1990-03-31	0.08250	1990-04-30	0.08250	0.00000	Neutral
4	1990-04-30	0.08250	1990-05-31	0.08250	0.00000	Neutral
...	...	...	...	...	...	...
429	2025-09-30	0.04250	2025-10-31	0.04000	-0.00250	Easing
430	2025-10-31	0.04000	2025-11-30	0.04000	0.00000	Neutral
431	2025-11-30	0.04000	2025-12-31	0.03750	-0.00250	Easing
432	2025-12-31	0.03750	2026-01-31	0.03750	0.00000	Neutral
433	2026-01-31	0.03750	NaT	NaN	NaN	Neutral

434 rows × 6 columns

Grouping the consecutive months of tightening and easing together, we can create a DataFrame that gives us the cumulative change in the FFR for each cycle, as well as the start and end dates for each cycle. This will be useful for evaluating asset class performance during these cycles.

fedfunds_grouped_cycles = fedfunds_cycles.copy()

# Create a group key that increments whenever the Cycle value changes
fedfunds_grouped_cycles['group'] = (fedfunds_grouped_cycles['cycle'] != fedfunds_grouped_cycles['cycle'].shift(1)).cumsum()

# Group by both the group key and Cycle label
cycle_ranges = (
    fedfunds_grouped_cycles.groupby(['group', 'cycle'], sort=False)
    .agg(
        start_date=('start_date', 'first'),
        end_date=('end_date', 'last'),
        fed_funds_start=('fed_funds_start', 'first'),
        fed_funds_end=('fed_funds_end', 'last')
    )
    .reset_index(drop=False)
    .drop(columns='group')
)

display(cycle_ranges)

	cycle	start_date	end_date	fed_funds_start	fed_funds_end
0	Neutral	1989-12-31	1990-06-30	0.08250	0.08250
1	Easing	1990-06-30	1990-07-31	0.08250	0.08000
2	Neutral	1990-07-31	1990-09-30	0.08000	0.08000
3	Easing	1990-09-30	1991-04-30	0.08000	0.05750
4	Neutral	1991-04-30	1991-07-31	0.05750	0.05750
...	...	...	...	...	...
116	Neutral	2024-12-31	2025-08-31	0.04500	0.04500
117	Easing	2025-08-31	2025-10-31	0.04500	0.04000
118	Neutral	2025-10-31	2025-11-30	0.04000	0.04000
119	Easing	2025-11-30	2025-12-31	0.04000	0.03750
120	Neutral	2025-12-31	2026-01-31	0.03750	0.03750

121 rows × 5 columns

Furthermore, we will make the assumption that any “Neutral” months (i.e. months where the FFR did not change) are part of the preceding cycle – to a point. For example, if we have a tightening month followed by a neutral month, we will consider the neutral month to be part of the tightening cycle. This is a simplifying assumption, but it allows us to categorize all months into either tightening or easing cycles without having to create a separate category for neutral months. From a practical standpoint, this also makes it easier to evaluate asset class performance during these cycles, as we can simply look at the performance during the tightening and easing cycles without having to worry about the neutral months. We will foward fill no more than 6 months of neutral months, however, as it would be unreasonable to assume that a month of neutral policy is part of the preceding cycle if it has been 6 months since the last change in policy.

fedfunds_grouped_cycles = fedfunds_cycles.copy()

# Forward fill cycle labels preceding "Neutral" cycles
fedfunds_grouped_cycles['cycle_filled'] = fedfunds_grouped_cycles['cycle'].replace('Neutral', pd.NA).ffill(limit=6)

# Replaced any remaining <NA> values with "Modified Tightening"
fedfunds_grouped_cycles['cycle_filled'] = fedfunds_grouped_cycles['cycle_filled'].fillna('Modified Tightening')

# Create a group key that increments whenever the Cycle value changes
fedfunds_grouped_cycles['group'] = (fedfunds_grouped_cycles['cycle_filled'] != fedfunds_grouped_cycles['cycle_filled'].shift(1)).cumsum()

# Group by both the group key and Cycle label
cycle_ranges = (
    fedfunds_grouped_cycles.groupby(['group', 'cycle_filled'], sort=False)
    .agg(
        start_date=('start_date', 'first'),
        end_date=('end_date', 'last'),
        fed_funds_start=('fed_funds_start', 'first'),
        fed_funds_end=('fed_funds_end', 'last')
    )
    .reset_index(drop=False)
    .drop(columns='group')
)

# Calc change in Fed Funds rate for each cycle
cycle_ranges["fed_funds_change"] = cycle_ranges["fed_funds_end"] - cycle_ranges["fed_funds_start"]

# Add cycle labels
cycle_labels = [f"Cycle {i+1}" for i in range(len(cycle_ranges))]

# Combine labels with cycle_ranges
cycle_ranges['cycle_label'] = cycle_labels

display(cycle_ranges)

	cycle_filled	start_date	end_date	fed_funds_start	fed_funds_end	fed_funds_change	cycle_label
0	Modified Tightening	1989-12-31	1990-06-30	0.08250	0.08250	0.00000	Cycle 1
1	Easing	1990-06-30	1993-03-31	0.08250	0.03000	-0.05250	Cycle 2
2	Modified Tightening	1993-03-31	1994-01-31	0.03000	0.03000	0.00000	Cycle 3
3	Tightening	1994-01-31	1995-06-30	0.03000	0.06000	0.03000	Cycle 4
4	Easing	1995-06-30	1996-07-31	0.06000	0.05250	-0.00750	Cycle 5
5	Modified Tightening	1996-07-31	1997-02-28	0.05250	0.05250	0.00000	Cycle 6
6	Tightening	1997-02-28	1997-09-30	0.05250	0.05500	0.00250	Cycle 7
7	Modified Tightening	1997-09-30	1998-08-31	0.05500	0.05500	0.00000	Cycle 8
8	Easing	1998-08-31	1999-05-31	0.05500	0.04750	-0.00750	Cycle 9
9	Tightening	1999-05-31	2000-11-30	0.04750	0.06500	0.01750	Cycle 10
10	Modified Tightening	2000-11-30	2000-12-31	0.06500	0.06500	0.00000	Cycle 11
11	Easing	2000-12-31	2002-06-30	0.06500	0.01750	-0.04750	Cycle 12
12	Modified Tightening	2002-06-30	2002-10-31	0.01750	0.01750	0.00000	Cycle 13
13	Easing	2002-10-31	2003-12-31	0.01750	0.01000	-0.00750	Cycle 14
14	Modified Tightening	2003-12-31	2004-05-31	0.01000	0.01000	0.00000	Cycle 15
15	Tightening	2004-05-31	2006-12-31	0.01000	0.05250	0.04250	Cycle 16
16	Modified Tightening	2006-12-31	2007-08-31	0.05250	0.05250	0.00000	Cycle 17
17	Easing	2007-08-31	2009-07-31	0.05250	0.00250	-0.05000	Cycle 18
18	Modified Tightening	2009-07-31	2015-11-30	0.00250	0.00250	0.00000	Cycle 19
19	Tightening	2015-11-30	2016-06-30	0.00250	0.00500	0.00250	Cycle 20
20	Modified Tightening	2016-06-30	2016-11-30	0.00500	0.00500	0.00000	Cycle 21
21	Tightening	2016-11-30	2019-06-30	0.00500	0.02500	0.02000	Cycle 22
22	Modified Tightening	2019-06-30	2019-07-31	0.02500	0.02500	0.00000	Cycle 23
23	Easing	2019-07-31	2020-09-30	0.02500	0.00250	-0.02250	Cycle 24
24	Modified Tightening	2020-09-30	2022-02-28	0.00250	0.00250	0.00000	Cycle 25
25	Tightening	2022-02-28	2024-01-31	0.00250	0.05500	0.05250	Cycle 26
26	Modified Tightening	2024-01-31	2024-08-31	0.05500	0.05500	0.00000	Cycle 27
27	Easing	2024-08-31	2025-06-30	0.05500	0.04500	-0.01000	Cycle 28
28	Modified Tightening	2025-06-30	2025-08-31	0.04500	0.04500	0.00000	Cycle 29
29	Easing	2025-08-31	2026-01-31	0.04500	0.03750	-0.00750	Cycle 30

Asset Class Performance By Fed Policy Cycle #

Moving on, we will now look at the performance of four (4) different asset classes during each Fed cycle. We’ll use the following Bloomberg indices:

SPXT_S&P 500 Total Return Index (Stocks)
SPBDU10T_S&P US Treasury Bond 7-10 Year Total Return Index (Bonds)
LF98TRUU_Bloomberg US Corporate High Yield Total Return Index Value Unhedged USD (High Yield Bonds)
XAU_Gold USD Spot (Gold)

Stocks #

First, we will clean and load the data for the S&P 500 Total Return Index (SPXT):

bb_clean_data(
    base_directory=DATA_DIR,
    fund_ticker_name="SPXT_S&P 500 Total Return Index",
    source="Bloomberg",
    asset_class="Indices",
    excel_export=True,
    pickle_export=True,
    output_confirmation=False,
)

spxt = load_data(
    base_directory=DATA_DIR,
    ticker="SPXT_S&P 500 Total Return Index_Clean",
    source="Bloomberg", 
    asset_class="Indices",
    timeframe="Daily",
    file_format="pickle",
)

# Filter SPXT to date range
spxt = spxt[(spxt.index >= pd.to_datetime(start_date)) & (spxt.index <= pd.to_datetime(end_date))]

# Drop everything except the "close" column
spxt = spxt[["Close"]]

# Resample to monthly frequency
spxt_monthly = spxt.resample("M").last()
spxt_monthly["Monthly_Return"] = spxt_monthly["Close"].pct_change()
spxt_monthly = spxt_monthly.dropna()

display(spxt_monthly)

	Close	Monthly_Return
Date
1990-01-31	353.94000	-0.06713
1990-02-28	358.50000	0.01288
1990-03-31	368	0.02650
1990-04-30	358.81000	-0.02497
1990-05-31	393.80000	0.09752
...	...	...
2025-09-30	14826.80000	0.03650
2025-10-31	15173.95000	0.02341
2025-11-30	15211.14000	0.00245
2025-12-31	15220.45000	0.00061
2026-01-31	15441.15000	0.01450

