SH

The IPO fade

Large IPOs tend to give back their first-day high. I measured how much across 63 of them, tested the catalyst everyone cites, and wrote down three predictions before SPCX’s first trade.

When
June 2026, before the listing
Sample
63 large-cap IPOs since 2010
Stack
Python: pandas, NumPy, SciPy, yfinance
Code
SukhmanH/SpaceX_Thesis
0%−50%−29.2%
All 63 paths, market-adjusted against QQQ. The median in white.

The IPO fade, measured

1531326863traditional IPOs above $5B, 2010 to nowwith clean price historylisted above $10Blarge-caps with a full 180-day history+50%0%−50%−100%+0+60+90+180trading days from the day-1 high, vs QQQP1−29.2%+1.58%2005-2009t 1.57+0.11%2010-2014t 0.21−0.10%2015-2019t −0.30−0.07%2020-presentt −0.12Rebalance-day abnormal return, S&P 500 additionsn = 18, 51, 101, 100 by era, vs SPYCommitted June 10, 2026, before the first tradeP1 lags QQQ by 25 to 40 points by +180P2 inclusion pop under 1 pointP3 flat four weeks, lower by +90Wider replication failed my pre-stated criteria. Null reported.

  1. The sample came first: 153 traditional IPOs above $5B since 2010, cut to 63 large-caps with a full 180-day history.
  2. Each line is one IPO, measured from its day-1 high against the Nasdaq-100.
  3. The median locks: −29.2% by trading day 180. 50 of 63 faded.
  4. Index inclusion, the usual reason to buy early, has added nothing on the rebalance day since 2010.
  5. Three predictions, committed before the first trade. A wider replication failed my pre-stated criteria, and I reported the null.

The question

SpaceX was set to list on Nasdaq as SPCX in one of the largest IPOs on record. The report makes no forecast about the company. It asks what usually happens to very large IPOs after the first day, and whether anything in this listing suggests it will be different.

The sample

153 traditional underwritten IPOs above $5B since 2010, built automatically from Jay Ritter’s IPO data, the Nasdaq calendar, and SEC EDGAR, with every exclusion logged. Direct listings, SPACs, and spinoffs are out because they are different liquidity events. 68 listed above $10B, and 63 of those have a full 180-day history.

The measure

Every number is market-adjusted: the return from the day-1 intraday high, minus QQQ over the same window. I report medians, because a few runaway winners skew the means.

What it found

The large-cap median is −29.2% by trading day 180 (t = −6.51), and 50 of 63 faded. It isn’t immediate: prices sit between −4.5% and −6.3% through the first month, then fall to −14.2% by day 60 and −24.3% by day 90.

Every robustness check stays deeply negative: other peak references, SPY and IWM as benchmarks, +120 and +250 days, high- and low-VIX listings. Every specification is significant at p < 0.001.

The catalyst

Across S&P 500 additions since 2005, the rebalance-day abnormal return was +1.58%in 2005 to 2009 (not significant) and statistically zero in every era since. What’s left is drift before the effective date. That’s front-running, not a return you can get by buying at inclusion.

What it can’t show

It’s a base rate, not a mechanism. 21 of 153 names had no complete price history, and they skew toward failures, so dropping them makes the fade look smaller, not larger. The t-statistics assume independent events; clustered or bootstrap errors would widen the intervals.

Committed first

Three predictions went in before the first trade, including one written to break the thesis if it was wrong. A replication on a wider sample failed my pre-stated criteria, and I reported the null.

Read the report or rebuild every number from the code.