S&P 500 members that were demoted, removed while their shares went on trading, trailed SPY by a median 57.6 percentage points of total return over their last 24 months in the index (285 removals, 1995-2024). Members whose shares stopped trading at removal, mostly acquired companies, finished a median 10.6 points ahead. Only 16 of the 702 removals measured belong to companies in today’s index, so a backtest built from the current list holds almost none of these paths.
How far do stocks fall before the S&P 500 drops them?
A member leaves the S&P 500 either because the company ceases to exist as a listed stock, through an acquisition or a failure, or because it no longer fits the index, usually after its market value has shrunk. Until the removal date every index fund holds it, so whatever it lost on the way out is part of the index return those funds delivered.
The replacement-pair study of S&P 500 deletions covers what removed stocks did after they left. This post looks backwards from the removal date: how large the shortfall is over the two years before it, when it builds up, and whether demoted and acquired companies differ. The measure is the market-adjusted return, a stock’s cumulative total return (price change plus dividends) minus SPY’s over the same trading days, in percentage points.
Following 775 removals back two years by entity id
- Take every removal from the dated log,
xfl.index_events("sp500"), effective 1995 to 2024: 775 removals. The same history appears in the historical S&P 500 constituents lookup; the call is described in the index events reference. - Follow each member by
entity_id, not ticker. AAL was Alexander & Alexander Services when it left the index in 1997 and American Airlines when it left in 2024. - Keep members that had been in the index for the full 504 trading days before removal; 73 recent additions and spin-offs drop out, leaving 702.
- Set month 0 at the last close on or before the removal date, compound the stock’s and SPY’s daily total returns from 504 trading days earlier, and read the gap every 21 trading days.
- Classify each removal from its own price series: at most 5 trading days of prices after the removal (stopped trading, 391), at least 63 (demoted, 285), or in between (26, shown separately). The still-trading group also holds removals made under index rules, such as the seven non-US companies dropped in July 2002.
Code
import pandas as pd
import xfinlink as xfl
xfl.set_api_key("YOUR_API_KEY") # free at https://xfinlink.com/signup
DAYS, STEP = 504, 21 # 24 months of trading days; one month = 21 trading days
ev = pd.concat([xfl.index_events("sp500", end="2025-06-30", offset=o)
for o in range(0, 4000, 1000)], ignore_index=True)
ev = ev.dropna(subset=["entity_id"]).drop_duplicates()
ev["eff"] = pd.to_datetime(ev["effective_date"])
ev["entity_id"] = ev["entity_id"].astype(int)
rem = ev[(ev["event_type"] == "removed") & ev["eff"].between("1995-01-01", "2024-12-31")]
adds = ev[ev["event_type"] == "added"]
spy = xfl.prices("SPY", start="1993-01-01", end="2025-06-30", fields=["return_daily"])
spy = spy.set_index("date")["return_daily"].fillna(0.0)
cal = spy.index
# Daily total returns for each year's removed members, followed by entity_id, not ticker
px = {}
for y, g in rem.groupby(rem["eff"].dt.year):
start = (pd.Timestamp(f"{y}-01-01") - pd.DateOffset(months=26)).date().isoformat()
end = min(pd.Timestamp(f"{y + 1}-04-30"), pd.Timestamp("2025-06-30")).date().isoformat()
p = xfl.prices(entity_id=sorted(set(g["entity_id"])), start=start, end=end,
fields=["return_daily"], max_rows=250_000)
for eid, s in p.dropna(subset=["return_daily"]).groupby("entity_id"):
px[(y, eid)] = s.drop_duplicates("date").set_index("date")["return_daily"]
