In a traditional IPO, an investment bank builds a book of buyers, sets an offer price overnight, allocates shares to chosen institutions, and often supports the stock afterward with a greenshoe. The company gets primary capital. Existing holders usually wait through a lock-up before they can sell.
A direct listing cuts that stack. The company registers shares, lists them, and lets the exchange opening auction find a price. There is no offer price, no allocation book, and no bank standing ready to stabilize the stock. Existing shareholders can sell from day one. The company typically raises nothing.
That legal act is the same for everybody. The economic decision underneath it is not.
When a large, well-capitalised company skips the bank, it is declining a service it does not need. It already has liquidity, a shareholder base, and a private-market price. The underwriter's core job (finding buyers and clearing an inventory of new shares) is redundant. That is a choice made from strength.
When a micro-cap skips the bank, it is usually because no bank would take the deal. The underwriter is not being declined; the underwriter is unavailable. Same legal structure, opposite meaning.
How the structure got here
Direct listings did not appear overnight. The timeline that matters:
- February 2018. The SEC approves NYSE rules for secondary direct listings: existing shareholders sell registered shares; the company raises no primary capital. An independent valuation of at least $250 million in publicly held shares, plus a financial advisor for the opening auction, replaces the old requirement of prior private-market trading.
- April 2018 / June 2019. Spotify and Slack list that way. Both already had deep private markets and famous brands. Financial advisors consult the designated market maker on the open. They do not book-build, allocate, or run a greenshoe.
- August–December 2020. NYSE wins SEC approval for primary direct floor listings: the company can sell new shares into the opening auction, alone or alongside selling shareholders. That closes the biggest historical objection ("you cannot raise money this way").
- May 2021. Nasdaq gets the parallel primary-capital rule.
- June 2023. In Slack Technologies v. Pirani, a unanimous Supreme Court holds that a Section 11 plaintiff must trace the shares they bought to the allegedly defective registration statement. In a direct listing, registered and unregistered stock trade side by side from the first print, so a buyer generally cannot prove which they got. The practical effect runs one way: it removed a liability worry for issuers and, in the same stroke, most of the disclosure-based legal recourse for anyone buying in the open market.
So there are two legal kinds: liquidity-only (secondary) and capital-raising (primary). Almost every direct listing we actually observe is still the first kind. Primary direct listings remain rare. The split that shows up in returns is not primary versus secondary. It is large versus small.
What you lose when there is no underwriter
Three operational differences matter more than the prospectus wording.
Price discovery sits in the opening auction. Before the open, the exchange (with a financial advisor in the loop) builds a book of buy and sell interest and prints a single opening price. There is no lead underwriter setting an offer overnight or allocating to long-only institutions it likes. Demand is whoever shows up.
No greenshoe. In a traditional IPO the underwriter can over-allot and buy back shares to cushion a weak open. Direct listings have no over-allotment option and no stabilization bid. Opening volatility is part of the design.
No standard 180-day lock-up. In a normal IPO, most pre-IPO holders are contractually barred from selling for about six months. In a direct listing that restraint is usually gone. Registered resale shares and Rule 144-eligible unregistered shares can both hit the tape immediately. That is why share-tracing became a Supreme Court issue, and why early float can be larger (and messier) than the IPO playbook assumes.
There is also a quiet but important detail: the reference price. Before the auction, the advisor and the exchange publish a reference that is not a transaction. Nothing trades there. For Spotify-scale names the open usually lands near it. For thin micro-caps, opens have printed multiples away from the reference on very little volume. Treating that number like an IPO offer price is a category error.
For a Coinbase-scale name, those absences are a cost of freedom: you skip fees and dilution control in exchange for a noisier open. For a thin micro-cap, they are the whole risk: a non-transaction reference, a thin book, no stabilizer, and no lock-up.
Why the pooled number lies
Direct listings are rare (a low single-digit share of listings), so most analyses dump them in one bucket. Pooled, the effect looks moderately bad and moderately significant. You would reasonably file it under "illiquid small deals underperform" and move on.
