(Reprinted from HKCER Letters, Vol. 85 January-March 2006)
Unlocking the Winner's Curse in the Application List System
Property prices in Hong Kong hit a record high in October 1997. Less than six years later, the market went down by 66% and hit rock bottom in July 2003. In January 2006, the market prices were still 48% below the peak.2 The government took drastic measures to stabilize the market by suspending land sales for nine months, from June 1998 to March 1999. After the moratorium, the government introduced the Application List System or ALS (see below) which operated in parallel with scheduled auctions (which had a long history in Hong Kong). Land sales were, however, suspended once again from November 2002 till December 2003. When land sales resumed in January 2004, scheduled auctions had been completely replaced by the ALS.
The Application List System (ALS) is a hybrid of tenders and auctions. The fundamental difference between tenders on the one hand and auctions on the other is that the former uses sealed bids, while the latter is based on open outcries.3 The system was introduced in 1999 (ALS 1.0) and subsequently modified in 2005 (ALS 2.0). The following section describes ALS 1.0, while the subsequent one describes ALS 2.0.
2. ALS 1.0
Each year, the government updates and announces an application list, which contains sites available for sale in the next financial year. If a developer is interested in a site listed in the application list, the developer can make an offer to the government. The government compares the offer to the undisclosed reserve price and accepts or rejects it depending on whether it exceeds or falls below that price.4 If the offer is accepted, the developer is required to pay a deposit.5 The government subsequently arranges an auction for the site, which takes around 10 weeks from the date of accepting the offer. The accepted offer is used as the starting bid (or called the upset price) of the auction. On the day of the auction, if the triggering developer, or other developers, bids at the upset price, the deposit is returned to the triggering developer (without interest). If no developers (including the triggering developer) bid at the upset price, the deposit is forfeited. The government withdraws the site if the auction fails to reach the updated reserve price (based on a reassessed OMV).
The displacement of scheduled auctions by the ALS eliminated the need for a moratorium on land sales to stabilize the market. Here is the argument. When the property market is weak, developers are less interested in making offers to the government. Even if offers are made, substantial discounts are factored into those offers, and are likely to be rejected by the government. The government can, in any case, reject offers if it believes that accepting such offers and putting up new sites for sale will destabilize the market.
In principle, the advantage of the ALS over scheduled auctions is more general by allowing the supply of land to be market-driven. The preceding paragraph also shows that land supply decreases when the property market is weak. On the other hand, if the market is robust, land supply should increase, because developers would be more interested in making offers to the government. The problem is that the ALS does not seem to work well in increasing land supply. This article attempts to shed light on this problem from a game theoretic perspective.
3. ALS 2.0
In the year after scheduled auctions were replaced by the ALS, there was a significant drop in the number of land auctions. Only six auctions were held in 2004-2005, whereas 12-13 auctions took place in each of the fiscal years 1999-2000, 2000-2001, and 2001-2002.6 Furthermore, no residential sites were successfully triggered for almost a year from October 13, 2004 to September 26, 2005.7
Developers blamed the small number of auctions on too high the trigger price required by ALS 1.0, which was set as the reserve price or higher. Also, under this framework, a successful trigger committed the developer to paying the trigger price. Note that the commitment was purely one-sided. The triggering developer received neither compensation nor favorable treatment from the government. The developers thought that this financial commitment was too high, and urged the government to lower the trigger price. They argued that a site would not be sold at a lower price once an auction had been triggered.
The government responded by modifying the ALS in 2005. ALS 2.0 reduced the commitment by (1) allowing the trigger price to be as low as 80% of the reserve price, (2) standardizing the deposit to 10% of the trigger price with a cap of $50 million, and (3) shortening the time between a successful trigger and the auction from 10 weeks to around seven. Subsequent to the modifications, three sites were triggered and the auctions were held in September 2005. However, no other site was triggered in 2005-2006. ALS 2.0 did not seem to be effective.
4. Two Proposals
The situation is still problematic, and industry practitioners have made two proposals. One proposal claims that the government's concessions of a 20% discount are not enough, and the proposal asks it to further lower the trigger price. While this proposal supports the use of the ALS in principle, and merely attempts to improve the operation of the ALS, the second proposal is more critical about the ALS and calls for a reinstating of the scheduled auctions.
The second proposal is, however, shortsighted. The 1998-1999 and 2002-2003 experiences show that reinstating scheduled auctions, whether they operated in parallel with the ALS or not, exposes the system to the risk of land sale suspension. A moratorium on land sales is a purely administrative decision. It interferes with market operations and violates the market-led philosophy of the Hong Kong Government on economic policies. More important, this proposal fails to appreciate the general advantage of the ALS over scheduled auctions: the ALS allows land supply to be market-driven, while scheduled auctions do not provide that flexibility.
