All the little birdies flying awaaaaaaay. Photographer: Andrew Harrer/Bloomberg
All the little birdies flying awaaaaaaay. Photographer: Andrew Harrer/Bloomberg

(Corrects Twitter's listing venue in 12th paragraph.)

Twitter's stock is off 21 percent in the two days since its employees and venture capital shareholders were allowed to sell stock, on not much other news.1 What does that tell you about the efficiency of markets? My colleague Jonathan Weil is unimpressed: "If the lockup expiration is the reason for Twitter's stock plunge, it would be an awful indictment of the efficient-market hypothesis, which holds that the prices of publicly traded securities reflect all publicly available information."

Now of course the stock market is not just about predicting the future cash flows of companies. It's also about balancing supply and demand for shares of stock. Those are related functions: Some people think Twitter will be worth a lot in the future and so would pay $50 for it; others think it will be worth very little in the future and so would only pay $20 for it. You can, if you squint, build a partial demand schedule:

That's pretty crude, but you get the idea. On Monday, there were 80.5 million shares you could buy, and 80.5 million shares someone could sell -- just the ones sold in the initial public offering (plus the greenshoe).2 And the stock closed at $38.75. Today, there are 588 million, and the stock closed yesterday at $30.66. Oh sure the holders of some 205 million of those 588 million shares have promised not to sell, but that seems like a nonbinding voluntary promise based mostly on the fact that they place a high valuation on Twitter3 -- as opposed to the lockup agreement, which was just a requirement of getting liquidity in the first place.

Now in some sense -- not a very accurate sense but whatever -- either the people who think Twitter is worth $30.66 are right, or the people who think it's worth $38.75 are right.4 The fact that there were more shares for sale yesterday than there were on Monday shouldn't change that.

The way financial markets are supposed to deal with this is called, loosely, arbitrage. If Twitter is actually worth more than $30.66, smart investors are supposed to come in and buy it until they bid the price up to the appropriate level. Notice, though, that the people who supplied tons of new stock at prices below $38.75 over the last two days are insiders: employees, ex-employees, early investors. It's not necessarily the case that a former Twitter employee would know more about Twitter's future value than a professional equity investor, but on the other hand, your risk of adverse selection in buying from the people dumping shares on the lockup release does seem pretty high. Higher than it was on Monday. You could see why smart money might take a few days off from buying.

But there is -- or was -- an arbitrage the other way too. Everyone knew this lockup expiry was coming. If Twitter was actually worth something like $30.66, why didn't its price fall to that level already?

The answer is partly in that (fake) demand schedule: To get to $30.66, you'd need 588 million shares for sale (or whatever), and there were only 80.5 million shares for sale. The fix for this is short selling. If there are 80 million shares floating, and you borrow 40 million and sell them short, then buyers can get 120 million shares. You can keep going: Borrow those 40 million shares again once they've settled, and you can get to 160 million. Or 200 million.5 There's no theoretical limit on this. But there is a practical limit; borrowing heavily shorted stocks gets expensive, and many shareholders won't lend you the stock (because you're using it to drive down the price).

So in fact, Twitter had a float of around 80.5 million shares as of Monday, but there were around 127 million shares available to buy.6 That is, there were about 46.7 million shares of short interest.7 That's over half of the shares that Twitter had actually sold to the public -- but still a drop in the bucket compared to the shares that came off lockup yesterday. This could mean that investors didn't expect as much selling as they got: "'More stock came for sale than people expected,' said Ian Winer, director of equity trading at Wedbush Securities Inc.," though he would say that. But it could just mean that it was tough to position for all the stock that was coming: You borrowed all you could, but you couldn't borrow as much as you wanted.