433 rows × 2 columns

Next, we can plot the price history before calculating the cycle performance:

plot_timeseries(
    price_df=spxt,
    plot_start_date=None,
    plot_end_date=None,
    plot_columns=["Close"],
    title="S&P 500 Total Return Index Daily Close Price",
    x_label="Date",
    x_format="Year",
    x_tick_spacing=2,
    x_tick_rotation=45,
    y_label="Price ($)",
    y_format="Decimal",
    y_format_decimal_places=0,
    y_tick_spacing=1000,
    grid=True,
    legend=False,
    export_plot=False,
    plot_file_name=None,
)

png

Next, we will calculate the performance for SPY based on the pre-defined Fed cycles:

spxt_cycles = calc_fed_cycle_asset_performance(
    start_date=cycle_ranges["start_date"],
    end_date=cycle_ranges["end_date"],
    label=cycle_ranges["cycle_label"],
    fed_funds_change=cycle_ranges["fed_funds_change"],
    monthly_returns=spxt_monthly,
)

display(spxt_cycles)

	Cycle	Start	End	Months	CumulativeReturn	CumulativeReturnPct	AverageMonthlyReturn	AverageMonthlyReturnPct	AnnualizedReturn	AnnualizedReturnPct	Volatility	FedFundsChange	FedFundsChange_bps	FFR_AnnualizedChange	FFR_AnnualizedChange_bps	Label
0	Cycle 1	1989-12-31	1990-06-30	6	0.03092	3.09164	0.00634	0.63403	0.06279	6.27887	0.19170	0.00000	0.00000	0.00000	0.00000	Cycle 1, 1989-12-31 to 1990-06-30
1	Cycle 2	1990-06-30	1993-03-31	34	0.36800	36.80041	0.00996	0.99573	0.11694	11.69426	0.13198	-0.05250	-525.00000	-0.01853	-185.29412	Cycle 2, 1990-06-30 to 1993-03-31
2	Cycle 3	1993-03-31	1994-01-31	11	0.11359	11.35920	0.01001	1.00089	0.12454	12.45375	0.06918	0.00000	0.00000	0.00000	0.00000	Cycle 3, 1993-03-31 to 1994-01-31
3	Cycle 4	1994-01-31	1995-06-30	18	0.21800	21.80042	0.01141	1.14060	0.14051	14.05103	0.09928	0.03000	300.00000	0.02000	200.00000	Cycle 4, 1994-01-31 to 1995-06-30
4	Cycle 5	1995-06-30	1996-07-31	14	0.23230	23.23023	0.01527	1.52704	0.19607	19.60729	0.07840	-0.00750	-75.00000	-0.00643	-64.28571	Cycle 5, 1995-06-30 to 1996-07-31
5	Cycle 6	1996-07-31	1997-02-28	8	0.19592	19.59152	0.02336	2.33576	0.30783	30.78278	0.14361	0.00000	0.00000	0.00000	0.00000	Cycle 6, 1996-07-31 to 1997-02-28
6	Cycle 7	1997-02-28	1997-09-30	8	0.22017	22.01713	0.02631	2.63056	0.34782	34.78178	0.17547	0.00250	25.00000	0.00375	37.50000	Cycle 7, 1997-02-28 to 1997-09-30
7	Cycle 8	1997-09-30	1998-08-31	12	0.08094	8.09434	0.00812	0.81242	0.08094	8.09434	0.20028	0.00000	0.00000	0.00000	0.00000	Cycle 8, 1997-09-30 to 1998-08-31
8	Cycle 9	1998-08-31	1999-05-31	10	0.17553	17.55334	0.01849	1.84915	0.21418	21.41769	0.23543	-0.00750	-75.00000	-0.00900	-90.00000	Cycle 9, 1998-08-31 to 1999-05-31
9	Cycle 10	1999-05-31	2000-11-30	19	0.00401	0.40140	0.00127	0.12700	0.00253	0.25333	0.16483	0.01750	175.00000	0.01105	110.52632	Cycle 10, 1999-05-31 to 2000-11-30
10	Cycle 11	2000-11-30	2000-12-31	2	-0.07433	-7.43308	-0.03697	-3.69725	-0.37088	-37.08781	0.20511	0.00000	0.00000	0.00000	0.00000	Cycle 11, 2000-11-30 to 2000-12-31
11	Cycle 12	2000-12-31	2002-06-30	19	-0.23106	-23.10629	-0.01254	-1.25394	-0.15291	-15.29071	0.17313	-0.04750	-475.00000	-0.03000	-300.00000	Cycle 12, 2000-12-31 to 2002-06-30
12	Cycle 13	2002-06-30	2002-10-31	5	-0.16407	-16.40672	-0.03266	-3.26564	-0.34955	-34.95539	0.27616	0.00000	0.00000	0.00000	0.00000	Cycle 13, 2002-06-30 to 2002-10-31
13	Cycle 14	2002-10-31	2003-12-31	15	0.39543	39.54292	0.02325	2.32521	0.30547	30.54681	0.14377	-0.00750	-75.00000	-0.00600	-60.00000	Cycle 14, 2002-10-31 to 2003-12-31
14	Cycle 15	2003-12-31	2004-05-31	6	0.06792	6.79152	0.01127	1.12720	0.14044	14.04429	0.08737	0.00000	0.00000	0.00000	0.00000	Cycle 15, 2003-12-31 to 2004-05-31
15	Cycle 16	2004-05-31	2006-12-31	32	0.34572	34.57166	0.00952	0.95192	0.11778	11.77833	0.06998	0.04250	425.00000	0.01594	159.37500	Cycle 16, 2004-05-31 to 2006-12-31
16	Cycle 17	2006-12-31	2007-08-31	9	0.06671	6.67103	0.00748	0.74817	0.08992	8.99217	0.08722	0.00000	0.00000	0.00000	0.00000	Cycle 17, 2006-12-31 to 2007-08-31
17	Cycle 18	2007-08-31	2009-07-31	24	-0.28840	-28.83990	-0.01191	-1.19125	-0.15644	-15.64355	0.22888	-0.05000	-500.00000	-0.02500	-250.00000	Cycle 18, 2007-08-31 to 2009-07-31
18	Cycle 19	2009-07-31	2015-11-30	77	1.59039	159.03905	0.01314	1.31414	0.15990	15.99000	0.13125	0.00000	0.00000	0.00000	0.00000	Cycle 19, 2009-07-31 to 2015-11-30
19	Cycle 20	2015-11-30	2016-06-30	8	0.02502	2.50250	0.00356	0.35617	0.03777	3.77714	0.11389	0.00250	25.00000	0.00375	37.50000	Cycle 20, 2015-11-30 to 2016-06-30
20	Cycle 21	2016-06-30	2016-11-30	6	0.06008	6.00766	0.00997	0.99744	0.12376	12.37623	0.07708	0.00000	0.00000	0.00000	0.00000	Cycle 21, 2016-06-30 to 2016-11-30
21	Cycle 22	2016-11-30	2019-06-30	32	0.46031	46.03091	0.01256	1.25618	0.15257	15.25686	0.12685	0.02000	200.00000	0.00750	75.00000	Cycle 22, 2016-11-30 to 2019-06-30
22	Cycle 23	2019-06-30	2019-07-31	2	0.08586	8.58628	0.04242	4.24249	0.63927	63.92672	0.13743	0.00000	0.00000	0.00000	0.00000	Cycle 23, 2019-06-30 to 2019-07-31
23	Cycle 24	2019-07-31	2020-09-30	15	0.17105	17.10456	0.01234	1.23424	0.13464	13.46425	0.21140	-0.02250	-225.00000	-0.01800	-180.00000	Cycle 24, 2019-07-31 to 2020-09-30
24	Cycle 25	2020-09-30	2022-02-28	18	0.27728	27.72844	0.01457	1.45670	0.17722	17.72221	0.15078	0.00000	0.00000	0.00000	0.00000	Cycle 25, 2020-09-30 to 2022-02-28
25	Cycle 26	2022-02-28	2024-01-31	24	0.10892	10.89196	0.00584	0.58363	0.05305	5.30525	0.19480	0.05250	525.00000	0.02625	262.50000	Cycle 26, 2022-02-28 to 2024-01-31
26	Cycle 27	2024-01-31	2024-08-31	8	0.19526	19.52588	0.02293	2.29282	0.30675	30.67513	0.10238	0.00000	0.00000	0.00000	0.00000	Cycle 27, 2024-01-31 to 2024-08-31
27	Cycle 28	2024-08-31	2025-06-30	11	0.13779	13.77869	0.01244	1.24436	0.15122	15.12175	0.13032	-0.01000	-100.00000	-0.01091	-109.09091	Cycle 28, 2024-08-31 to 2025-06-30
28	Cycle 29	2025-06-30	2025-08-31	3	0.09622	9.62171	0.03119	3.11890	0.44406	44.40637	0.05911	0.00000	0.00000	0.00000	0.00000	Cycle 29, 2025-06-30 to 2025-08-31
29	Cycle 30	2025-08-31	2026-01-31	6	0.10133	10.13298	0.01629	1.62914	0.21293	21.29273	0.04688	-0.00750	-75.00000	-0.01500	-150.00000	Cycle 30, 2025-08-31 to 2026-01-31

This gives us the following data points:

Cycle start date
Cycle end date
Number of months in the cycle
Cumulative return during the cycle (decimal and percent)
Average monthly return during the cycle (decimal and percent)
Annualized return during the cycle (decimal and percent)
Return volatility during the cycle
Cumulative change in FFR during the cycle (decimal and basis points)
Annualized change in FFR during the cycle (decimal and basis points)

From the above DataFrame, we can then plot the cumulative and annualized returns for each cycle in a bar chart. First, the cumulative returns along with the cumulative change in FFR:

plot_bar_returns_ffr_change(
    cycle_df=spxt_cycles,
    asset_label="SPXT",
    annualized_or_cumulative="Cumulative",
)

png

And then the annualized returns along with the annualized change in FFR:

plot_bar_returns_ffr_change(
    cycle_df=spxt_cycles,
    asset_label="SPXT",
    annualized_or_cumulative="Annualized",
)

png

The cumulative returns plot is not particularly insightful, but there are some interesting observations to be gained from the annualized returns plot. During the past two (2) rate cutting cycles (cycles 11/12/13/14 and 18), stocks have exhibited negative returns during the rate cutting cycle. However, after the rate cutting cycle was complete, returns during the following cycle (when rates were usually flat) were quite strong and higher than the historical mean return for the S&P 500. The economic intuition for this behavior is valid; as the economy weakens, investors are concerned about the pricing of equities, the returns become negative, and the Fed responds with cutting rates. The exact timing of when the Fed begins cutting rates is one of the unknowns; the Fed could be ahead of the curve, cutting rates as economic data begins to prompt that action, or behind the curve, where the ecomony rolls over rapidly and even the Fed’s actions are not enough to halt the economic contraction.

