Category: Uncategorized

In the weeks leading up to Google’s IPO, few people had anything good to say about the company or its decision to go public using a modified Dutch auction. (Here’s one notable exception.) But now we’re seeing a welcome backlash to the anti-Google backlash, with a host of articles arguing that, glitches notwithstanding, the IPO worked. (My take is here, but unfortunately you need to subscribe to the Financial Times to read it.)

Most discussions of the IPO have focused, appropriately, on the fact that Google maximized the amount of money it raised by reducing the commissions it paid its investment bankers and by getting itself a fairer price than it would have under the traditional system. (Even though Google’s price did jump 18% on the first day, that was a relatively reasonable discount given all the fear and uncertainty Wall Street had tried to sow about the company and the offering.) As Alex wrote last week, the true test of the success of an IPO is the “cost per dollar of raised funds,” and by that standard Google did well.

But the offering was also a success for another reason, which is that it forced institutional investors to compete, for once, on a level playing field. The problem with the current IPO system isn’t just that companies end up leaving billions of dollars on the table when they go public, but that select mutual-fund and hedge-fund managers (as well as well-connected individuals) are handed what amounts to free money. In a traditional IPO, the investment bank underwriting the offering controls the allocation of shares. In the late 1990s in particular, that allocation process became a way of doling out favors and securing future business. For instance, if you were a mutual-fund manager who funneled a lot of trades through an investment bank — or who agreed to do so — then you were more likely to get a hefty allocation of IPO shares.

This made money managers look a lot smarter than they were — even if you set the bubble aside, there are lots of fund managers whose returns from the late nineties need an asterisk next to them — and it wrecked the price-setting process, since there was no real attempt to let the price reflect the real demand for a stock. It also sabotaged one of the best things about capital markets, which is that in theory they aggregate the opinions of anyone with enough capital and enough risk tolerance to participate, and not just the opinions of those with the right connections. (There should be no velvet ropes in capital markets: if you can pay, you can play.) Google turned all this around: the only way to get shares in the Dutch auction was to do the valuation work and make a reasonable bid. The traditional IPO relies on the power of cronyism. Google’s IPO, flawed as it was, relied on the power of markets. Bad for the Street, good for everyone else.

At this point, it seems clear that Venezuelan president Hugo Chavez won a definitive victory in the recall referendum that the country held a week ago Sunday. The opposition, though, continues to insist that there was massive fraud. There doesn’t seem to be any proof of this, but one piece of evidence that Chavez’s opponents seized on almost immediately was the curious fact that at hundreds of polling stations around the country more than one voting machine recorded the exact same number of “yes” votes (“yes” was a vote for Chavez’s removal). For instance, the Wall Street Journal reported that at one polling station in Bolivar, two machines each recorded 153 “yes” votes while recording 215 and 237 “no” votes.

The opposition argued that this was proof that the number of “yes” votes had been “capped,” so that after a certain number of votes had been recorded, every additional “yes” vote was changed to a “no” vote instead. (Venezuela uses computerized touch-screen voting machines.) And at first glance, this might seem suspicious. But at second glance, it seems like a simple product of chance, as the Journal pointed out:

Aviel Rubin, a computer-science professor at Johns Hopkins University, said he calculated odds of roughly one in 17 that two of three computers at a voting table would have identical results. That compares to about one in 15 that so far have shown similar results in Venezuela’s referendum.

In other words, with twelve thousand voting “tables,” many with multiple machines, it was inevitable that some would end up with matching scores. (It’s similar to the fact that if there are 23 people in a room, the chances are 50-50 that two of them have the same birthday.) Not surprisingly, then, when international observers audited a sample of the results, they found that while there were 402 tables with matching anti-Chavez votes, there were 311 tables with matching pro-Chavez votes, too. What seemed to be proof of fraud was most likely just a statistical artifact.

This is a classic example of what Nassim Taleb calls being “fooled by randomness,” in his intriguing book of the same name. We think that randomness means there will be no clusters or sequences of similar behavior, and therefore when we see them we assume they’re evidence of some hidden pattern. (You can see this in the way people interpret everything from clusters of cancer cases to hitting streaks in baseball.) But they’re really just evidence of the numbers working themselves out.