Compare Facebook, which was up on its big lockup expiration in November 2012. Between its IPO and intervening smaller lockup expirations, Facebook had around 884 million shares that were freely trading, not so much less than the 1.2 billion coming in off lockup. And investors weren't shorting everything they could -- perhaps because they could predict that insiders wouldn't want to sell.8 And it did much better:

Shares in millions
Shares in millions

That seems neatly explanatory! But I'm mostly messing with you. In fact, if you look at a broader sample of recent tech lockup releases, they're all over the place, and those statistics have no explanatory power at all.9 The behavior of insiders seems to be unpredictable, and the market seems to do a near-random job of predicting it. The limits to arbitrage, in immediately-post-IPO tech companies, seem to be constraining.10

When Twitter priced its IPO, I wrote about how difficult it is to transition from a private company to a public one:

The transition from "private company valued essentially by bankers' appraisals and insiders' hopes" to "public company valued by a ruthless and shifting stock market" is a necessarily harsh one, and the underwriters' role is to smooth that transition as best they can. ... One obvious move that the underwriters can make to smooth that transition is to allocate a lot of shares in the IPO to a smallish number of large, long-only, "real money" investors who believe in Twitter's story and want to be long-term investors. You know: Like the pre-IPO investors. The rumor is that that's what happened, with hedge funds looking for a quick flip getting few or no shares, and with a few large long-only investors getting big allocations.

That transition is hard, but it is only partially accomplished by the IPO. Once Twitter went public, sure it traded on the New York Stock Exchange, but most of its stock was held by locked-up employees and insiders whose views were not reflected in the market price (because they couldn't trade11 ). The actual NYSE trading was a small stub of the company -- more informative than the pre-IPO secondary market trading, but not a perfect reflection of market views either. With Tuesday's lockup expiry, the transition to public markets was (more or less) completed.

And the transition is hard! It's hard to get the IPO price right, which is why so many stocks trade way up (or, less often, way down) on their first day. And it's hard for the post-IPO market to get the post-lockup price right, which is why so many stocks trade way down (or, less often, way up12 ) on the day the lockup expires. The fact that the market can't accurately predict the market price should make you more sympathetic to the bankers, who can't do it either.

That's the weird thing: A simple story you could tell about Twitter is that it priced its IPO at $26 a share and, when it finally became fully public six months later, its shares ended up around 18 percent above the IPO price. That's ... pretty good?13 The much criticized first-day pop reflects stiff competition to buy the relatively small stub of Twitter that was outstanding for the last six months. The similarly-criticized post-lockup drop reflects a move closer to the "real" market. If you bought Twitter as a "real money" investment six months ago, you got pretty much a reasonable return on your money. There were some twists and turns along the way, but the market pretty much worked.

1 Except investors clearing out room for Alibaba or whatever, but that's not really "news" either.

2 This oversimplifies -- there are 70 million shares in the IPO, 10.5 million in the greenshoe (which was exercised), and some bits and bobs (9,867,228 shares) that came off lockup early but only for tax vesting. I'm ignoring those, though the number is non-trivial.

3 This obviously makes my demand schedule wrong -- many of them would be buyers at $38.75, or above. But I meant that mostly as a joke.

4 Or there's been new information about its future cash flows in the last two days but let's discount that.

5 No one believes this, but it is nonetheless true. There can be more short interest than there are shares outstanding, even without "phantom shares" or naked shorts or anything creepy or illegal. See footnote 6 here, but really there can be an infinite number of shares sold sort with a long enough chain of rehypothecation.

6 Again my demand table is fake! Sorry.

7 This is reported only every two weeks, and on a delay, so the most recent data is as of mid-April. But it was on its (slow) way up then.

8 Maybe because it was trading at like half the IPO price?

9 Like, here:

Shares in millions
Shares in millions

10 By the way, which form of efficient markets hypothesis is this? One thing you can think is that the employees, venture capitalists and so forth selling in the lockup expiry have nonpublic information that is not available to the broader market and is only incorporated by their selling.

11 Couldn't sell, technically. They could buy all they wanted, mostly.

12 By the way the reasons for these directions are straightforward supply/demand imbalances. Lots of people who want to buy private companies can't, so when they finally can (at IPO), they bid the price up. Lots of people who want to sell just-post-IPO companies can't (they're locked up), so when they finally can (at expiry), they push the price down.

13 Like, if you think that the right IPO first-day pop is 10-15 percent (to compensate for the risk and work of investing in the IPO), then 18 percent in six months feels pretty good, right? Like $30.66 in six months is worth say $28-$29.25 now, discounted at 10-20 percent, equivalent (???) to a 7.6 to 12.4 percent IPO "pop." Heh.

To contact the writer of this article: Matt Levine at mlevine51@bloomberg.net.

To contact the editor responsible for this article: Zara Kessler at zkessler@bloomberg.net.