Finally, we can run an OLS regression to check fit:

df = spxt_cycles

#=== Don't modify below this line ===

# Run OLS regression with statsmodels
X = df["FFR_AnnualizedChange_bps"]
y = df["AnnualizedReturnPct"]
X = sm.add_constant(X)
model = sm.OLS(y, X).fit()
print(model.summary())
print(f"Intercept: {model.params[0]}, Slope: {model.params[1]}")  # Intercept and slope

# Calc X and Y values for regression line
X_vals = np.linspace(X.min(), X.max(), 100)
Y_vals = model.params[0] + model.params[1] * X_vals

                             OLS Regression Results                            
===============================================================================
Dep. Variable:     AnnualizedReturnPct   R-squared:                       0.017
Model:                             OLS   Adj. R-squared:                 -0.018
Method:                  Least Squares   F-statistic:                    0.4786
Date:                 Tue, 24 Feb 2026   Prob (F-statistic):              0.495
Time:                         14:11:31   Log-Likelihood:                -132.24
No. Observations:                   30   AIC:                             268.5
Df Residuals:                       28   BIC:                             271.3
Df Model:                            1                                         
Covariance Type:             nonrobust                                         
============================================================================================
                               coef    std err          t      P>|t|      [0.025      0.975]
--------------------------------------------------------------------------------------------
const                       13.0760      3.792      3.448      0.002       5.308      20.844
FFR_AnnualizedChange_bps     0.0221      0.032      0.692      0.495      -0.043       0.087
==============================================================================
Omnibus:                        4.070   Durbin-Watson:                   1.523
Prob(Omnibus):                  0.131   Jarque-Bera (JB):                2.894
Skew:                          -0.320   Prob(JB):                        0.235
Kurtosis:                       4.381   Cond. No.                         120.
==============================================================================

Notes:
[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.
Intercept: 13.076015539043272, Slope: 0.022056651501767076

And then plot the regression line along with the values:

plot_scatter_regression_ffr_vs_returns(
    cycle_df=spxt_cycles,
    asset_label="SPXT",
    x_vals=X_vals,
    y_vals=Y_vals,
    intercept=model.params[0],
    slope=model.params[1],
)

png

Here we can see the data points for cycles 11/12/13/14 and 18 as mentioned above. Interestingly, cycles 28/29/30 (which is the current rate cutting cycle) appears to be an outlier. Of course, the book is not yet finished for cycles 28/29/30, and we could certainly see a bear market in stocks over the next several years.

Bonds #

Next, we’ll run a similar process for medium term bonds using SPBDU10T_S&P US Treasury Bond 7-10 Year Total Return Index.

First, we pull data with the following:

bb_clean_data(
    base_directory=DATA_DIR,
    fund_ticker_name="SPBDU10T_S&P US Treasury Bond 7-10 Year Total Return Index",
    source="Bloomberg",
    asset_class="Indices",
    excel_export=True,
    pickle_export=True,
    output_confirmation=False,
)

treas_10y = load_data(
    base_directory=DATA_DIR,
    ticker="SPBDU10T_S&P US Treasury Bond 7-10 Year Total Return Index_Clean",
    source="Bloomberg", 
    asset_class="Indices",
    timeframe="Daily",
    file_format="pickle",
)

# Filter TREAS_10Y to date range
treas_10y = treas_10y[(treas_10y.index >= pd.to_datetime(start_date)) & (treas_10y.index <= pd.to_datetime(end_date))]

# Drop everything except the "close" column
treas_10y = treas_10y[["Close"]]

# Resample to monthly frequency
treas_10y_monthly = treas_10y.resample("M").last()
treas_10y_monthly["Monthly_Return"] = treas_10y_monthly["Close"].pct_change()
treas_10y_monthly = treas_10y_monthly.dropna()

display(treas_10y_monthly)

	Close	Monthly_Return
Date
1990-01-31	98.01300	-0.01987
1990-02-28	97.99000	-0.00023
1990-03-31	97.98900	-0.00001
1990-04-30	96.60600	-0.01411
1990-05-31	99.64700	0.03148
...	...	...
2025-09-30	645.58400	0.00675
2025-10-31	650.00500	0.00685
2025-11-30	656.63600	0.01020
2025-12-31	652.28800	-0.00662
2026-01-31	650.16900	-0.00325

433 rows × 2 columns

Next, we can plot the price history before calculating the cycle performance:

plot_timeseries(
    price_df=treas_10y,
    plot_start_date=None,
    plot_end_date=None,
    plot_columns=["Close"],
    title="S&P US Treasury Bond 7-10 Year Total Return Index Daily Close Price",
    x_label="Date",
    x_format="Year",
    x_tick_spacing=2,
    x_tick_rotation=45,
    y_label="Price ($)",
    y_format="Decimal",
    y_format_decimal_places=0,
    y_tick_spacing=50,
    grid=True,
    legend=False,
    export_plot=False,
    plot_file_name=None,
)

png

Next, we will calculate the performance for SPY based on the pre-defined Fed cycles:

treas_10y_cycles = calc_fed_cycle_asset_performance(
    start_date=cycle_ranges["start_date"],
    end_date=cycle_ranges["end_date"],
    label=cycle_ranges["cycle_label"],
    fed_funds_change=cycle_ranges["fed_funds_change"],
    monthly_returns=treas_10y_monthly,
)

display(treas_10y_cycles)

	Cycle	Start	End	Months	CumulativeReturn	CumulativeReturnPct	AverageMonthlyReturn	AverageMonthlyReturnPct	AnnualizedReturn	AnnualizedReturnPct	Volatility	FedFundsChange	FedFundsChange_bps	FFR_AnnualizedChange	FFR_AnnualizedChange_bps	Label
0	Cycle 1	1989-12-31	1990-06-30	6	0.01362	1.36200	0.00241	0.24101	0.02743	2.74255	0.06657	0.00000	0.00000	0.00000	0.00000	Cycle 1, 1989-12-31 to 1990-06-30
1	Cycle 2	1990-06-30	1993-03-31	34	0.45682	45.68226	0.01124	1.12428	0.14202	14.20180	0.05346	-0.05250	-525.00000	-0.01853	-185.29412	Cycle 2, 1990-06-30 to 1993-03-31
2	Cycle 3	1993-03-31	1994-01-31	11	0.08876	8.87610	0.00784	0.78356	0.09721	9.72108	0.04458	0.00000	0.00000	0.00000	0.00000	Cycle 3, 1993-03-31 to 1994-01-31
3	Cycle 4	1994-01-31	1995-06-30	18	0.07796	7.79638	0.00439	0.43870	0.05132	5.13229	0.07287	0.03000	300.00000	0.02000	200.00000	Cycle 4, 1994-01-31 to 1995-06-30
4	Cycle 5	1995-06-30	1996-07-31	14	0.04785	4.78467	0.00344	0.34387	0.04087	4.08738	0.04940	-0.00750	-75.00000	-0.00643	-64.28571	Cycle 5, 1995-06-30 to 1996-07-31
5	Cycle 6	1996-07-31	1997-02-28	8	0.05301	5.30118	0.00658	0.65819	0.08056	8.05623	0.05365	0.00000	0.00000	0.00000	0.00000	Cycle 6, 1996-07-31 to 1997-02-28
6	Cycle 7	1997-02-28	1997-09-30	8	0.06453	6.45298	0.00799	0.79904	0.09834	9.83398	0.06285	0.00250	25.00000	0.00375	37.50000	Cycle 7, 1997-02-28 to 1997-09-30
7	Cycle 8	1997-09-30	1998-08-31	12	0.14428	14.42797	0.01135	1.13516	0.14428	14.42797	0.03891	0.00000	0.00000	0.00000	0.00000	Cycle 8, 1997-09-30 to 1998-08-31
8	Cycle 9	1998-08-31	1999-05-31	10	0.03052	3.05162	0.00326	0.32630	0.03673	3.67303	0.08244	-0.00750	-75.00000	-0.00900	-90.00000	Cycle 9, 1998-08-31 to 1999-05-31
9	Cycle 10	1999-05-31	2000-11-30	19	0.08744	8.74418	0.00449	0.44926	0.05437	5.43706	0.04250	0.01750	175.00000	0.01105	110.52632	Cycle 10, 1999-05-31 to 2000-11-30
10	Cycle 11	2000-11-30	2000-12-31	2	0.04918	4.91839	0.02430	2.42972	0.33386	33.38584	0.00458	0.00000	0.00000	0.00000	0.00000	Cycle 11, 2000-11-30 to 2000-12-31
11	Cycle 12	2000-12-31	2002-06-30	19	0.14331	14.33071	0.00721	0.72113	0.08826	8.82645	0.05901	-0.04750	-475.00000	-0.03000	-300.00000	Cycle 12, 2000-12-31 to 2002-06-30
12	Cycle 13	2002-06-30	2002-10-31	5	0.09945	9.94496	0.01929	1.92905	0.25551	25.55118	0.06682	0.00000	0.00000	0.00000	0.00000	Cycle 13, 2002-06-30 to 2002-10-31
13	Cycle 14	2002-10-31	2003-12-31	15	0.02161	2.16130	0.00172	0.17218	0.01725	1.72534	0.08696	-0.00750	-75.00000	-0.00600	-60.00000	Cycle 14, 2002-10-31 to 2003-12-31
14	Cycle 15	2003-12-31	2004-05-31	6	-0.00088	-0.08831	0.00006	0.00600	-0.00177	-0.17654	0.07653	0.00000	0.00000	0.00000	0.00000	Cycle 15, 2003-12-31 to 2004-05-31
15	Cycle 16	2004-05-31	2006-12-31	32	0.10267	10.26670	0.00314	0.31382	0.03733	3.73293	0.04440	0.04250	425.00000	0.01594	159.37500	Cycle 16, 2004-05-31 to 2006-12-31
16	Cycle 17	2006-12-31	2007-08-31	9	0.03300	3.30047	0.00371	0.37102	0.04425	4.42466	0.05098	0.00000	0.00000	0.00000	0.00000	Cycle 17, 2006-12-31 to 2007-08-31
17	Cycle 18	2007-08-31	2009-07-31	24	0.19843	19.84327	0.00792	0.79193	0.09473	9.47295	0.09432	-0.05000	-500.00000	-0.02500	-250.00000	Cycle 18, 2007-08-31 to 2009-07-31
18	Cycle 19	2009-07-31	2015-11-30	77	0.38995	38.99485	0.00443	0.44302	0.05265	5.26537	0.05945	0.00000	0.00000	0.00000	0.00000	Cycle 19, 2009-07-31 to 2015-11-30
19	Cycle 20	2015-11-30	2016-06-30	8	0.06790	6.78998	0.00835	0.83470	0.10356	10.35595	0.05317	0.00250	25.00000	0.00375	37.50000	Cycle 20, 2015-11-30 to 2016-06-30
20	Cycle 21	2016-06-30	2016-11-30	6	-0.03175	-3.17490	-0.00513	-0.51254	-0.06249	-6.24899	0.08242	0.00000	0.00000	0.00000	0.00000	Cycle 21, 2016-06-30 to 2016-11-30
21	Cycle 22	2016-11-30	2019-06-30	32	0.05923	5.92338	0.00190	0.19002	0.02181	2.18142	0.04975	0.02000	200.00000	0.00750	75.00000	Cycle 22, 2016-11-30 to 2019-06-30
22	Cycle 23	2019-06-30	2019-07-31	2	0.01306	1.30568	0.00653	0.65261	0.08094	8.09429	0.03024	0.00000	0.00000	0.00000	0.00000	Cycle 23, 2019-06-30 to 2019-07-31
23	Cycle 24	2019-07-31	2020-09-30	15	0.13083	13.08268	0.00837	0.83706	0.10336	10.33591	0.06056	-0.02250	-225.00000	-0.01800	-180.00000	Cycle 24, 2019-07-31 to 2020-09-30
24	Cycle 25	2020-09-30	2022-02-28	18	-0.06522	-6.52177	-0.00366	-0.36622	-0.04397	-4.39653	0.04430	0.00000	0.00000	0.00000	0.00000	Cycle 25, 2020-09-30 to 2022-02-28
25	Cycle 26	2022-02-28	2024-01-31	24	-0.09878	-9.87840	-0.00394	-0.39369	-0.05068	-5.06760	0.09858	0.05250	525.00000	0.02625	262.50000	Cycle 26, 2022-02-28 to 2024-01-31
26	Cycle 27	2024-01-31	2024-08-31	8	0.02748	2.74817	0.00357	0.35718	0.04150	4.15045	0.06957	0.00000	0.00000	0.00000	0.00000	Cycle 27, 2024-01-31 to 2024-08-31
27	Cycle 28	2024-08-31	2025-06-30	11	0.03294	3.29408	0.00310	0.31010	0.03599	3.59887	0.06280	-0.01000	-100.00000	-0.01091	-109.09091	Cycle 28, 2024-08-31 to 2025-06-30
28	Cycle 29	2025-06-30	2025-08-31	3	0.02670	2.66978	0.00888	0.88750	0.11114	11.11445	0.04409	0.00000	0.00000	0.00000	0.00000	Cycle 29, 2025-06-30 to 2025-08-31
29	Cycle 30	2025-08-31	2026-01-31	6	0.03081	3.08068	0.00510	0.51007	0.06256	6.25626	0.02992	-0.00750	-75.00000	-0.01500	-150.00000	Cycle 30, 2025-08-31 to 2026-01-31

