Michael Giberson
The RGGI auction design final report released in October 2007, which contains the proposed auction design for the sale of RGGI CO2 permits, overall presents a thorough examination of the market design problem and a well-thought out justification for their final design proposal. There is, however, one serious flaw that threatens the efficiency of the auction: they proposed to auction two goods which are asymmetric substitutes in separate, simultaneous sealed bid auctions.
The goods being auctioned are carbon emission permits for the 10-state RGGI region. Permits are identified by the year (“vintage”) in which they can first be used. In each auction it is proposed that both the current vintage permit and a future vintage permit be sold. Because the current vintage can be banked, and used to meeting requirements in later periods, the current vintage is a substitute for the later vintage. The reverse is not true, you can’t borrow forward, making the goods asymmetric substitutes.
A similar situation arose when Virginia sought to auction off both 2004 and 2005 NOx permits at the same time, with the 2004 permits an asymmetric substitute for the 2005. In a paper describing the research and development of the Virginia NOx auction design*, the researchers observed that offering the permits in separate auction creates a puzzle for bidders and problems for the auctioneer:
Participants submitting such bids would be unable to indicate to the auctioneer whether they were willing to accept 2004 allowances in place of 2005 allowances, and if so, the exchange rate at which they would be willing to do so. This would force bidders to choose between bidding too conservatively or exposing themselves to financial loss. It would also deny the auctioneer flexibility in selecting an allocation to maximize revenue and efficiency.
The design team recommended a combinatorial ascending clock auction, but due to the limited time available to develop and deploy the auction, the state chose to conduct sequential ascending clock auctions instead. The auction raised about $10.8 billion in revenue, nearly 20 percent over pre-sale estimates. They considered, but rejected, a combinatorial sealed-bid auction, and didn’t consider a simultaneous separate, sealed-bid auction for the reasons stated.
In the case of the RGGI, the failure of the proposal to accommodate substitutability in the auction design was noted by Peter Cramton, an auction design expert working on behalf of the regional power markets in New England and New York.** By employing separate simultaneous auctions, Cramton explained, the proposed auction design puts bidders in a position where they have to guess which product is going to be a better value, and if they guess wrong they could lose big. (Like the Virginia NOx folks, Porter et al., said: in such a position bidders have “to choose between bidding too conservatively or exposing themselves to financial loss.”) Cramton also said the sealed-bid design may not perform price discovery as well as alternative designs.
Incredibly, the only mention of key words like “substitute” in the final RGGI auction design report come in the annotated bibliography (in the summary of the Porter et al. paper!). The report discusses banking of permits, but fails to examine how banking interacts with the bidders’ problem when substitutable goods are offered in separate, simultaneous auctions. In fact, despite the overall thoroughness of the auction design team testing in the lab, it appears from their reports that they did not test the simultaneous sale of two vintage products. Banking was tested solely by running a simple spot market after the lab auction, but only one product was involved.
Because the auction team did not examine the simultaneous auction of two products in their experiments, their recommendation to conduct separate, simultaneous sealed-bid auctions should be considered untested in the lab and not directly supported by their analysis.
Cramton, in his report on behalf of the New England and New York power markets, recommends a simultaneous ascending clock auction design. Worth reading if you are into these kinds of issues. Simultaneous ascending clocks pretty much are, as Cramton describes them, becoming the industry standard for these sorts of market situations. The Virginia NOx design team tested similar designs in their laboratory testing. There is little doubt that such a design would do the job.
My sense of the RGGI situation is that their design issue could be remedied with a much less drastic accommodation. A linked sealed-bid auction would allow bids for the future vintage to specify the premium, if any, which they would pay to receive the current vintage product. Permits could be allocated using a surplus-maximization rule, with a uniform clearing prices for each of the products. Of course, this suggestion is untested in the lab as well.***
While keeping the sealed-bid nature of the original proposal sacrifices the superior price discovery properties of the ascending clock, in this particular case a lot of the price discovery will occur in the secondary market between the proposed quarterly auctions . If, as the RGGI team assumed, the secondary market is relatively active, price discovery becomes a less important consideration in the selection of an auction design.
My recommendation is that the RGGI folks send their design team back to the lab for one more set of experiments to test the efficiency effects of various auction designs for the simultaneous sale of two asymmetric products. I’d suggest the separate, sealed-bid auction proposed with an ascending clock design and the linked sealed-bid design.
While the RGGI folks are anxious to have their design in place so the auctions can begin in 2008, better to address this important issue first.
NOTES: *David Porter, Stephen Rassenti, William Shobe, Vernon Smith, and Abel Winn, “The Design, Testing and Implementation of Virginia’s NOx Allowance Auction,” forthcoming in the Journal of Economic Behavior and Organization.
**Peter Cramton, Comments on the RGGI Market Design, submitted to the RGGI board by ISO New England Inc. and the New York Independent System Operator, November 15, 2007.
***The “linked sealed-bid” proposal is similar to the combinatorial sealed-bid auction tested in the Virginia NOx research.