IPOs are underpriced
And what to do about them
An initial public offering, or an IPO, is when a company goes public for the first time and sells part of itself as stock. Most commonly, a firm is underwritten by an investment bank, who offers some guarantees against it not selling well in exchange for a fee. A substantial quantity of stocks are then put up for sale, all at the same price, to whoever wants to buy. This strikes me as strange. Why sell at one price? Why not have an auction? For that matter, why are IPOs consistently underpriced relative to what they will move to later that day?
i. The evidence for underpricing
There’s a pretty clear body of evidence that they are, in fact, consistently underpriced. A couple papers disagree, but most do find they are underpriced, and are concerned with explaining why. Loughran and Ritter, 2002 find an average underpricing of $9.1 million per IPO, for a total loss of $27 billion — twice the level of fees paid to investment bankers to underwrite. Purnandam and Swaminathan (2004) find that, while they are indeed underpriced on the first day, they are not all that underpriced if you compare to the long-run, arguing that investors are simply a bit irrational on the first day. I am skeptical of markets being inefficient, but could find a single party being biased toward underpricing reasonable. I’m also concerned that they are simply picking up the tech bubble popping — they could have been correctly valued during the dotcom bubble, because their upside potential is much higher than downside, and yet seem to be money-losing because their risks are correlated.
ii. Why? Can we do better?
It is entirely plausible that an auction is impractical. There is substantial uncertainty about the true value of the stock. If we assume risk-averse individuals, then bidders would prefer to underestimate than overestimate the value of the stock, making single-bid auction forms not profit-maximizing for the seller. You would therefore want ascending auctions (which are isomorphic to a second-price auction), which for millions of shares, cannot be run in a reasonable time frame.
What doesn’t fit with this story is that IPOs have been done via auctions before. Google’s initial IPO was run over the web by having everyone interested write what stocks they’re willing to buy and at what price, and then allocating. This was noted as quirky and different at the time, however. You could just run a Dutch auction, and sell them all in one go, in a way which is informative to other bidders. Start the price of a stock at an unrealistically high level, then slowly lower it until all shares are sold. There’s lots of ways around it — selling it at one price is clearly not optimal.
It’s possible that simply taking the stock price, and multiplying it out over the number of shares, exaggerates the true value of the stock. The current market price is what the people who feel most strongly about it would be able to buy it at — it is not implausible that you would be unable to sell the whole lot at that price. Demand curves slope down, after all. Of course, while “mark-to-market” pricing (which is what the multiplying out method is called) exaggerates the price, you will never know quite how much the price will change when you try to sell it all until you do.
So what to do? I think everything would be improved if you released the shares slowly. If you plan to sell 20% of the company, sell 2% at first. If you can, auction them off, but if you can’t that’s fine too. Let people trade and arrive at the market price, and then and only then release the remaining equity you want to sell. You don’t leave money on the table, and you don’t overprice them.
A common knock on auction theory is that it doesn’t tell us anything new, only explain why what people have evolved by happenstance is optimal. This is different. The way we do IPOs is simply incorrect. Companies are needlessly mispricing their assets. We can do better than this!