This gives us the following data points:

Cycle start date
Cycle end date
Number of months in the cycle
Cumulative return during the cycle (decimal and percent)
Average monthly return during the cycle (decimal and percent)
Annualized return during the cycle (decimal and percent)
Return volatility during the cycle
Cumulative change in FFR during the cycle (decimal and basis points)
Annualized change in FFR during the cycle (decimal and basis points)

From the above DataFrame, we can then plot the cumulative and annualized returns for each cycle in a bar chart. First, the cumulative returns along with the cumulative change in FFR:

plot_bar_returns_ffr_change(
    cycle_df=treas_10y_cycles,
    asset_label="TREAS_10Y",
    annualized_or_cumulative="Cumulative",
)

png

And then the annualized returns along with the annualized change in FFR:

plot_bar_returns_ffr_change(
    cycle_df=treas_10y_cycles,
    asset_label="TREAS_10Y",
    annualized_or_cumulative="Annualized",
)

png

For this dataset, we have cycles 11/12/13/14 exhibiting strong returns, which is consistent with the economic intuition that bonds should perform well during periods of economic weakness and rate cuts. We also see this outcome with cycle 18. It becomes a little more interesting during cycles 25 and 26, where the correlations of stocks and bond returns seemed to align, so we see negative bond returns there. Finally, cycles 28/29/30 also exhibit positive bond returns, which is consistent with our thesis that bonds should perform well during periods of economic weakness and rate cuts.

Finally, we can run an OLS regression to check fit:

df = treas_10y_cycles

#=== Don't modify below this line ===

# Run OLS regression with statsmodels
X = df["FFR_AnnualizedChange_bps"]
y = df["AnnualizedReturnPct"]
X = sm.add_constant(X)
model = sm.OLS(y, X).fit()
print(model.summary())
print(f"Intercept: {model.params[0]}, Slope: {model.params[1]}")  # Intercept and slope

# Calc X and Y values for regression line
X_vals = np.linspace(X.min(), X.max(), 100)
Y_vals = model.params[0] + model.params[1] * X_vals

                             OLS Regression Results                            
===============================================================================
Dep. Variable:     AnnualizedReturnPct   R-squared:                       0.050
Model:                             OLS   Adj. R-squared:                  0.016
Method:                  Least Squares   F-statistic:                     1.462
Date:                 Tue, 24 Feb 2026   Prob (F-statistic):              0.237
Time:                         14:11:34   Log-Likelihood:                -103.64
No. Observations:                   30   AIC:                             211.3
Df Residuals:                       28   BIC:                             214.1
Df Model:                            1                                         
Covariance Type:             nonrobust                                         
============================================================================================
                               coef    std err          t      P>|t|      [0.025      0.975]
--------------------------------------------------------------------------------------------
const                        6.7457      1.462      4.613      0.000       3.751       9.741
FFR_AnnualizedChange_bps    -0.0149      0.012     -1.209      0.237      -0.040       0.010
==============================================================================
Omnibus:                       19.978   Durbin-Watson:                   2.116
Prob(Omnibus):                  0.000   Jarque-Bera (JB):               29.185
Skew:                           1.582   Prob(JB):                     4.60e-07
Kurtosis:                       6.651   Cond. No.                         120.
==============================================================================

Notes:
[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.
Intercept: 6.745718627069434, Slope: -0.01486255000436152

And then plot the regression line along with the values:

plot_scatter_regression_ffr_vs_returns(
    cycle_df=treas_10y_cycles,
    asset_label="TREAS_10Y",
    x_vals=X_vals,
    y_vals=Y_vals,
    intercept=model.params[0],
    slope=model.params[1],
)

png

The above plot is intriguing because of how well the OLS regression appears to fit the data. It certainly appears that during rate-cutting cycles, bonds are an asset that performs well.

High Yield Bonds #

Next, we’ll run a similar process for high yield bonds using LF98TRUU_Bloomberg US Corporate High Yield Total Return Index Value Unhedged USD.

First, we pull data with the following:

bb_clean_data(
    base_directory=DATA_DIR,
    fund_ticker_name="LF98TRUU_Bloomberg US Corporate High Yield Total Return Index Value Unhedged USD",
    source="Bloomberg",
    asset_class="Indices",
    excel_export=True,
    pickle_export=True,
    output_confirmation=False,
)

hy_bonds = load_data(
    base_directory=DATA_DIR,
    ticker="LF98TRUU_Bloomberg US Corporate High Yield Total Return Index Value Unhedged USD_Clean",
    source="Bloomberg", 
    asset_class="Indices",
    timeframe="Daily",
    file_format="pickle",
)

# Filter HY_BONDS to date range
hy_bonds = hy_bonds[(hy_bonds.index >= pd.to_datetime(start_date)) & (hy_bonds.index <= pd.to_datetime(end_date))]

# Drop everything except the "close" column
hy_bonds = hy_bonds[["Close"]]

# Resample to monthly frequency
hy_bonds_monthly = hy_bonds.resample("M").last()
hy_bonds_monthly["Monthly_Return"] = hy_bonds_monthly["Close"].pct_change()
hy_bonds_monthly = hy_bonds_monthly.dropna()

display(hy_bonds_monthly)

	Close	Monthly_Return
Date
1990-01-31	194.21000	-0.02146
1990-02-28	190.20000	-0.02065
1990-03-31	195.19000	0.02624
1990-04-30	194.86000	-0.00169
1990-05-31	198.62000	0.01930
...	...	...
2025-09-30	2876.85000	0.00816
2025-10-31	2881.38000	0.00157
2025-11-30	2898.07000	0.00579
2025-12-31	2914.49000	0.00567
2026-01-31	2929.32000	0.00509

433 rows × 2 columns

Next, we can plot the price history before calculating the cycle performance:

plot_timeseries(
    price_df=hy_bonds,
    plot_start_date=None,
    plot_end_date=None,
    plot_columns=["Close"],
    title="Bloomberg US Corporate High Yield Total Return Index Value Unhedged USD Daily Close Price",
    x_label="Date",
    x_format="Year",
    x_tick_spacing=2,
    x_tick_rotation=45,
    y_label="Price ($)",
    y_format="Decimal",
    y_format_decimal_places=0,
    y_tick_spacing=200,
    grid=True,
    legend=False,
    export_plot=False,
    plot_file_name=None,
)

png

Next, we will calculate the performance of high yield bonds based on the pre-defined Fed cycles:

hy_bonds_cycles = calc_fed_cycle_asset_performance(
    start_date=cycle_ranges["start_date"],
    end_date=cycle_ranges["end_date"],
    label=cycle_ranges["cycle_label"],
    fed_funds_change=cycle_ranges["fed_funds_change"],
    monthly_returns=hy_bonds_monthly,
)

display(hy_bonds_cycles)

