What held up when bond yields spiked? 20 episodes since 1963

The short version: in the 7 episodes since 1963 when the 10-year Treasury yield jumped 1 point or more in a year and the S&P 500 fell, 10-year Treasuries lost money all 7 times. One-month Treasury bills made money all 7 times. A 60/40 portfolio with Treasuries lost a median 6.6%, barely less than stocks' 7.4%; the same 60/40 holding T-bills lost 2.5%.

Part 2 of a series on rising yields and your portfolio. Part 1 found the 20 yield spikes and showed that stocks fell only when inflation was 3% or more. This part asks what, in the same 20 episodes, held up.

The usual answer to "what do I own when yields rise?" is a bond fund, because the 60/40 portfolio counts on bonds to rise when stocks fall. That works when stocks fall because the economy is slowing and yields drop. A yield spike is the opposite case: the bonds are what is falling.

In the table, "bad 7" are the spikes when the S&P 500 fell, "hot 14" those with inflation at 3% or more, "cool 6" those under 3%. Each count is how many finished their 12 months up.

The 7 bad years, one by one

Each row is the 12 months ending at the sharpest point of the climb, the same windows as in part 1. "60/40 + 10y" is 60% US stocks and 40% 10-year Treasuries; "60/40 + bills" holds T-bills instead. Inflation over these windows ran from 3.5% (1966) to 11.8% (1974).

The 60/40 with Treasuries beat stocks alone in 5 of the 7, but by little: in the 12 months to October 2022 it lost 17.3% against 17.5%. Swapping the 40% of Treasuries for T-bills did better than the Treasury version in all 7, and in all 20 episodes.

June 1978 is the one year stocks come out positive here while part 1 counts it as a loss: part 1 used the S&P 500 price index, this table counts dividends and the whole US market, and that is the difference.

Why Treasuries couldn't cushion this fall

A bond's price moves against its yield. Over a 1-point rise, the price a 10-year Treasury lost outweighed the year of coupon it paid in 19 of the 20 episodes. It finished positive once, in the 12 months to February 1989 (+1.3%). The median was −6.4%.

So the thing a 60/40 relies on, bonds up while stocks are down, needs yields to fall. In a spike, both sides lose together. That is what happened in 2022, and it is the same mechanism our study of long-term Treasuries at 5% measured on the 30-year bond. Today's level is on the 10-year Treasury yield page and the whole curve on Treasury yields.

Cash won the spike, not the inflation

T-bills are what the rising rate is paid on, so they rise with it: +14.8% in the 12 months to September 1981. They never lost money in the 20 episodes (the lowest was 0.0%, in 2013).

After inflation it is a different score. In the 7 bad episodes T-bills beat US inflation 4 times, with a median real return of +0.4%. In the 12 months to October 2022 they returned 0.8% while prices rose 7.8%. Nothing held its purchasing power every time: gold beat inflation in 2 of the 7, US stocks and every stock-bond mix in none.

Gold: the biggest wins and the biggest loss

Gold's median in the bad 7 is 0.0%, and that median hides the widest spread in the table: +59.0% in the 12 months to October 1974, −34.1% in the 12 months to September 1981. It rose in 3 of the 7. Adding 10% of it to a 60/40 (60/30/10) did not change the median loss in the bad years (−7.0%). Before 1968 the price was fixed at $35, so 1966 shows no move. Our oil-shock study found the same pattern over weeks: gold helps, then gives it back. The fund most people hold it through is GLD; GLD vs TLT puts it next to long Treasuries.

Which stocks held up

Ken French's 12 US industry portfolios, same 7 bad episodes:

Staples came closest to holding flat; the usual fund for them is XLP. Energy is the trap: it led across the 14 high-inflation spikes, with a median of +13.3%, and still lost 7.8% in the ones where the market fell.

And after the spike

The loss on bonds was mostly deferred income. In the 12 months after the sharpest month of each spike, 10-year Treasuries gained 18 times out of 20, a median 8.3%, and beat T-bills 14 times. The two exceptions were the 12 months to March 2022 (−3.6%) and to October 2023 (−2.5%).

The catch is timing. The sharpest month is only known afterwards; in 2021 the next year brought a bigger spike, not a rebound. Nobody we tested could call the turn in yields: in our walk-forward test no model beat "no change".

What it means for a portfolio today

On September 24, 2026 the 10-year yield was 5.18%, up 1.02 points on the year, with US inflation at 3.4% in August: the high-inflation side of the table, where the 7 bad episodes sit. The S&P 500 was still up 16.7% on the year. This is not a forecast, and it is not investment advice.

What the 20 episodes say about a portfolio is narrower: in a yield spike, the part meant to cushion a stock fall is only as good as its maturity. Long bonds (TLT) and a total bond fund (BND vs TLT) carried the loss; T-bills didn't. Here is the 2022 rate shock, from the January 2022 high to the recovery in early 2024, replayed on what you hold today:

The crash test runs the same holdings through the COVID crash and the 2024 yen unwind, and the portfolio X-ray shows how much of them sits in long bonds and rate-sensitive funds. Every study in the series is on the research page.

How we measured

Monthly returns from February 1962 to August 2026. US stocks are the CRSP value-weighted US market with dividends (Ken French's Mkt-RF plus RF); T-bills are the 1-month Treasury bill (Ken French RF). The 10-year Treasury is built from FRED's DGS10: each month a par 10-year bond is bought at last month's close and repriced at this month's, one month shorter, plus a month of coupon. On calendar years it returns −7.1% for 1994 and −16.4% for 2022. Gold is the World Bank Pink Sheet monthly average price, so it smooths month-end swings. Industries are Ken French's 12 value-weighted portfolios. Portfolios are rebalanced every month, with no fees or taxes. Real returns deduct CPI (FRED CPIAUCSL) over the same 12 months.