Split by size and that average falls apart. One cohort looks like the broad IPO market. The other does not merely lag; its three-month win rate sits near zero. The useful question is not "do direct listings work?" It is "which companies are using this structure, and why?"
What the mechanism implies
If size is really a proxy for why the underwriter is absent, the useful signal is not "direct listing." It is the absence of a party willing to underwrite. That reframes the label as a screening variable rather than a structural one. The prediction is specific: the drag should be strongest where an underwriter would have been cheapest to hire, and should weaken as company quality rises.
It also means investors should treat day-one prints on small direct listings with suspicion. The auction can clear. That does not mean the first price was informative.
The size split
Across 1,266 underwritten IPOs and 33 size-classified direct listings, the ≥$1B cohort (think Palantir, Warby Parker, Amplitude, ZipRecruiter, Coinbase, Roblox, Asana) posts a three-month median of 5.5% with a 55.6% win rate, statistically indistinguishable from the rest of the market. The sub-$1B cohort posts -63.9% with a 4.8% win rate. Underwritten IPOs sit in between at -13.3% / 35.1%.
That is not a mild tilt. Names like reAlpha Tech (-97.5%), TG-17 (-97.4%), FreeCast (-92.0%), and 20/20 Biolabs (-91.8%) are not outliers decorating an otherwise mixed sample. They are the sample.
Returns by listing type
| Listing type | n | 3M median | Win rate (3M) | p vs rest |
|---|---|---|---|---|
| Direct listing ≥$1B | 9 | 5.5% | 55.6% | 0.9353 |
| Direct listing <$1B | 24 | -63.9% | 4.8% | <0.0001 |
| Underwritten IPO | 1266 | -13.3% | 35.1% | <0.0001 |
Direct listings: market cap at open vs 3-month return
Each point is one size-classified direct listing. The x-axis is log-scaled market cap at the opening print; the y-axis is three-month return from that open.
The 2020–21 cohort sits on the right and scatters around flat: Palantir well up, Roblox and ZipRecruiter modestly positive, Coinbase down but nowhere near the left-hand cloud. Below and left, the sub-$1B names cluster tightly between −50% and −100%.
The interesting exceptions are the three points sitting high on the x-axis and deep negative: OBOOK, VenHub, and Polaryx, all 2026 listings. They are not counterexamples to the thesis so much as an artefact of it. Each cleared $1B only because its opening auction printed multiples above the reference price on very thin volume — OBOOK opened 580% above its reference. Size them on the reference price instead and all three fall into the sub-$1B cohort, where their returns say they belong. That is the reference-price warning from earlier, in chart form: for a thin listing, the first print is a number, not a valuation.
As a regression on three-month return, the pooled direct-listing indicator looks like a real effect (-38pp, p=0.0021). Split it and the story collapses onto one side: the ≥$1B indicator is noise (-2.2pp, p=0.9256); the sub-$1B indicator carries essentially everything (-51pp, p=0.0010).
One listing (FreeCast) is omitted from the scatter: its stored market cap is a bad source value and would plot two orders of magnitude too far left. It is included in every statistic above, where the cap is not used.
Caveats
The small-cap sample is small in absolute terms and weighted toward recent listings, so the longest holding periods are thinly covered. Absolute returns here are not market-beta-adjusted; the comparison is to underwritten IPOs over the same horizons, not to the S&P 500 on matching windows. The opening print for illiquid names is volatile enough to shove a company across the $1B line on its own. And a handful of multi-class issuers still cannot be sized from public share counts.
None of that changes the direction or the consistency of the split.
Methodology: listing type is read from each company's registration statement (a direct listing registers a resale with no underwriting table and no offer price) and cross-checked against a hand-maintained academic census of US direct listings. Market cap is the first traded open × shares outstanding at listing; prices are un-adjusted for post-listing splits before that multiplication. Returns are measured from the first traded open. Significance is a Welch t-test of each group against the rest of the population.
The legal history explains how direct listings work. These numbers explain which ones are a problem.