	Cycle	Start	End	Months	CumulativeReturn	CumulativeReturnPct	AverageMonthlyReturn	AverageMonthlyReturnPct	AnnualizedReturn	AnnualizedReturnPct	Volatility	FedFundsChange	FedFundsChange_bps	FFR_AnnualizedChange	FFR_AnnualizedChange_bps	Label
0	Cycle 1	1989-12-31	1990-06-30	6	0.02494	2.49408	0.00432	0.43159	0.05050	5.05036	0.07625	0.00000	0.00000	0.00000	0.00000	Cycle 1, 1989-12-31 to 1990-06-30
1	Cycle 2	1990-06-30	1993-03-31	34	0.62149	62.14883	0.01480	1.47959	0.18601	18.60069	0.10898	-0.05250	-525.00000	-0.01853	-185.29412	Cycle 2, 1990-06-30 to 1993-03-31
2	Cycle 3	1993-03-31	1994-01-31	11	0.14262	14.26235	0.01221	1.22129	0.15656	15.65571	0.02227	0.00000	0.00000	0.00000	0.00000	Cycle 3, 1993-03-31 to 1994-01-31
3	Cycle 4	1994-01-31	1995-06-30	18	0.11254	11.25422	0.00607	0.60701	0.07369	7.36869	0.05691	0.03000	300.00000	0.02000	200.00000	Cycle 4, 1994-01-31 to 1995-06-30
4	Cycle 5	1995-06-30	1996-07-31	14	0.10892	10.89235	0.00743	0.74262	0.09267	9.26651	0.01906	-0.00750	-75.00000	-0.00643	-64.28571	Cycle 5, 1995-06-30 to 1996-07-31
5	Cycle 6	1996-07-31	1997-02-28	8	0.10479	10.47889	0.01255	1.25548	0.16123	16.12319	0.02366	0.00000	0.00000	0.00000	0.00000	Cycle 6, 1996-07-31 to 1997-02-28
6	Cycle 7	1997-02-28	1997-09-30	8	0.09557	9.55703	0.01156	1.15581	0.14673	14.67279	0.04792	0.00250	25.00000	0.00375	37.50000	Cycle 7, 1997-02-28 to 1997-09-30
7	Cycle 8	1997-09-30	1998-08-31	12	0.03219	3.21938	0.00282	0.28169	0.03219	3.21938	0.06631	0.00000	0.00000	0.00000	0.00000	Cycle 8, 1997-09-30 to 1998-08-31
8	Cycle 9	1998-08-31	1999-05-31	10	-0.00728	-0.72813	-0.00042	-0.04217	-0.00873	-0.87312	0.09026	-0.00750	-75.00000	-0.00900	-90.00000	Cycle 9, 1998-08-31 to 1999-05-31
9	Cycle 10	1999-05-31	2000-11-30	19	-0.08916	-8.91594	-0.00480	-0.47963	-0.05728	-5.72758	0.05172	0.01750	175.00000	0.01105	110.52632	Cycle 10, 1999-05-31 to 2000-11-30
10	Cycle 11	2000-11-30	2000-12-31	2	-0.02105	-2.10510	-0.01014	-1.01430	-0.11984	-11.98427	0.14432	0.00000	0.00000	0.00000	0.00000	Cycle 11, 2000-11-30 to 2000-12-31
11	Cycle 12	2000-12-31	2002-06-30	19	0.02117	2.11701	0.00167	0.16693	0.01332	1.33189	0.11914	-0.04750	-475.00000	-0.03000	-300.00000	Cycle 12, 2000-12-31 to 2002-06-30
12	Cycle 13	2002-06-30	2002-10-31	5	-0.10874	-10.87431	-0.02215	-2.21511	-0.24141	-24.14102	0.13364	0.00000	0.00000	0.00000	0.00000	Cycle 13, 2002-06-30 to 2002-10-31
13	Cycle 14	2002-10-31	2003-12-31	15	0.37662	37.66182	0.02172	2.17230	0.29137	29.13703	0.07004	-0.00750	-75.00000	-0.00600	-60.00000	Cycle 14, 2002-10-31 to 2003-12-31
14	Cycle 15	2003-12-31	2004-05-31	6	0.02188	2.18782	0.00371	0.37115	0.04424	4.42350	0.05322	0.00000	0.00000	0.00000	0.00000	Cycle 15, 2003-12-31 to 2004-05-31
15	Cycle 16	2004-05-31	2006-12-31	32	0.25625	25.62459	0.00722	0.72211	0.08931	8.93138	0.04056	0.04250	425.00000	0.01594	159.37500	Cycle 16, 2004-05-31 to 2006-12-31
16	Cycle 17	2006-12-31	2007-08-31	9	0.01683	1.68350	0.00199	0.19914	0.02251	2.25094	0.05992	0.00000	0.00000	0.00000	0.00000	Cycle 17, 2006-12-31 to 2007-08-31
17	Cycle 18	2007-08-31	2009-07-31	24	0.04908	4.90805	0.00371	0.37092	0.02425	2.42463	0.20404	-0.05000	-500.00000	-0.02500	-250.00000	Cycle 18, 2007-08-31 to 2009-07-31
18	Cycle 19	2009-07-31	2015-11-30	77	0.83146	83.14599	0.00808	0.80820	0.09889	9.88931	0.06867	0.00000	0.00000	0.00000	0.00000	Cycle 19, 2009-07-31 to 2015-11-30
19	Cycle 20	2015-11-30	2016-06-30	8	0.03948	3.94781	0.00515	0.51508	0.05980	5.97977	0.09107	0.00250	25.00000	0.00375	37.50000	Cycle 20, 2015-11-30 to 2016-06-30
20	Cycle 21	2016-06-30	2016-11-30	6	0.06426	6.42628	0.01049	1.04900	0.13266	13.26554	0.04025	0.00000	0.00000	0.00000	0.00000	Cycle 21, 2016-06-30 to 2016-11-30
21	Cycle 22	2016-11-30	2019-06-30	32	0.17306	17.30568	0.00508	0.50759	0.06168	6.16825	0.04347	0.02000	200.00000	0.00750	75.00000	Cycle 22, 2016-11-30 to 2019-06-30
22	Cycle 23	2019-06-30	2019-07-31	2	0.02856	2.85572	0.01421	1.42144	0.18405	18.40520	0.04203	0.00000	0.00000	0.00000	0.00000	Cycle 23, 2019-06-30 to 2019-07-31
23	Cycle 24	2019-07-31	2020-09-30	15	0.04631	4.63140	0.00372	0.37244	0.03688	3.68827	0.13113	-0.02250	-225.00000	-0.01800	-180.00000	Cycle 24, 2019-07-31 to 2020-09-30
24	Cycle 25	2020-09-30	2022-02-28	18	0.06775	6.77496	0.00374	0.37444	0.04467	4.46712	0.04952	0.00000	0.00000	0.00000	0.00000	Cycle 25, 2020-09-30 to 2022-02-28
25	Cycle 26	2022-02-28	2024-01-31	24	0.03580	3.58042	0.00186	0.18605	0.01774	1.77447	0.09916	0.05250	525.00000	0.02625	262.50000	Cycle 26, 2022-02-28 to 2024-01-31
26	Cycle 27	2024-01-31	2024-08-31	8	0.06285	6.28480	0.00769	0.76866	0.09574	9.57381	0.03255	0.00000	0.00000	0.00000	0.00000	Cycle 27, 2024-01-31 to 2024-08-31
27	Cycle 28	2024-08-31	2025-06-30	11	0.08182	8.18158	0.00722	0.72241	0.08958	8.95776	0.03617	-0.01000	-100.00000	-0.01091	-109.09091	Cycle 28, 2024-08-31 to 2025-06-30
28	Cycle 29	2025-06-30	2025-08-31	3	0.03577	3.57673	0.01180	1.17991	0.15093	15.09297	0.02414	0.00000	0.00000	0.00000	0.00000	Cycle 29, 2025-06-30 to 2025-08-31
29	Cycle 30	2025-08-31	2026-01-31	6	0.03937	3.93741	0.00646	0.64627	0.08030	8.02985	0.01260	-0.00750	-75.00000	-0.01500	-150.00000	Cycle 30, 2025-08-31 to 2026-01-31

This gives us the following data points:

Cycle start date
Cycle end date
Number of months in the cycle
Cumulative return during the cycle (decimal and percent)
Average monthly return during the cycle (decimal and percent)
Annualized return during the cycle (decimal and percent)
Return volatility during the cycle
Cumulative change in FFR during the cycle (decimal and basis points)
Annualized change in FFR during the cycle (decimal and basis points)

From the above DataFrame, we can then plot the cumulative and annualized returns for each cycle in a bar chart. First, the cumulative returns along with the cumulative change in FFR:

plot_bar_returns_ffr_change(
    cycle_df=hy_bonds_cycles,
    asset_label="HY_BONDS",
    annualized_or_cumulative="Cumulative",
)

png

And then the annualized returns along with the annualized change in FFR:

plot_bar_returns_ffr_change(
    cycle_df=hy_bonds_cycles,
    asset_label="HY_BONDS",
    annualized_or_cumulative="Annualized",
)

png

Finally, we can run an OLS regression to check fit:

df = hy_bonds_cycles

#=== Don't modify below this line ===

# Run OLS regression with statsmodels
X = df["FFR_AnnualizedChange_bps"]
y = df["AnnualizedReturnPct"]
X = sm.add_constant(X)
model = sm.OLS(y, X).fit()
print(model.summary())
print(f"Intercept: {model.params[0]}, Slope: {model.params[1]}")  # Intercept and slope

# Calc X and Y values for regression line
X_vals = np.linspace(X.min(), X.max(), 100)
Y_vals = model.params[0] + model.params[1] * X_vals

                             OLS Regression Results                            
===============================================================================
Dep. Variable:     AnnualizedReturnPct   R-squared:                       0.004
Model:                             OLS   Adj. R-squared:                 -0.031
Method:                  Least Squares   F-statistic:                    0.1171
Date:                 Tue, 24 Feb 2026   Prob (F-statistic):              0.735
Time:                         14:11:36   Log-Likelihood:                -110.57
No. Observations:                   30   AIC:                             225.1
Df Residuals:                       28   BIC:                             227.9
Df Model:                            1                                         
Covariance Type:             nonrobust                                         
============================================================================================
                               coef    std err          t      P>|t|      [0.025      0.975]
--------------------------------------------------------------------------------------------
const                        6.6113      1.842      3.590      0.001       2.839      10.384
FFR_AnnualizedChange_bps    -0.0053      0.015     -0.342      0.735      -0.037       0.026
==============================================================================
Omnibus:                        9.087   Durbin-Watson:                   1.936
Prob(Omnibus):                  0.011   Jarque-Bera (JB):                9.047
Skew:                          -0.790   Prob(JB):                       0.0109
Kurtosis:                       5.178   Cond. No.                         120.
==============================================================================

Notes:
[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.
Intercept: 6.611340605920211, Slope: -0.005299161729553994

And then plot the regression line along with the values:

plot_scatter_regression_ffr_vs_returns(
    cycle_df=treas_10y_cycles,
    asset_label="TREAS_10Y",
    x_vals=X_vals,
    y_vals=Y_vals,
    intercept=model.params[0],
    slope=model.params[1],
)

png

Gold #

Finally, we’ll run a similar process for gold using XAU_Gold USD Spot.