The government¡¦s response to the second proposal is to consider having scheduled auctions for a limited number of commercial sites, but to refuse the use of scheduled auctions for residential sites. It is unclear why the government believes that the dual system of using the ALS and scheduled auctions would work better for commercial sites than for residential sites. Recall that the only successful trigger in the one-year period from October 13, 2004 to September 26, 2005 was a commercial site, when only the ALS was in operation. Reinstating scheduled auctions is a premature decision and it merely detracts from the discussion, which should focus on how to refine the ALS instead.
Can the first proposal of pushing for lower trigger prices improve the ALS? The government reveals that some developers' offers are only 30-40% of the reserve price. Such low offers are considered by the government to be unacceptable. The government reiterates its policy of not selling land cheaply. Given that the government maintains the same reserve price, the problem with the first proposal is examined next.
Since some offers are only 30-40% of the reserve price, the trigger price needs to be lowered substantially to induce more successful triggers. Suppose the trigger price is lowered to 50% of the reserve price. The triggerer can get back the deposit as long as it bids at half of the reserve price. The commitment is minimal, and with such a low hurdle, more sites would be triggered for purchase. However, in these auctions, it is not unlikely that the highest bid would fail to reach the reserve price, and the auction would then have to be withdrawn. An auction withdrawal will introduce uncertainties in the property market and should be avoided.
The discussion so far shows that (1) reinstating scheduled auctions is a premature decision and (2) lowering the trigger price any further is not advisable. Can the ALS be fixed? A solution is proposed below. To introduce the solution, however, the following concept needs to be understood first.
5. The Winner's Curse
The concept is known as the winner's curse, which arises when bidders have incomplete information of the value of the object. Each bidder needs to assess this unknown value. The assessments will be varied if they are formed independently of each other. The representative case assumes that the most favorable assessment exceeds the true value of the object and the least favorable assessment lies on the opposite side. If bidders are naive, they bid according to their assessments. The bidder with the most favorable assessment submits the highest bid. The bidder wins the object but suffers a loss because the most favorable assessment exceeds the true value of the object. The winner is cursed.
The more interesting behavioral assumption for the bidders (or developers, in this case) is that they are sophisticated. They are aware of the winner's curse, and to avoid it they submit bids that are lower than their assessments.8 This conservative bidding reduces the seller's revenue. Hence, the winner¡¦s curse backfires i.e. it becomes the "seller's curse". (In the rest of the article, bidders as well as the seller are assumed to be sophisticated.)
To prevent backfiring, the general principle (for the seller) of designing an auction is to make the auction less prone to the winner's curse. Comparing a tender to an auction provides an illustration of the general principle. Recall that tenders use sealed bids and auctions use open outcries. The bidding behavior is not observed under tenders, but is common knowledge under auctions. Consequently, bidders reveal and share their information via bidding under auction. This process reduces information incompleteness of the bidders. With more information, or less uncertainty, the winner's curse is less of a concern to the bidders, and hence, they bid more aggressively under auction. On the other hand, tenders do not facilitate information sharing and result in more conservative bidding. So to maximize its revenue, the seller should choose auctions over tenders.
6. A Counter-Intuitive Proposal
It should be pointed out that the general solution of replacing the tender (in the pre-auction stage) with an auction is not applicable to the ALS, since there is already an auction in the second stage. Neither is scrapping the tender a solution, because it is equivalent to reinstating scheduled auctions. The question is how the winner's curse can be used to devise a solution for the ALS. A solution is presented below, which is quite counter-intuitive, so it is interesting in its own right.
Before we proceed, recall that the ALS has two stages. The first stage, the pre-auction stage, is a tender, and the second stage is an auction. The discussion in the last section has shown that the tender in the first stage of the ALS is more prone to the winner's curse. More specifically, in both ALS 1.0 and ALS 2.0, a single developer can trigger an auction if its offer is accepted by the government. Once the offer is accepted, the triggering developer knows that its offer is the highest. Once the auction is triggered, the developer is committed to bid at this offer, or stands to lose the deposit. This is the winner's curse, and explains why offers are conservative in the pre-auction stage.
The culprit of the winner's curse is the winning bidder's concern of overbidding. This concern can be eased if the bidder knows that another bidder is willing to bid similarly, as in the case of an auction. Based on this observation, ALS 2.0 is modified to require two bidders to trigger the auction, a proposition that is clearly counter-intuitive. The following questions need to be addressed.
