Optimal Auctions Blog

Designing a Better Google IPO Auction

Posted by Mike on August 19th, 2012
Designing a Better Google IPO Auction

On the very first day of microeconomics you learn that demand is represented by a curve. If something costs a lot, you want a low quantity of it, and if it costs less, you will want more of it. So if every Intro to Micro student can grasp this concept, why do Dutch Auction IPOs like Google's get it so wrong by making the bidders choose a single demand point with their bid? Demand is a curve, not a single point! You'll see why this mistake led to the big first day price pop of Google, and see just how much money Google left on the table during its IPO.

Demand is a curve, not a single point!

The "Dutch Auction" Google Used

Before we continue further, we should dispel one myth - that the auction format Google used is a Dutch Auction. In fact, as you can read about in my other blog post about the history of dutch auctions, this was not a dutch auction at all. Dutch auctions sell 1 indivisible product to multiple buyers. This is definitely not the case when you are selling 19,605,052 shares of stock.

Instead, the Google IPO process is properly called a "Second Item Auction". Other synonyms include "Dutch Auction IPO", "OpenIPO Auction", or "Google IPO Auction". Whatever term you choose though, don't call it a dutch auction. Though it sounds catchier, it's not the correct term.

The way a Second Item Auction works is by taking 1 bid from any interested buyer. For example, in the Google IPO, you might have seen the following bids:

  • Alex bids for 500 shares @ $200/share
  • Bob bids for 10 shares @ $10/share
  • Carl bids for 100 shares @ $100/share
  • Doug bids for 1000 shares @ $50/share

Alex placed 1 bid: 500 shares @ $200/share. What if he was also interested in 250 shares of Google @ $300/share? What if he was interested in 1000 shares of Google @ $85/share? Unfortunately for Alex, he has to choose 1 (and only 1) bid to place in this auction.

When all the bids are received, the "clearing price" is determined. This is the price where all 19,605,052 shares are sold. In economic terms, the clearing price is the price where supply meets demand. The key part of the Second Price Auction though is that every bidder only pays the clearing price for their shares.

Google's clearing price turned out to be $85/share. So, if we look at our 4 bidders from above, we can see that Alex would win 500 shares @ $85, Carl would win 100 shares @ $85, and the other 2 would not win any shares of Google.

But Alex was willing to buy 1000 shares at $85/share, and now he only has 500 shares! Alex is the perfect example of the problems in this auction design. Obviously demand did not meet supply because Alex wants more shares at the $85 clearing price than he actually purchased. This should also make it obvious why the share price popped the first day - people like Alex were willing to buy more shares at a higher price than $85.

The Problems in This Auction

The problem in this auction is that it breaks with the basic fundamental rule of microeconomics! Demand is represented by a curve. Demand is NOT represented by a single point. The way the Google IPO auction was set up, it required bidders to pick a single point on their demand curve and to use that point as their bid.

Alex's Demand Point and the Clearing PriceWe've shown the problem in the auction using our example bidder Alex. He has a very simple demand curve, with only 3 points. If we can break the auction design with just 1 bidder and a demand curve with only 3 points, imagine how broken the Google IPO was with thousands of bidders.

Look at the results of the Google IPO and Alex again. Google's clearing price was $85/share, and Alex received 500 shares. However, Alex's demand curve says he wants to own 1000 shares when the price is that low. So, when Google starts trading on the market immediately after the IPO, Alex will quickly try to buy 500 additional shares, to fill his demand. With thousands of bidders like Alex all trying to buy additional shares, the result will be the big price pop on the first day, which Google saw exactly. Google's shares popped to over $100 on the first day, as bidders just like Alex sought to fill their demand.

How to Fix This Auction

The solution to this problem is to go back to basics - allow bidders to enter their entire demand curve instead of making them choose a single demand point. The Google IPO simply aggregated all the individual demand points into an aggregate demand curve, from which it calculated the clearing price. The correct solution would be no different - aggregating the demand curves from every individual into an aggregate demand curve, and again finding the clearing price for the auction.

What are the drawbacks of switching to this method? Of course, nothing comes free and there is a minor drawback to using this auction format instead of a simple Second Price Auction. Bidders may not understand how to properly construct their demand curve. Entering a single point from their demand curve is simply easier to understand for bidders.

The Optimal Auctions Solutions to the Google IPO

Optimal Auctions has a solution to the Dutch Auction IPO problem, which we call our Advanced Dutch Auction solution. Using our software, we allow bidders to go in and create their demand curve in its entirety. Our software also allows bidders to create as many, or as few, points on their demand curve as they want to (This is even configurable by the auction administrators). To make sure that bidders understand the bids they are placing and the demand curve they are creating, we visually reinforce their bids by showing them their demand curve dynamically as it is created. Finally, we ensure that their demand curve is truly a downward sloping curve, and we run error checking before accepting the bidders' bids.

Once all of the bids have been placed, we run our solving algorithm to create an aggregate demand curve from all of the individual demand curves, and we determine the clearing price in the auction. Then we award the shares to the bidders at their desired price, calculated using the demand curve they created themselves.

Optimal Auctions Advanced Dutch Auction format

How the Google IPO Should  Have Happened

The results of the Google auction would have been different had they been using the Optimal Auctions Advanced Dutch Auction format for their IPO. Instead of individual demand points, every bidder would have been allowed to create their entire demand curve.

The errors in the auction design of the Google IPO Auction cost Google and shareholders $273,000,000!

The clearing price would have been about $99/share (The final day closing price was $100.34). This means a proper auction design would have netted Google an additional $197,000,000 and the selling shareholders an additional $76,000,000. Because the "opening pop" would be eliminated with the proper auction design, the shareholders (Google and the sellers) would have netted all that gain themselves, instead of yielding it to the market.

Interesting Note: Several members of the Optimal Auctions team placed bids in the Google IPO knowing that the auction was improperly designed and knowing that the shares would pop on the first day because of that. They sold their shares quickly when the price reached $100 that first day and pocketed a quick 15% profit. To them, this was easy money.