First, we pull data with the following:

bb_clean_data(
    base_directory=DATA_DIR,
    fund_ticker_name="XAU_Gold USD Spot",
    source="Bloomberg",
    asset_class="Commodities",
    excel_export=True,
    pickle_export=True,
    output_confirmation=False,
)

gold = load_data(
    base_directory=DATA_DIR,
    ticker="XAU_Gold USD Spot_Clean",
    source="Bloomberg", 
    asset_class="Commodities",
    timeframe="Daily",
    file_format="pickle",
)

# Filter GOLD to date range
gold = gold[(gold.index >= pd.to_datetime(start_date)) & (gold.index <= pd.to_datetime(end_date))]

# Drop everything except the "close" column
gold = gold[["Close"]]

# Resample to monthly frequency
gold_monthly = gold.resample("M").last()
gold_monthly["Monthly_Return"] = gold_monthly["Close"].pct_change()
gold_monthly = gold_monthly.dropna()

display(gold_monthly)

	Close	Monthly_Return
Date
1990-01-31	415.05000	0.03439
1990-02-28	407.70000	-0.01771
1990-03-31	368.50000	-0.09615
1990-04-30	367.75000	-0.00204
1990-05-31	363.05000	-0.01278
...	...	...
2025-09-30	3858.96000	0.11920
2025-10-31	4002.92000	0.03731
2025-11-30	4239.43000	0.05908
2025-12-31	4319.37000	0.01886
2026-01-31	4894.23000	0.13309

433 rows × 2 columns

Next, we can plot the price history before calculating the cycle performance:

plot_timeseries(
    price_df=gold,
    plot_start_date=None,
    plot_end_date=None,
    plot_columns=["Close"],
    title="XAU Gold USD Spot Daily Close Price",
    x_label="Date",
    x_format="Year",
    x_tick_spacing=2,
    x_tick_rotation=45,
    y_label="Price ($)",
    y_format="Decimal",
    y_format_decimal_places=0,
    y_tick_spacing=400,
    grid=True,
    legend=False,
    export_plot=False,
    plot_file_name=None,
)

png

Next, we will calculate the performance of gold based on the pre-defined Fed cycles:

gold_cycles = calc_fed_cycle_asset_performance(
    start_date=cycle_ranges["start_date"],
    end_date=cycle_ranges["end_date"],
    label=cycle_ranges["cycle_label"],
    fed_funds_change=cycle_ranges["fed_funds_change"],
    monthly_returns=gold_monthly,
)

display(gold_cycles)

	Cycle	Start	End	Months	CumulativeReturn	CumulativeReturnPct	AverageMonthlyReturn	AverageMonthlyReturnPct	AnnualizedReturn	AnnualizedReturnPct	Volatility	FedFundsChange	FedFundsChange_bps	FFR_AnnualizedChange	FFR_AnnualizedChange_bps	Label
0	Cycle 1	1989-12-31	1990-06-30	6	-0.12224	-12.22430	-0.02069	-2.06945	-0.22954	-22.95426	0.14885	0.00000	0.00000	0.00000	0.00000	Cycle 1, 1989-12-31 to 1990-06-30
1	Cycle 2	1990-06-30	1993-03-31	34	-0.06624	-6.62443	-0.00158	-0.15839	-0.02390	-2.39005	0.10292	-0.05250	-525.00000	-0.01853	-185.29412	Cycle 2, 1990-06-30 to 1993-03-31
2	Cycle 3	1993-03-31	1994-01-31	11	0.15903	15.90288	0.01462	1.46238	0.17468	17.46838	0.17075	0.00000	0.00000	0.00000	0.00000	Cycle 3, 1993-03-31 to 1994-01-31
3	Cycle 4	1994-01-31	1995-06-30	18	-0.01561	-1.56130	-0.00070	-0.06966	-0.01044	-1.04359	0.06717	0.03000	300.00000	0.02000	200.00000	Cycle 4, 1994-01-31 to 1995-06-30
4	Cycle 5	1995-06-30	1996-07-31	14	0.00716	0.71559	0.00066	0.06552	0.00613	0.61305	0.06186	-0.00750	-75.00000	-0.00643	-64.28571	Cycle 5, 1995-06-30 to 1996-07-31
5	Cycle 6	1996-07-31	1997-02-28	8	-0.04468	-4.46839	-0.00519	-0.51936	-0.06627	-6.62715	0.11737	0.00000	0.00000	0.00000	0.00000	Cycle 6, 1996-07-31 to 1997-02-28
6	Cycle 7	1997-02-28	1997-09-30	8	-0.02875	-2.87498	-0.00313	-0.31292	-0.04281	-4.28133	0.11886	0.00250	25.00000	0.00375	37.50000	Cycle 7, 1997-02-28 to 1997-09-30
7	Cycle 8	1997-09-30	1998-08-31	12	-0.14993	-14.99306	-0.01285	-1.28489	-0.14993	-14.99306	0.12433	0.00000	0.00000	0.00000	0.00000	Cycle 8, 1997-09-30 to 1998-08-31
8	Cycle 9	1998-08-31	1999-05-31	10	-0.05621	-5.62053	-0.00517	-0.51718	-0.06706	-6.70614	0.12709	-0.00750	-75.00000	-0.00900	-90.00000	Cycle 9, 1998-08-31 to 1999-05-31
9	Cycle 10	1999-05-31	2000-11-30	19	-0.05619	-5.61857	-0.00193	-0.19269	-0.03586	-3.58627	0.17333	0.01750	175.00000	0.01105	110.52632	Cycle 10, 1999-05-31 to 2000-11-30
10	Cycle 11	2000-11-30	2000-12-31	2	0.02678	2.67773	0.01332	1.33221	0.17181	17.18109	0.03266	0.00000	0.00000	0.00000	0.00000	Cycle 11, 2000-11-30 to 2000-12-31
11	Cycle 12	2000-12-31	2002-06-30	19	0.16269	16.26918	0.00844	0.84406	0.09988	9.98819	0.11022	-0.04750	-475.00000	-0.03000	-300.00000	Cycle 12, 2000-12-31 to 2002-06-30
12	Cycle 13	2002-06-30	2002-10-31	5	-0.02695	-2.69484	-0.00496	-0.49607	-0.06346	-6.34605	0.12104	0.00000	0.00000	0.00000	0.00000	Cycle 13, 2002-06-30 to 2002-10-31
13	Cycle 14	2002-10-31	2003-12-31	15	0.28404	28.40365	0.01772	1.77192	0.22141	22.14112	0.15439	-0.00750	-75.00000	-0.00600	-60.00000	Cycle 14, 2002-10-31 to 2003-12-31
14	Cycle 15	2003-12-31	2004-05-31	6	-0.00653	-0.65302	0.00044	0.04419	-0.01302	-1.30178	0.20868	0.00000	0.00000	0.00000	0.00000	Cycle 15, 2003-12-31 to 2004-05-31
15	Cycle 16	2004-05-31	2006-12-31	32	0.64628	64.62831	0.01653	1.65326	0.20556	20.55610	0.14568	0.04250	425.00000	0.01594	159.37500	Cycle 16, 2004-05-31 to 2006-12-31
16	Cycle 17	2006-12-31	2007-08-31	9	0.03904	3.90432	0.00447	0.44660	0.05239	5.23935	0.07374	0.00000	0.00000	0.00000	0.00000	Cycle 17, 2006-12-31 to 2007-08-31
17	Cycle 18	2007-08-31	2009-07-31	24	0.43610	43.60981	0.01766	1.76641	0.19837	19.83731	0.24658	-0.05000	-500.00000	-0.02500	-250.00000	Cycle 18, 2007-08-31 to 2009-07-31
18	Cycle 19	2009-07-31	2015-11-30	77	0.14924	14.92391	0.00317	0.31742	0.02191	2.19146	0.18253	0.00000	0.00000	0.00000	0.00000	Cycle 19, 2009-07-31 to 2015-11-30
19	Cycle 20	2015-11-30	2016-06-30	8	0.15742	15.74192	0.02027	2.02679	0.24519	24.51911	0.22531	0.00250	25.00000	0.00375	37.50000	Cycle 20, 2015-11-30 to 2016-06-30
20	Cycle 21	2016-06-30	2016-11-30	6	-0.03466	-3.46575	-0.00449	-0.44887	-0.06811	-6.81139	0.19940	0.00000	0.00000	0.00000	0.00000	Cycle 21, 2016-06-30 to 2016-11-30
21	Cycle 22	2016-11-30	2019-06-30	32	0.10362	10.36165	0.00355	0.35519	0.03766	3.76642	0.10758	0.02000	200.00000	0.00750	75.00000	Cycle 22, 2016-11-30 to 2019-06-30
22	Cycle 23	2019-06-30	2019-07-31	2	0.08288	8.28750	0.04132	4.13180	0.61239	61.23899	0.18771	0.00000	0.00000	0.00000	0.00000	Cycle 23, 2019-06-30 to 2019-07-31
23	Cycle 24	2019-07-31	2020-09-30	15	0.33789	33.78880	0.02043	2.04336	0.26222	26.22222	0.14899	-0.02250	-225.00000	-0.01800	-180.00000	Cycle 24, 2019-07-31 to 2020-09-30
24	Cycle 25	2020-09-30	2022-02-28	18	-0.02989	-2.98862	-0.00076	-0.07561	-0.02002	-2.00247	0.15388	0.00000	0.00000	0.00000	0.00000	Cycle 25, 2020-09-30 to 2022-02-28
25	Cycle 26	2022-02-28	2024-01-31	24	0.13485	13.48509	0.00605	0.60516	0.06529	6.52938	0.13996	0.05250	525.00000	0.02625	262.50000	Cycle 26, 2022-02-28 to 2024-01-31
26	Cycle 27	2024-01-31	2024-08-31	8	0.21348	21.34824	0.02494	2.49353	0.33675	33.67502	0.11399	0.00000	0.00000	0.00000	0.00000	Cycle 27, 2024-01-31 to 2024-08-31
27	Cycle 28	2024-08-31	2025-06-30	11	0.34954	34.95424	0.02824	2.82419	0.38683	38.68250	0.12923	-0.01000	-100.00000	-0.01091	-109.09091	Cycle 28, 2024-08-31 to 2025-06-30
28	Cycle 29	2025-06-30	2025-08-31	3	0.04825	4.82481	0.01609	1.60850	0.20741	20.74143	0.09689	0.00000	0.00000	0.00000	0.00000	Cycle 29, 2025-06-30 to 2025-08-31
29	Cycle 30	2025-08-31	2026-01-31	6	0.48764	48.76396	0.06926	6.92618	1.21307	121.30714	0.16013	-0.00750	-75.00000	-0.01500	-150.00000	Cycle 30, 2025-08-31 to 2026-01-31

This gives us the following data points:

Cycle start date
Cycle end date
Number of months in the cycle
Cumulative return during the cycle (decimal and percent)
Average monthly return during the cycle (decimal and percent)
Annualized return during the cycle (decimal and percent)
Return volatility during the cycle
Cumulative change in FFR during the cycle (decimal and basis points)
Annualized change in FFR during the cycle (decimal and basis points)

From the above DataFrame, we can then plot the cumulative and annualized returns for each cycle in a bar chart. First, the cumulative returns along with the cumulative change in FFR:

plot_bar_returns_ffr_change(
    cycle_df=gold_cycles,
    asset_label="GOLD",
    annualized_or_cumulative="Cumulative",
)

png

And then the annualized returns along with the annualized change in FFR:

plot_bar_returns_ffr_change(
    cycle_df=gold_cycles,
    asset_label="GOLD",
    annualized_or_cumulative="Annualized",
)

png

We see strong returns for gold across several different Fed cycles, so it is difficult to draw any kind of initial conclusion based on the bar charts.

Finally, we can run an OLS regression to check fit:

df = gold_cycles

#=== Don't modify below this line ===

# Run OLS regression with statsmodels
X = df["FFR_AnnualizedChange_bps"]
y = df["AnnualizedReturnPct"]
X = sm.add_constant(X)
model = sm.OLS(y, X).fit()
print(model.summary())
print(f"Intercept: {model.params[0]}, Slope: {model.params[1]}")  # Intercept and slope

# Calc X and Y values for regression line
X_vals = np.linspace(X.min(), X.max(), 100)
Y_vals = model.params[0] + model.params[1] * X_vals

                             OLS Regression Results                            
===============================================================================
Dep. Variable:     AnnualizedReturnPct   R-squared:                       0.064
Model:                             OLS   Adj. R-squared:                  0.030
Method:                  Least Squares   F-statistic:                     1.900
Date:                 Tue, 24 Feb 2026   Prob (F-statistic):              0.179
Time:                         14:11:39   Log-Likelihood:                -140.03
No. Observations:                   30   AIC:                             284.1
Df Residuals:                       28   BIC:                             286.9
Df Model:                            1                                         
Covariance Type:             nonrobust                                         
============================================================================================
                               coef    std err          t      P>|t|      [0.025      0.975]
--------------------------------------------------------------------------------------------
const                       11.4670      4.917      2.332      0.027       1.396      21.539
FFR_AnnualizedChange_bps    -0.0570      0.041     -1.378      0.179      -0.142       0.028
==============================================================================
Omnibus:                       28.991   Durbin-Watson:                   1.081
Prob(Omnibus):                  0.000   Jarque-Bera (JB):               62.894
Skew:                           2.093   Prob(JB):                     2.20e-14
Kurtosis:                       8.727   Cond. No.                         120.
==============================================================================

Notes:
[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.
Intercept: 11.467013480569243, Slope: -0.056974234027166226

And then plot the regression line along with the values:

plot_scatter_regression_ffr_vs_returns(
    cycle_df=gold_cycles,
    asset_label="GOLD",
    x_vals=X_vals,
    y_vals=Y_vals,
    intercept=model.params[0],
    slope=model.params[1],
)

png

It’s difficult to draw any strong conclusions with the above plot. Gold has traditionally been considered a hedge for inflation, and while one of the Fed’s mandates is to manage inflation, there may not be a conclusion to draw in relationship to the historical returns that gold has exhibited.

Asset Allocation #

With the above analysis (somewhat) complete, let’s look at the optimal allocation for a portfolio based on the data and the hypythetical historical results.

We have to be careful with our criteria for when to hold stocks, bonds, or gold, as hindsight bias is certainly possible. So, without overanalyzing the results, let’s assume that we hold stocks as the default position during tightening cycles, and then hold bonds during easing cycles when the Fed starts cutting rates, and then resume holding stocks when the Fed stops cutting rates. If there is not any change in FFR, then we still hold stocks.

We can then combine the return series based on the above with the following:

# Shift the "cycle_filled" column down by one row to create a new column called "cycle_invested" that represents the cycle label that an investor would be invested in for each month (i.e. the cycle label from the previous month)

fedfunds_grouped_cycles['cycle_invested'] = fedfunds_grouped_cycles['cycle_filled'].shift(1)

display(fedfunds_grouped_cycles)

	start_date	fed_funds_start	end_date	fed_funds_end	fed_funds_change	cycle	cycle_filled	group	cycle_invested
0	1989-12-31	0.08250	1990-01-31	0.08250	0.00000	Neutral	Modified Tightening	1	None
1	1990-01-31	0.08250	1990-02-28	0.08250	0.00000	Neutral	Modified Tightening	1	Modified Tightening
2	1990-02-28	0.08250	1990-03-31	0.08250	0.00000	Neutral	Modified Tightening	1	Modified Tightening
3	1990-03-31	0.08250	1990-04-30	0.08250	0.00000	Neutral	Modified Tightening	1	Modified Tightening
4	1990-04-30	0.08250	1990-05-31	0.08250	0.00000	Neutral	Modified Tightening	1	Modified Tightening
...	...	...	...	...	...	...	...	...	...
429	2025-09-30	0.04250	2025-10-31	0.04000	-0.00250	Easing	Easing	30	Easing
430	2025-10-31	0.04000	2025-11-30	0.04000	0.00000	Neutral	Easing	30	Easing
431	2025-11-30	0.04000	2025-12-31	0.03750	-0.00250	Easing	Easing	30	Easing
432	2025-12-31	0.03750	2026-01-31	0.03750	0.00000	Neutral	Easing	30	Easing
433	2026-01-31	0.03750	NaT	NaN	NaN	Neutral	Easing	30	Easing

434 rows × 9 columns

# Reset index to merge on date
stocks_merged = pd.merge_asof(
    spxt_monthly.reset_index(),
    fedfunds_grouped_cycles[['start_date', 'end_date', 'cycle_invested', 'group']],
    left_on='Date',
    right_on='start_date',
    direction='backward'
)

# Drop rows where the date falls outside the cycle's end_date
stocks_merged = stocks_merged[stocks_merged['Date'] <= stocks_merged['end_date']]

display(stocks_merged)

	Date	Close	Monthly_Return	start_date	end_date	cycle_invested	group
0	1990-01-31	353.94000	-0.06713	1990-01-31	1990-02-28	Modified Tightening	1
1	1990-02-28	358.50000	0.01288	1990-02-28	1990-03-31	Modified Tightening	1
2	1990-03-31	368	0.02650	1990-03-31	1990-04-30	Modified Tightening	1
3	1990-04-30	358.81000	-0.02497	1990-04-30	1990-05-31	Modified Tightening	1
4	1990-05-31	393.80000	0.09752	1990-05-31	1990-06-30	Modified Tightening	1
...	...	...	...	...	...	...	...
427	2025-08-31	14304.68000	0.02027	2025-08-31	2025-09-30	Modified Tightening	30
428	2025-09-30	14826.80000	0.03650	2025-09-30	2025-10-31	Easing	30
429	2025-10-31	15173.95000	0.02341	2025-10-31	2025-11-30	Easing	30
430	2025-11-30	15211.14000	0.00245	2025-11-30	2025-12-31	Easing	30
431	2025-12-31	15220.45000	0.00061	2025-12-31	2026-01-31	Easing	30

432 rows × 7 columns

# Reset index to merge on date
bonds_merged = pd.merge_asof(
    treas_10y_monthly.reset_index(),
    fedfunds_grouped_cycles[['start_date', 'end_date', 'cycle_invested', 'group']],
    left_on='Date',
    right_on='start_date',
    direction='backward'
)

# Drop rows where the date falls outside the cycle's end_date
bonds_merged = bonds_merged[bonds_merged['Date'] <= bonds_merged['end_date']]

display(bonds_merged)

	Date	Close	Monthly_Return	start_date	end_date	cycle_invested	group
0	1990-01-31	98.01300	-0.01987	1990-01-31	1990-02-28	Modified Tightening	1
1	1990-02-28	97.99000	-0.00023	1990-02-28	1990-03-31	Modified Tightening	1
2	1990-03-31	97.98900	-0.00001	1990-03-31	1990-04-30	Modified Tightening	1
3	1990-04-30	96.60600	-0.01411	1990-04-30	1990-05-31	Modified Tightening	1
4	1990-05-31	99.64700	0.03148	1990-05-31	1990-06-30	Modified Tightening	1
...	...	...	...	...	...	...	...
427	2025-08-31	641.25800	0.01668	2025-08-31	2025-09-30	Modified Tightening	30
428	2025-09-30	645.58400	0.00675	2025-09-30	2025-10-31	Easing	30
429	2025-10-31	650.00500	0.00685	2025-10-31	2025-11-30	Easing	30
430	2025-11-30	656.63600	0.01020	2025-11-30	2025-12-31	Easing	30
431	2025-12-31	652.28800	-0.00662	2025-12-31	2026-01-31	Easing	30

432 rows × 7 columns

# Select the appropriate return based on cycle
stocks_merged['strategy_return'] = stocks_merged.apply(
    lambda row: row['Monthly_Return'] if row['cycle_invested'] in ['Tightening', 'Modified Tightening'] else None,
    axis=1
)

bonds_merged['strategy_return'] = bonds_merged.apply(
    lambda row: row['Monthly_Return'] if row['cycle_invested'] == 'Easing' else None,
    axis=1
)

# Combine
strategy = pd.concat([stocks_merged, bonds_merged]).dropna(subset=['strategy_return'])
strategy = strategy.sort_values('Date')

display(strategy.head(20))

	Date	Close	Monthly_Return	start_date	end_date	cycle_invested	group	strategy_return
0	1990-01-31	353.94000	-0.06713	1990-01-31	1990-02-28	Modified Tightening	1	-0.06713
1	1990-02-28	358.50000	0.01288	1990-02-28	1990-03-31	Modified Tightening	1	0.01288
2	1990-03-31	368	0.02650	1990-03-31	1990-04-30	Modified Tightening	1	0.02650
3	1990-04-30	358.81000	-0.02497	1990-04-30	1990-05-31	Modified Tightening	1	-0.02497
4	1990-05-31	393.80000	0.09752	1990-05-31	1990-06-30	Modified Tightening	1	0.09752
5	1990-06-30	391.14000	-0.00675	1990-06-30	1990-07-31	Modified Tightening	2	-0.00675
6	1990-07-31	102.84500	0.01463	1990-07-31	1990-08-31	Easing	2	0.01463
7	1990-08-31	100.78200	-0.02006	1990-08-31	1990-09-30	Easing	2	-0.02006
8	1990-09-30	101.74000	0.00951	1990-09-30	1990-10-31	Easing	2	0.00951
9	1990-10-31	103.65200	0.01879	1990-10-31	1990-11-30	Easing	2	0.01879
10	1990-11-30	106.43100	0.02681	1990-11-30	1990-12-31	Easing	2	0.02681
11	1990-12-31	108.24900	0.01708	1990-12-31	1991-01-31	Easing	2	0.01708
12	1991-01-31	109.52500	0.01179	1991-01-31	1991-02-28	Easing	2	0.01179
13	1991-02-28	110.10400	0.00529	1991-02-28	1991-03-31	Easing	2	0.00529
14	1991-03-31	110.43400	0.00300	1991-03-31	1991-04-30	Easing	2	0.00300
15	1991-04-30	111.59800	0.01054	1991-04-30	1991-05-31	Easing	2	0.01054
16	1991-05-31	112.07800	0.00430	1991-05-31	1991-06-30	Easing	2	0.00430
17	1991-06-30	111.50200	-0.00514	1991-06-30	1991-07-31	Easing	2	-0.00514
18	1991-07-31	113.05200	0.01390	1991-07-31	1991-08-31	Easing	2	0.01390
19	1991-08-31	116.06500	0.02665	1991-08-31	1991-09-30	Easing	2	0.02665

