1. Introduction
Auctions are used by governments for a variety of purpose. Governments use auction to allocate various publically owned natural resources. State-owned land is commonly sold at auctions. Central and/or local governments may auction large amount of undeveloped land, residential housing, and commercial real estate.
Auctions provide several purposes. Auctions can raise revenue for the government, and can generate an efficient allocation (Ashenfelter and Genesove, 1992; Lusht, 1996).1 More importantly, an auction can reveal information: how strong the demand is, and how bidders value the asset.
Auctions are transparent and fair. In an auction, the government has no allocation discretion to influence the allocation outcome. In that sense, impropriety, or even corruption can be prevented. Auction outcomes are determined independent of policy makers’ preferences. In summary, land auctions allow governments to obtain maximum pricing through an efficient and transparent bidding process that sets a realistic benchmark for pricing through broader market exposure.
Auctions put power over pricing into the hands of bidders, with little or no role for the government in pricing and allocation. Nonetheless, auctions are challenging for bidders. Whether it is best to place pricing entirely in the hands of investors depends in part on how sophisticated those investors are at evaluating the object.
When there are common values, the Vickrey auction is efficient. When a buyer's information is multidimensional, no auction is generally efficient. The auction model of incomplete information predicts ‘winner’s curse’, since the winner of a sealed-bid
1Lusht (1996) finds that auction sales generate a premium of 8% for properties sold
auction of unknown common value tends to overestimate the true value of the auction object (Giliberto & Varaiya, 1989)
In a model for auction with endogenous entry, rational investors will choose to acquire information and enter an auction if they expect to recover their costs evaluating the object (Sherman, 2005). If they follow the optimal entry and bidding strategies, the entry of these informed bidders and the aggressiveness of their bids will be positively related to returns to winning bidders. I therefore want to test whether the bidders’ entry decisions depend on information costs, and how the aggressiveness of their bids depends on the entry of bids2. Chiang, Qian, and Sherman (2010) use Taiwan’s auction IPO data to empirically test Sherman’s information production model. They find that the unexpected entry of more institutional investors is related to higher expected returns, suggesting that institutional investors bid based on information. The bids of individual investors, in contrast, exhibit evidence of return chasing, and the unexpected entry of more individual investors is related to lower returns, a sign of systematic overbidding perhaps due to inadequate bid shaving. It means that unexpected entry not based on information will lead to a higher clearing price and hence lower returns (Engelbrecht-Wiggans and Katok, 2005). In this paper, I will test the effect of unexpected entry on winning price of land auctions in Taiwan.
Other papers discuss land auctions include Ashenfelter and Genesove (1992); and Lusht (1996). These literatures, however, do not examine bidders’ entry decisions and the effects of entry on bid premium.
My sample includes five-hundred-ninety-seven land auctions held by the government in Taiwan during 09/2007–02/2010. Not like many countries that all land
virtually owned by the government, land auction by the government in Taiwan is not
2 In a first-price, sealed-bid auction, number of bid is equivalent to the number of bidders, since each bidder may place only one bid.
the only way for developers to obtain new developable land. Developers can obtain land from private sectors. They may decide to enter an auction held by governments based on their own judgment.
This gives us less constraint to test whether investors may acquire information and enter an auction held by the government. The dataset, containing land in Taipei City (the core metropolitan area) and in Taipei County (the suburb), allows us to examine potential differences between bidders in different locations. Taipei City is a core area that attracts more attentions. There will be more information on land auction in Taipei City. On the contrast, it is more costly to collect information on land auctions in Taipei County. I want to test whether location will affect how investors collect information, their decisions to enter an auction, and their bidding strategies.I find that individual bidders’ behavior in Taipei City is very different from those in Taipei County. I first run logistics regressions to test what contribute the success of a land auction. An auction to be successful needs at least one bidder to participate.
Empirical results show that previous auction premium (the ratio of winning price over reserve price), building attached to land or not, land use code for residential dummy, and market return are significant factors to induce at least one investor to bid in Taipei City. Evidences show that investors tend to bid in Taipei City’s land auctions when it is a hot market. However, only the dummy of land use code for residential is the significant factor affecting the participation tendency of investors in Taipei County.
The number of bid regression shows a similar result that more and more investors tend to bid in hot market in Taipei City, but not in Taipei County.
The bid premium regression shows that in Taipei City, the entry of more investors is not related to higher bids placed by these investors, suggesting that even
contrast, the entry regression of Taipei County auction data tells us that number of bids in Taipei County is hard to predict, making the entry of bids have a significantly impact on winning prices. It indicates that bidders in Taipei County’s auctions tend to be uninformed.
My findings add to the literature regarding investors’ bidding behavior in different locations, a core area and a suburb. I find that bids of investors in Taipei City’s land auctions are largely consistent with the predictions of auction theory regarding informed bidders. Investors in Taipei City’s land auctions place their bids based on information. The bids of investors in Taipei County, in contrast, exhibit evidence of unpredictable, and the entry of investors is related to higher bids placed by these investors, a sign of systematic overbidding.
The article is organized as follows: Section 2 lays out the hypotheses that I will be testing. Section 3 describes the institutional features of land auctions in Taiwan.
Section 4 discusses data and methodology used in this paper. Section 5 presents evidence results on entry, while Section 6 presents results on bid premium. Section 7 is the conclusion.
2. Hypotheses
Based on Sherman’s (2005) information production models, investors spend resources on evaluation and choose to enter only if they expect to recover their evaluation and entry costs on average. Sherman’s model predicts that entry decreases with information uncertainty or costs of information production. One important implication of Sherman’s model is that since entry is uncoordinated, there will be ex post fluctuations in entry and in bidders’ bid premium (the ratio of winning price over the reservation price). For an auction to be successful, the auction must at least attract one investor to bid. Therefore, the success of auctions and the number of bids in an
auction will be negatively related to information uncertainty or information cost.
Therefore, I have two hypotheses regarding entry of bidders: