When IPO volumes surge and first-day performance is an extreme pop, it generally means investors are euphoric and signals that underwriters can price just about anything. This classic late-cycle signal points to weak forward returns. 1999 (484) and 2000 (382) IPOs priced with average first-day pops of 70% and 56%, respectively. This was just before the S&P 500 fell 10% in 2000 and 13% in 2001. 2021 rhymes: 315 IPOs (plus separate wave of SPACs), a 32% average pop, then a 19% market decline in 2022.
The regressions above test this, and I can conclude that the relationship is directionally consistent with the "froth" story given the number of IPOs and the first-day pop both tend to carry a negative slope against next year's return. On average, hot IPO markets are followed by softer years. However, it's statistically weak over the 30-year sample (n≈30, R² in single digits, p-values missing the conventional 0.05 threshold). The two dot-com and SPAC-boom years do a lot of the work. The jackknife row shows how much the fit moves when the single most influential year is dropped.
IPO volume should be treated as one qualitative input into a sentiment indicator (with others like options positioning, margin debt, and valuation multiples) as opposed to a standalone quantitative timing signal. The first-day pop has a cleaner behavioral story than the raw count, but both are mechanically suppressed in weak years simply because fewer companies choose to go public at all.
| 1995–1998 | Steady bull-market IPO issuance (310–690 deals/yr) as the 1990s expansion matured, still well short of dot-com extremes. |
| 1999–2000 | Dot-com peak: 484 and 382 IPOs with average first-day pops of 70% and 56%. The most extreme underpricing on record. Followed by the 2000–2002 bear market. |
| 2001–2003 | IPO count collapses to 60–80/year post-bear and post-9/11. |
| 2004–2007 | Recovery cycle (160–180 IPOs/yr, pops back to a more normal 10–14%) tracking the mid-2000s credit-fueled bull market. |
| 2008–2009 | GFC. 21 and 42 IPOs priced. IPO window effectively closed. |
| 2013–2014 | Post-GFC bull market IPO revival. Led by tech and biotech (163 and 222 issues). |
| 2020–2021 | Pandemic-liquidity boom. Record proceeds ($62B and $120B). 2021 with a 32% average pop, SPAC wave (not counted above). 2022's bear market followed. |
| 2022 | IPO window effectively closed again (39 deals). Fed hiked into a bear market. Lowest gross proceeds since GFC. |
| 2023–2025 | Gradual reopening, 2025's 94 deals and 28% average pop (led by high-profile tech listings) suggest renewed risk appetite. |
The chart shows three series pulled directly from Jay Ritter's continuously updated IPO database, a standard academic and industry source for U.S. IPO statistics going back to 1960. The headline series includes bank/thrift conversions and excludes ADRs, SPACs, unit offers, closed-end funds, REITs, and very small or best-efforts deals, which helps to use a conservative definition for year-over-year consistency instead of the broadest possible count. *S&P 500 returns are price return only (no dividends)*
The two scatter plots test the "IPO froth as a contrarian indicator" hypothesis directly: does this year's IPO activity (count, or average pop) predict next year's S&P 500 return? A negative, statistically significant slope would support the classic narrative that heavy IPO issuance marks late-cycle excess. As the methodology tables show, the sign is usually right but the significance is marginal. Consistent with IPO volume being a real but noisy and non-mechanical signal, unlike the implied ERP series on this site's Implied Equity Risk Premium page, which by construction can't avoid some contemporaneous correlation with market pricing.
Limitations: a 31-year annual sample is small, and two outlier years (1999–2000, 2021) drive much of the relationship. The jackknife row in each methodology table shows by exactly how much. Less focused on is that IPO count is as much a supply-side measure as it is a demand-side measure. Companies choose whether to go public based on the same market conditions that are being tested as the predictor.