# Calculate cumulative returns and drawdown for spxt
spxt_monthly['Cumulative_Return'] = (1 + spxt_monthly['Monthly_Return']).cumprod() - 1
spxt_monthly['Cumulative_Return_Plus_One'] = 1 + spxt_monthly['Cumulative_Return']
spxt_monthly['Rolling_Max'] = spxt_monthly['Cumulative_Return_Plus_One'].cummax()
spxt_monthly['Drawdown'] = spxt_monthly['Cumulative_Return_Plus_One'] / spxt_monthly['Rolling_Max'] - 1
spxt_monthly.drop(columns=['Cumulative_Return_Plus_One', 'Rolling_Max'], inplace=True)

# Calculate cumulative returns and drawdown for treas_10y
treas_10y_monthly['Cumulative_Return'] = (1 + treas_10y_monthly['Monthly_Return']).cumprod() - 1
treas_10y_monthly['Cumulative_Return_Plus_One'] = 1 + treas_10y_monthly['Cumulative_Return']
treas_10y_monthly['Rolling_Max'] = treas_10y_monthly['Cumulative_Return_Plus_One'].cummax()
treas_10y_monthly['Drawdown'] = treas_10y_monthly['Cumulative_Return_Plus_One'] / treas_10y_monthly['Rolling_Max'] - 1
treas_10y_monthly.drop(columns=['Cumulative_Return_Plus_One', 'Rolling_Max'], inplace=True)

# Convert to DataFrame
portfolio_monthly = strategy[['Date', 'strategy_return']].copy().set_index('Date')
portfolio_monthly = portfolio_monthly.rename(columns={'strategy_return': 'Portfolio_Monthly_Return'})

# Calculate cumulative returns and drawdown for the portfolio
portfolio_monthly['Portfolio_Cumulative_Return'] = (1 + portfolio_monthly['Portfolio_Monthly_Return']).cumprod() - 1
portfolio_monthly['Portfolio_Cumulative_Return_Plus_One'] = 1 + portfolio_monthly['Portfolio_Cumulative_Return']
portfolio_monthly['Portfolio_Rolling_Max'] = portfolio_monthly['Portfolio_Cumulative_Return_Plus_One'].cummax()
portfolio_monthly['Portfolio_Drawdown'] = portfolio_monthly['Portfolio_Cumulative_Return_Plus_One'] / portfolio_monthly['Portfolio_Rolling_Max'] - 1
portfolio_monthly.drop(columns=['Portfolio_Cumulative_Return_Plus_One', 'Portfolio_Rolling_Max'], inplace=True)

# Merge "spxt_monthly" and "treas_10y_monthly" into "portfolio_monthly" to compare cumulative returns
portfolio_monthly = portfolio_monthly.join(
    spxt_monthly['Monthly_Return'].rename('SPXT_Monthly_Return'),
    how='left'
).join(
    spxt_monthly['Cumulative_Return'].rename('SPXT_Cumulative_Return'),
    how='left'
).join(
    spxt_monthly['Drawdown'].rename('SPXT_Drawdown'),
    how='left'
).join(
    treas_10y_monthly['Monthly_Return'].rename('10Y_Monthly_Return'),
    how='left'
).join(
    treas_10y_monthly['Cumulative_Return'].rename('10Y_Cumulative_Return'),
    how='left'
).join(
    treas_10y_monthly['Drawdown'].rename('10Y_Drawdown'),
    how='left'
)

display(portfolio_monthly)

	Portfolio_Monthly_Return	Portfolio_Cumulative_Return	Portfolio_Drawdown	SPXT_Monthly_Return	SPXT_Cumulative_Return	SPXT_Drawdown	10Y_Monthly_Return	10Y_Cumulative_Return	10Y_Drawdown
Date
1990-01-31	-0.06713	-0.06713	0.00000	-0.06713	-0.06713	0.00000	-0.01987	-0.01987	0.00000
1990-02-28	0.01288	-0.05511	0.00000	0.01288	-0.05511	0.00000	-0.00023	-0.02010	-0.00023
1990-03-31	0.02650	-0.03007	0.00000	0.02650	-0.03007	0.00000	-0.00001	-0.02011	-0.00024
1990-04-30	-0.02497	-0.05429	-0.02497	-0.02497	-0.05429	-0.02497	-0.01411	-0.03394	-0.01436
1990-05-31	0.09752	0.03793	0.00000	0.09752	0.03793	0.00000	0.03148	-0.00353	0.00000
...	...	...	...	...	...	...	...	...	...
2025-08-31	0.02027	42.28506	0.00000	0.02027	36.70243	0.00000	0.01668	5.41258	-0.11316
2025-09-30	0.00675	42.57706	0.00000	0.03650	38.07857	0.00000	0.00675	5.45584	-0.10717
2025-10-31	0.00685	42.87548	0.00000	0.02341	38.99354	0.00000	0.00685	5.50005	-0.10106
2025-11-30	0.01020	43.32308	0.00000	0.00245	39.09156	0.00000	0.01020	5.56636	-0.09189
2025-12-31	-0.00662	43.02959	-0.00662	0.00061	39.11610	0.00000	-0.00662	5.52288	-0.09790

432 rows × 9 columns

Next, we’ll look at performance for the assets and portfolio.

Performance Statistics #

We can then plot the monthly returns:

plot_timeseries(
    price_df=portfolio_monthly,
    plot_start_date=start_date,
    plot_end_date=end_date,
    plot_columns=["Portfolio_Monthly_Return", "SPXT_Monthly_Return", "10Y_Monthly_Return"],
    title="Portfolio, SPXT, and 10Y Monthly Returns",
    x_label="Date",
    x_format="Year",
    x_tick_spacing=2,
    x_tick_rotation=45,
    y_label="Return",
    y_format="Decimal",
    y_format_decimal_places=2,
    y_tick_spacing=0.02,
    grid=True,
    legend=True,
    export_plot=False,
    plot_file_name=None,
)

png

And cumulative returns:

plot_timeseries(
    price_df=portfolio_monthly,
    plot_start_date=start_date,
    plot_end_date=end_date,
    plot_columns=["Portfolio_Cumulative_Return", "SPXT_Cumulative_Return", "10Y_Cumulative_Return"],
    title="Portfolio, SPXT, and 10Y Cumulative Returns",
    x_label="Date",
    x_format="Year",
    x_tick_spacing=2,
    x_tick_rotation=45,
    y_label="Cumulative Return",
    y_format="Decimal",
    y_format_decimal_places=0,
    y_tick_spacing=3,
    grid=True,
    legend=True,
    export_plot=False,
    plot_file_name=None,
)

png

And drawdowns:

plot_timeseries(
    price_df=portfolio_monthly,
    plot_start_date=start_date,
    plot_end_date=end_date,
    plot_columns=["Portfolio_Drawdown", "SPXT_Drawdown", "10Y_Drawdown"],
    title="Portfolio, SPXT, and 10Y Drawdowns",
    x_label="Date",
    x_format="Year",
    x_tick_spacing=2,
    x_tick_rotation=45,
    y_label="Drawdown",
    y_format="Decimal",
    y_format_decimal_places=2,
    y_tick_spacing=0.05,
    grid=True,
    legend=True,
    export_plot=False,
    plot_file_name=None,
)

png

Finally, we can run the summary statistics on the strategy portfolio, SPY, and TLT with the following code:

port_sum_stats = summary_stats(
    fund_list=["Portfolio", "SPXT", "10Y"],
    df=portfolio_monthly[["Portfolio_Monthly_Return"]],
    period="Monthly",
    use_calendar_days=False,
    excel_export=False,
    pickle_export=False,
    output_confirmation=False,
)

spy_sum_stats = summary_stats(
    fund_list=["Portfolio", "SPXT", "10Y"],
    df=portfolio_monthly[["SPXT_Monthly_Return"]],
    period="Monthly",
    use_calendar_days=False,
    excel_export=False,
    pickle_export=False,
    output_confirmation=False,
)

tlt_sum_stats = summary_stats(
    fund_list=["Portfolio", "SPXT", "10Y"],
    df=portfolio_monthly[["10Y_Monthly_Return"]],
    period="Monthly",
    use_calendar_days=False,
    excel_export=False,
    pickle_export=False,
    output_confirmation=False,
)

sum_stats = pd.concat([port_sum_stats, spy_sum_stats, tlt_sum_stats])
sum_stats

	Annual Mean Return (Arithmetic)	Annualized Volatility	Annualized Sharpe Ratio	CAGR (Geometric)	Monthly Max Return	Monthly Max Return (Date)	Monthly Min Return	Monthly Min Return (Date)	Max Drawdown	Peak	Trough	Recovery Date	Days to Recovery	MAR Ratio
Portfolio_Monthly_Return	0.11267	0.11897	0.94701	0.11086	0.10946	2020-11-30	-0.14458	1998-08-31	-0.23867	2021-12-31	2022-09-30	2023-12-31	457.00000	0.46449
SPXT_Monthly_Return	0.11392	0.14735	0.77311	0.10799	0.12819	2020-04-30	-0.16795	2008-10-31	-0.50949	2007-10-31	2009-02-28	2012-03-31	1127.00000	0.21196
10Y_Monthly_Return	0.05420	0.06331	0.85606	0.05347	0.08169	2008-11-30	-0.05558	2003-07-31	-0.22865	2020-07-31	2023-10-31	NaT	NaN	0.23386

Based on the above, our strategy portfolio outperforms both stocks and bonds, with lower drawdowns.

Conclusions #

This was a interesting exercise to evaluate the performance of different asset classes during Fed tightening and easing cycles. The results are not particularly surprising, but it is interesting to see the data and plots to confirm the economic intuition that stocks perform well during tightening cycles (economic strength) and bonds perform well during easing cycles (economic weakness). The results are certainly dependent on the specific time period or regime, and also on the assumption made for how to handle the periods of neutral policy (i.e. no change in FFR).

Future Investigation #

A couple of ideas sound intriguing for future investigation:

Does a commodity index (such as GSCI) exhibit differing behavior than gold?
How does leverage affect the returns that are observed for the strategy portfolio, stocks, and bonds?
Do other Fed tightening/loosening cycles exhibit the same behavior for returns?

References #

Code #

The Jupyter notebook with the functions and all other code is available here.
The HTML export of the jupyter notebook is available here.
The PDF export of the jupyter notebook is available here.