rows = []
for _, r in rem.iterrows():
s = px.get((r["eff"].year, r["entity_id"]))
joined = adds.loc[(adds["entity_id"] == r["entity_id"]) & (adds["eff"] <= r["eff"]), "eff"].max()
if s is None or (pd.notna(joined) and joined > cal[cal.searchsorted(r["eff"], "right") - 1 - DAYS]):
rows.append({"group": "member under 24 months"})
continue
k0 = cal.get_loc(s.index[s.index <= r["eff"]].max()) # last close as a member = month 0
window = cal[k0 - DAYS + 1:k0 + 1]
after = int((s.index > r["eff"]).sum()) # trading days after the removal
group = ("stopped trading" if after <= 5 else
"still trading" if after >= 63 else "stopped within 3 months")
stock = (1 + s.reindex(window).fillna(0.0)).cumprod() - 1
mkt = (1 + spy.loc[window]).cumprod() - 1
adj = (stock - mkt).iloc[STEP - 1::STEP].values # months -23 ... 0
rows.append({"group": group, "eff": r["eff"], "entity_id": r["entity_id"],
"raw": stock.iloc[-1], "spy": mkt.iloc[-1],
**{m: v for m, v in zip(range(-23, 1), adj)}, -24: 0.0})
res = pd.DataFrame(rows)
kept = res[res["group"] != "member under 24 months"]
print(res["group"].value_counts())
print((kept.groupby("group")[[-18, -12, -6, -3, -1, 0]].median() * 100).round(1))Full script with formatting and visualisation: stocks-removed-from-sp500-before-removal-python.py
Output
S&P 500 removals 1995-2024: 775
member for under 24 months before removal (left out): 73
stopped trading within 5 trading days of removal: 391
still trading 63+ trading days after removal: 285
stopped trading 6-62 trading days after removal: 26
of the 702 measured, members of today's S&P 500: 16
Return minus SPY's since month -24 (total return, percentage points)
month -24 -18 -12 -9 -6 -3 -1 0
still trading 25th pct 0.0 -20.9 -40.9 -49.0 -58.5 -68.7 -79.3 -84.5
still trading median 0.0 -7.1 -22.8 -31.6 -38.7 -47.2 -55.4 -57.6
still trading 75th pct 0.0 3.3 -7.1 -9.7 -12.6 -20.9 -27.0 -27.7
stopped trading 25th pct 0.0 -15.0 -23.4 -26.3 -26.7 -22.5 -24.3 -20.8
stopped trading median 0.0 -3.1 -6.2 -5.2 -2.0 5.5 8.4 10.6
stopped trading 75th pct 0.0 9.2 10.4 14.8 23.6 35.3 39.8 40.9
stopped within 25th pct 0.0 -21.7 -45.7 -51.3 -62.0 -68.0 -55.7 -65.6
stopped within median 0.0 -8.1 -13.9 -28.6 -46.3 -39.2 -37.3 -33.4
stopped within 75th pct 0.0 6.4 12.9 -3.1 7.6 2.8 4.6 16.6
still trading (n=285): own return median -33.7%, SPY median +24.0%; behind SPY at month -12 79%, at month 0 91%; mean gap at month 0 -52.9 pts
stopped trading (n=391): own return median +39.7%, SPY median +28.0%; behind SPY at month -12 60%, at month 0 42%; mean gap at month 0 +15.2 pts
stopped within 3 months (n=26): own return median +3.5%, SPY median +30.6%; behind SPY at month -12 69%, at month 0 69%; mean gap at month 0 -24.0 pts
stopped trading with an own 24-month loss beyond 80% (failures and rescues): 15 of 391
still trading: median shortfall reached by month -12 40%, by month -6 67%; median relative wealth (1+stock)/(1+SPY)-1 -46.9%
By period of removal: median gap to SPY at month -12 and month 0 (points)
1995-2004 still trading n= 97 month -12 -25.5 month 0 -62.1 behind SPY 84%
1995-2004 stopped trading n=162 month -12 -6.1 month 0 +6.5 behind SPY 46%
2005-2014 still trading n= 86 month -12 -19.0 month 0 -45.8 behind SPY 90%
2005-2014 stopped trading n=131 month -12 -6.2 month 0 +16.0 behind SPY 40%
2015-2024 still trading n=102 month -12 -23.9 month 0 -65.1 behind SPY 98%
2015-2024 stopped trading n= 98 month -12 -6.1 month 0 +10.5 behind SPY 40%
Still trading, five widest gaps
1999-06-08 HPH HARNISCHFEGER INDUSTRIES INC own -97.3% SPY +57.3% gap -154.6 pts
1998-01-16 ECO ECHO BAY MINES LTD own -86.9% SPY +62.9% gap -149.8 pts
1998-12-31 Z VENATOR GROUP INC own -69.4% SPY +71.3% gap -140.7 pts
2000-09-29 OWC OWENS CORNING own -91.3% SPY +48.5% gap -139.8 pts
1999-09-30 FTL FRUIT OF THE LOOM INC own -87.8% SPY +39.9% gap -127.7 pts
Still trading, three that beat SPY by most
1995-10-20 FRM FIRST MISSISSIPPI CORP own +296.6% SPY +32.4% gap +264.2 pts
2000-07-27 COMS 3COM CORP own +155.4% SPY +32.1% gap +123.3 pts
2001-08-03 CS CABLETRON SYSTEMS INC own +73.8% SPY -5.5% gap +79.3 pts
Named checks (AAL appears twice: two different companies)
2020-04-03 M MACYS INC own -81.2% SPY -2.0% gap -79.2 pts
2020-09-18 KSS KOHLS CORP own -67.5% SPY +18.7% gap -86.2 pts
2013-11-29 JCP PENNEY J C CO INC own -65.7% SPY +57.8% gap -123.6 pts
2020-04-03 RTN RAYTHEON CO own -43.2% SPY +0.5% gap -43.7 pts
2017-08-28 WFM WHOLE FOODS MARKET INC own +34.3% SPY +31.1% gap +3.1 pts
2008-09-16 LEH LEHMAN BROTHERS HOLDINGS INC own -99.6% SPY -4.1% gap -95.5 pts
1997-01-14 AAL ALEXANDER & ALEXANDER SVCS INC own -14.4% SPY +70.8% gap -85.2 pts
2024-09-20 AAL AMERICAN AIRLINES GROUP INC own -22.5% SPY +50.6% gap -73.1 pts
M 2020: still trading
KSS 2020: still trading
JCP 2013: still trading
RTN 2020: stopped trading
WFM 2017: stopped trading
LEH 2008: stopped trading
AAL 1997: stopped within 3 months
AAL 2024: still tradingDemoted members lost most of the ground in their last year
The demoted group’s median path steepens as removal nears: 7.1 points behind SPY at month -18, 22.8 at month -12, 38.7 at month -6 and 57.6 at month 0. About 40% of the final median shortfall had accrued a year before removal and 67% six months before. In plain returns, the median demoted member lost 33.7% while the median SPY return over the same windows was +24.0%; even the 75th percentile ends 27.7 points behind, and 91% trailed.
Several of the widest gaps belong to companies near bankruptcy or in deep restructuring: Harnischfeger (-97.3% before its 1999 removal), Owens Corning, Fruit of the Loom, and J.C. Penney, down 65.7% while SPY gained 57.8%. Macy’s and Kohl’s, both moved out in 2020, trailed by 79.2 and 86.2 points. The few that beat SPY include 3Com, removed in 2000 as it spun off Palm, and Cabletron, removed in 2001 as it split into separate companies.
Companies that stopped trading at removal followed another path. Their median hovered a few points below SPY until month -6 and then climbed to +10.6, consistent with deal premiums entering prices. Failures and rescue takeovers sit in the same group: 15 of 391 lost more than 80%, Lehman Brothers, Washington Mutual, SVB Financial and First Republic among them.
The pattern has not faded. Median demoted shortfalls were 62.1 points for 1995-2004 removals, 45.8 for 2005-2014 and 65.1 for 2015-2024, when 98% of the 102 demoted members trailed SPY.
What the pre-removal slide means for index holders and backtests
These losses happened inside the index, while the names were members. Cap weighting softens them as a falling stock’s weight shrinks; an equal-weight portfolio does not. A universe drawn from today’s constituents drops 686 of the 702 removals and nearly every demoted member’s slide, so its backtest never holds the index’s worst members on their way out. The guide to S&P 500 additions and deletions explains how to rebuild each date’s membership from the dated log.
A year before removal, 79% of eventual demotions already trailed SPY, so a trailing-return screen would have flagged most of them. Its false-alarm rate, the members that trailed as much and stayed, is not measured here and decides whether the screen works. Measure it on each date’s roster before treating weak relative returns as a deletion forecast; acquired companies need a separate test, since their final months run the other way.
Built with xfinlink — free financial data API for Python.