Will the new condition make it more difficult to trigger an auction?
The answer is, surprisingly, not positive. The difficulty of triggering an auction will depend on whether the new condition induces bidders to bid more aggressively; this in turn is a result of the new setup. An auction can now only be triggered when there are two independent offers exceeding 80% of the reserve price. When the government receives an acceptable offer, the developer and the public are informed. The details, however, such as the identity of the developer and the amount of the offer are not disclosed. Bidders contemplating their bidding decisions know that either the government has accepted an earlier offer or that theirs is the first one. Both cases are examined below.
Will a bidder be more aggressive if he/she knows that an earlier offer has been accepted?
The answer is "yes". Suppose a second developer makes an offer high enough to trigger an auction. The two developers probably think that their offers are similar in magnitude. Each is committed to bid at its own offer, or stands to lose the deposit. Roughly speaking, the chance of winning by overbidding is halved, because there is always another developer preparing to bid at a similar level. This eases the winner's curse and induces developers to bid more aggressively.
Will a bidder be more aggressive if he/she knows that no offer has been accepted?
The answer is, once again, "yes". Suppose a developer's offer is the first offer accepted by the government. The developer has no commitment unless the government accepts a second offer. When the second offer is accepted, it triggers an auction, as well as the commitment of the first developer. The two developers have similar commitments, which brings us back to the preceding case.
7. ALS 3.0
This section refines the proposal outlined in the last section, which uses the winner¡¦s curse to understand the problem faced by the ALS and to devise a solution for it.
The new condition will lengthen the time between the first acceptable offer and the auction. The government should consider requiring the first developer to pay the deposit only when the second acceptable offer is received, or paying interest to the developers on their deposit amounts. Furthermore, the condition allows the government to relax the deposit of each developer to 5% of its offer, subject to a maximum of $25 million. Halving the deposit amount will ease the winner's curse further and induce more aggressive bidding.
It is possible that there may be some delay before a second acceptable offer is received. The government may be criticized for waiting too long and may be forced to launch an auction without the second acceptable offer. Such a credibility problem is known as time inconsistency. One way to deal with it is to allow the first developer to choose between (1) triggering an auction on his/her own and paying a 10% deposit (subject to a maximum of $50 million) or (2) waiting for the second acceptable offer to trigger an auction and paying a 5% deposit (subject to a maximum of $25 million). Time inconsistency does not arise if the developer chooses option (1). If it chooses option (2), requiring a second acceptable offer to trigger an auction becomes a commercial agreement between the first developer and the government. The government has to honor the commercial agreement or bear the legal and reputation consequences of breaching it.
A game theoretic approach is adopted in this article to tackle the problem with the prevailing ALS. The concept of the winner's curse is used to devise a solution, and the main features of the solution (ALS 3.0) are summarized below:
¡P Maintaining the trigger price at 80% of the reserve price or higher.
¡P Allowing the developer who submits the first acceptable offer to choose between:
¡P Triggering an auction on his/her own and paying a 10% deposit (subject to a maximum of $50 million); or
¡P Waiting for the second acceptable offer to trigger an auction and paying a 5% deposit (subject to a maximum of $25 million).
¡P Paying interest(s) on the deposit(s).
The highlighted feature is the most important one. The analysis of the winner¡¦s curse shows that requiring two independent offers to trigger an auction is, per se, already sufficient to induce developers to bid more aggressively.
1 School of Economics and Finance, The University of Hong Kong, Pokfulam Road, Hong Kong. I am grateful to Charles Lai for his excellent research assistance and the SEF and HIEBS for research support.
2 The figures are based on the Territory-Wide Private Domestic Price Index of All Classes found in the Hong Kong Property Review published by the Rating and Valuation Department. The index levels were 172.9, 58.4, and 89.7 in October 1997, July 2003, and January 2006, respectively.
3 This follows the common usage of the two terms. It is different from the formal usage in the auction literature, where auction is a generic term. Tenders are formally referred to as sealed-bid auctions, and auctions are called open auctions.
4 The reserve price is based on the assessment by the government called the Open Market Value (OMV).
5 The amount of deposit is not standardized in ALS 1.0 (and was rectified by ALS 2.0).
6 The second moratorium on land sale spanned from November 2002 to December 2003.
7 Only one commercial site was triggered during the period and the auction was held in February 2005.
8 The winner¡¦s curse is exacerbated if the bids are higher than the assessments.