The 20 episodes are part 1's: runs of months when the 10-year yield was up 1 point or more on the year, each measured over the 12 months ending at its sharpest month. "Bad" means the S&P 500 price index fell over those months. What would change the result: the episodes are chosen in hindsight, so a live strategy would enter late and leave late; a shorter bond than 10 years would lose less in each spike and earn less after it.

Reproduce it

The headline counts, T-bills up in 7 of the 7 bad episodes and the 10-year Treasury in 0, in about 60 lines of Python on public data only: FRED for the 10-year yield and CPI, Yahoo Finance for the S&P 500, Ken French for stocks and T-bills, month-ends through August 2026. Gold is left out, because the Pink Sheet is an Excel file. We ran it on September 29, 2026; it finds part 1's 20 episodes, then prints the counts.

# Reproduces https://portfolio-terminal.com/blog/portfolio-when-bond-yields-rise
# pip install pandas yfinance
# Month-ends through August 2026. Gold is left out: its source is an Excel file.
import pandas as pd
import yfinance as yf

END = "2026-08"
FRED = "https://fred.stlouisfed.org/graph/fredgraph.csv?id={}&coed=2026-09-25"
FRENCH = "https://mba.tuck.dartmouth.edu/pages/faculty/ken.french/ftp/F-F_Research_Data_Factors_CSV.zip"


def month_end(s):
    return s.groupby(s.index.to_period("M")).last()


def fred(series):
    s = pd.read_csv(FRED.format(series), index_col=0, parse_dates=True, na_values=".").iloc[:, 0]
    return month_end(s.dropna())


y10, cpi = fred("DGS10"), fred("CPIAUCSL")  # 10-year yield in %, CPI
spx = yf.Ticker("^GSPC").history(start="1960-01-01", end="2026-09-01", auto_adjust=False)["Close"]
spx = month_end(spx.tz_localize(None))

ff = pd.read_csv(FRENCH, skiprows=3, index_col=0)  # monthly block, then annual
ff = ff[ff.index.astype(str).str.strip().str.fullmatch(r"\d{6}")].astype(float) / 100
ff.index = pd.PeriodIndex(ff.index.str.strip(), freq="M")

# Par 10-year bond bought at last month's yield, repriced a month later, plus a month of coupon.
c, y, t = y10.shift(1) / 100, y10 / 100, 10 - 1 / 12
v = (1 + y / 2) ** (-2 * t)
bond = c / y * (1 - v) + v - 1 + c / 12

r = pd.DataFrame({"stocks": ff["Mkt-RF"] + ff["RF"], "T-bills": ff["RF"], "10y Treasury": bond}).loc["1962-02":END]
r["60/40 Treasuries"] = 0.6 * r["stocks"] + 0.4 * r["10y Treasury"]
r["60/40 T-bills"] = 0.6 * r["stocks"] + 0.4 * r["T-bills"]

# Part 1's episodes: runs of months with the yield up 1 point or more on the year
# (gaps under a quarter joined), each taken at its sharpest month.
m = pd.DataFrame({"dy": y10 - y10.shift(12), "spx": spx / spx.shift(12) - 1, "cpi": cpi / cpi.shift(12) - 1}).loc[:END].dropna()
hits = m[m["dy"] >= 1]
run = (pd.Series([p.ordinal for p in hits.index]).diff() > 3).cumsum().values
peaks = hits.groupby(run)["dy"].idxmax()


def pct(p, first, last):
    return ((1 + r.loc[p + first:p + last]).prod() - 1).mul(100).round(1)


ep = pd.DataFrame({p: pct(p, -11, 0) for p in peaks}).T
after = pd.DataFrame({p: pct(p, 1, 12) for p in peaks}).T
bad = m.loc[peaks, "spx"].values < 0
print(f"{len(ep)} episodes, {(m.loc[peaks, 'cpi'] >= 0.03).sum()} with inflation at 3%+, stocks fell in {bad.sum()}")
print("During the 12 months of the spike:")
for col in r.columns:
    print(f"  {col:17} up in {(ep[col][bad] > 0).sum()} of {bad.sum()} bad episodes (median {ep[col][bad].median():.1f}%),"
          f" up in {(ep[col] > 0).sum()} of {len(ep)}")
print(f"  T-bills beat the 10y Treasury in {(ep['T-bills'] > ep['10y Treasury']).sum()} of {len(ep)}")
print(f"In the 12 months after: 10y Treasury up in {(after['10y Treasury'] > 0).sum()} of {len(after)}"
      f" (median {after['10y Treasury'].median():.1f}%), beat T-bills in {(after['10y Treasury'] > after['T-bills']).sum()}")

20 episodes, 14 with inflation at 3%+, stocks fell in 7
During the 12 months of the spike:
  stocks            up in 1 of 7 bad episodes (median -7.4%), up in 13 of 20
  T-bills           up in 7 of 7 bad episodes (median 6.1%), up in 19 of 20
  10y Treasury      up in 0 of 7 bad episodes (median -5.5%), up in 1 of 20
  60/40 Treasuries  up in 1 of 7 bad episodes (median -6.6%), up in 12 of 20
  60/40 T-bills     up in 2 of 7 bad episodes (median -2.5%), up in 15 of 20
  T-bills beat the 10y Treasury in 20 of 20
In the 12 months after: 10y Treasury up in 18 of 20 (median 8.3%), beat T-bills in 14

Sources: FRED, 10-year Treasury constant maturity rate (DGS10) · FRED, consumer price index (CPIAUCSL) · Kenneth R. French Data Library · World Bank commodity prices (Pink Sheet) · S&P 500 index history, Yahoo Finance