11 — Scenario Analysis¶

Learning objectives¶

Apply a named, hypothetical scenario to a multi-sleeve portfolio by computing each shock's P&L contribution separately, then combining them — and see shocks that partially offset each other in the same scenario.

Free learning pack¶

1. reference/concepts/stress_testing.md
2. CFA active portfolio management page
3. PIMCO fixed-income education hub

Do not search for more material until these are insufficient.

Distinction¶

Scenario analysis asks: what would happen if these shocks occurred? — as opposed to VaR, which asks "what's the loss at a given statistical confidence?" (see notebooks/active/19_var_and_expected_shortfall.ipynb).

This repo keeps shocks explicit and separate rather than summing them in one aggregator function — each leg is auditable on its own. Reuse key_rate_return_approximation (curve), spread_pnl (credit), and simple multiplication (equity) per leg.

PREDICT¶

A risk-off scenario shocks Treasury yields down (curve rallies) and widens credit spreads and sells off equities, all at once. For a portfolio holding both rate-sensitive bonds and credit risk: does the curve P&L help or hurt in this scenario? Could it be large enough to meaningfully offset the credit and equity losses, or is that wishful thinking?

In [ ]:

from pm.scenarios import Scenario

risk_off = Scenario(
    name="Risk Off",
    curve_bp={"2Y": -25, "5Y": -40, "10Y": -50, "30Y": -55},
    ig_spread_bp=75,
    hy_spread_bp=200,
    equity_return=-0.20,
)

# A hypothetical multi-sleeve portfolio:
rates_sleeve_mv = 12_000_000
rates_sleeve_krds = [1.5, 2.0, 3.0, 1.0]  # 2Y, 5Y, 10Y, 30Y, portfolio-level

ig_mv, ig_spread_duration = 8_000_000, 4.8
hy_mv, hy_spread_duration = 3_000_000, 3.2

equity_mv = 5_000_000

from pm.scenarios import Scenario risk_off = Scenario( name="Risk Off", curve_bp={"2Y": -25, "5Y": -40, "10Y": -50, "30Y": -55}, ig_spread_bp=75, hy_spread_bp=200, equity_return=-0.20, ) # A hypothetical multi-sleeve portfolio: rates_sleeve_mv = 12_000_000 rates_sleeve_krds = [1.5, 2.0, 3.0, 1.0] # 2Y, 5Y, 10Y, 30Y, portfolio-level ig_mv, ig_spread_duration = 8_000_000, 4.8 hy_mv, hy_spread_duration = 3_000_000, 3.2 equity_mv = 5_000_000

MANUAL FIRST¶

Calculate each contribution separately before aggregating:

1. Curve: key_rate_return_approximation(krds, shocks_in_decimal) gives a return; multiply by rates_sleeve_mv for dollars. Remember to convert curve_bp values from basis points to decimal (divide by 10,000), and match the order of rates_sleeve_krds to ["2Y","5Y","10Y","30Y"].
2. Credit: spread_pnl(market_value, spread_duration, spread_change_bp) for the IG and HY legs separately.
3. Equity: just equity_mv * risk_off.equity_return.

In [ ]:

from pm.fixed_income.credit import spread_pnl
from pm.fixed_income.curve import key_rate_return_approximation

tenors = ["2Y", "5Y", "10Y", "30Y"]

# MANUAL FIRST:
curve_shocks_decimal = None   # [risk_off.curve_bp[t] / 10000 for t in tenors]
curve_pnl = None              # key_rate_return_approximation(...) * rates_sleeve_mv
ig_pnl = None
hy_pnl = None
equity_pnl = None
total_pnl = None

print("curve:", curve_pnl, " IG:", ig_pnl, " HY:", hy_pnl, " equity:", equity_pnl)
print("total risk-off P&L:", total_pnl)

# CHECK (uncomment after your attempt):
# assert curve_pnl > 0, "yields falling should help a duration-long rates sleeve"
# assert ig_pnl < 0 and hy_pnl < 0 and equity_pnl < 0
# import numpy as np
# assert np.isclose(total_pnl, -1_093_000, atol=1000)
# assert abs(curve_pnl) < abs(ig_pnl) + abs(hy_pnl) + abs(equity_pnl), (
#     "the curve rally helps, but check whether it's actually enough to offset everything else"
# )

from pm.fixed_income.credit import spread_pnl from pm.fixed_income.curve import key_rate_return_approximation tenors = ["2Y", "5Y", "10Y", "30Y"] # MANUAL FIRST: curve_shocks_decimal = None # [risk_off.curve_bp[t] / 10000 for t in tenors] curve_pnl = None # key_rate_return_approximation(...) * rates_sleeve_mv ig_pnl = None hy_pnl = None equity_pnl = None total_pnl = None print("curve:", curve_pnl, " IG:", ig_pnl, " HY:", hy_pnl, " equity:", equity_pnl) print("total risk-off P&L:", total_pnl) # CHECK (uncomment after your attempt): # assert curve_pnl > 0, "yields falling should help a duration-long rates sleeve" # assert ig_pnl < 0 and hy_pnl < 0 and equity_pnl < 0 # import numpy as np # assert np.isclose(total_pnl, -1_093_000, atol=1000) # assert abs(curve_pnl) < abs(ig_pnl) + abs(hy_pnl) + abs(equity_pnl), ( # "the curve rally helps, but check whether it's actually enough to offset everything else" # )

ORAL CHECK¶

Why might falling Treasury yields and widening credit spreads partially offset one another in a credit portfolio? Using your own numbers above: did the offset in this particular scenario end up being large, negligible, or somewhere in between — and what would need to be true about the portfolio's mix for the offset to fully cancel the loss?

Try /tutor stress testing for an adaptive walkthrough.