Lynne Kiesling
Today’s Opinion Journal has a commentary by Art Laffer and Patrick Giordano on proposed wholesale power procurement auctions in Illinois. They claim that the uniform price reverse auction (in which the price falls until Q supplied=Q demanded, and all suppliers still in the market receive the market-clearing price) is a ComEd setup that’s tailor-made for suppliers to collude and for ComEd’s generation company to get a higher price than they would receive under another auction design. Laffer and Giordano recommend, instead, a pay-as-offer reverse auction, in which (I suppose, but the article is not entirely clear on the design points) the suppliers drop out as the auctioneer ticks down the price, stopping when the last supplier drops out. Then the auctioneer goes back up that offer curve to where the quantity demanded equals the quantity supplied, chooses the suppliers that submitted the lowest offers, and you’re done. Laffer and Giordano argue that such an auction would deliver lower prices to consumers.
Not necessarily. Laffer and Giordano assume that suppliers will reveal their reservation prices truthfully. But we have decades of field and experimental data to show that pay-as-bid/pay-as-offer auctions induce participants to misrepresent their true values. Think about it: if you know that you are going to receive exactly what you offer, are you going to offer at your marginal cost? Heck no! You are going to shade your offer upward. In experiments ranging from Treasury bond auctions to wholesale power markets, there’s a four-decade-long experimental literature exploring the intricacies of uniform vs. pay-as-bid auctions, and in general uniform price auctions are more efficient, although whether the buyers or the sellers benefit more depends on the application.
Their point about sharing offer information with participants in real time is a valid point, and worthy of research. Their point that the proposed procurement auction favors ComEd’s generation company is uncontroversial. But they miss the point entirely when they assume that suppliers will make truthful offers in a pay-as-offer reverse auction. It’s a fatal assumption, not borne out in reality or in the laboratory.
I would say that pay-as-bid demands much more than ?shading upward.? In essence you have to bid somewhere near what you think the clearing price will be, i.e., you have to be a forecaster in addition to being a producer, even if you?re infra-marginal. In a single-price situation, a bidder with the potential for setting the price may shade upward from its marginal cost to explore or take advantage of his headroom, his cost advantage over his nearest higher-cost competitor. He has to study the expected market conditions, but his bids are at least disciplined by his nearest competitor.
Alas, pay-as-bid will never go away, seeing as how facts don?t matter and people never learn. ELCON complained about single-price auctions in its paper ?Problems in Organized Markets? earlier in the year. But what seemed to really be at the root of this concern is that the single-price auction allowed the infra-marginal units to be lazy, giving them no incentive to go out and find customers whose load shapes matched the baseload output. If the baseload generators had to sell only to people with flat load shapes, then the pool of potential buyers would be small and the value of baseload generation would be less. Followed to its logical end, this objection was aimed at the efficiency of the clearing market mechanism, because it ruined what they had hoped would be an advantage. I prefer to think of it as giving the baseload generators a clear signal of what their power is worth, so that they contract on that basis instead of on monopsony pricing.
Interestingly, Hogan and Harvey showed in 2001 that the separate, sequentially clearing markets for energy and ancillary services in California introduced a pay-as-bid characteristic for some bidders, causing them to bid into all of the auctions at the price they expected the last market to clear. A separate observation (my own) on California is that with single-part bidding in separate hourly auctions, generators would often be unwilling to run at their marginal cost since it was lower than average cost. Some generators, in fact, would never be willing to run at their marginal cost at any level of output. So the strategy in California for many generators would have been to bid an hourly price ?shape? that would keep the unit online overnight while recouping the cycling costs during the peak hours. As in pay-as-bid, they were having to forecast price instead of bidding marginal cost. Nevertheless, regulators and monitors were aghast that generators didn?t bid their marginal costs. For that market design, it was an unrealistic expectation.