Results on Land Auction Premium

In this section, I examine bid premium, the ratio of the winning price and the reservation price, from land auctions. To see whether bidders optimally shave their bids in different locations, I test Hypotheses 2 by running structural equation models:

entry regression on previous premium and measures of information-uncertainty/

information costs, while bid premium regression on previous premium and measures

of information-uncertainty/ information costs, and entry. We see from Table 2 that the average bid premium in Taipei County’s auction samples is smaller than that in Taipei City. However, this phenomenon does not imply that bidders in Taipei County’s land auctions bid more smartly. We must consider how number of bids affect the bid premium in order to make any judgment. I therefore include the number of bids to estimate the bid premium regression.

I also include the previous auction premium in the bid premium regression. If there is a hot market, bidders tend to bid high and one can expect the bid premium on the current land auction to be positively related to the latest auction premium.

Bid premium regression results are reported in Table 5. On Panel A and Panel C, we see bid premiums are positively related to entry for all samples and samples in Taipei County. This result is consistent as predicted in Hypotheses 2B for unsophisticated uninformed bidders. Based on the hypotheses, the positive coefficient for entry suggests that bidders in Taipei County are uninformed and not bidding optimally. On the contrast, the coefficient on entry in Taipei City’s sample is not significant. Bidders in Taipei City’s land auction tend to collect information and bid optimally.

To evaluate the impact of entry on bid premium in Taipei County, I gauge the economic significance of the entry of bids by computing the change in bid premiums when the entry of bids increases by one standard deviation. I find that a one-standard-deviation (0.8698) increase in entry of bids increases the bid premium by 0.2916 (22.09% of the mean bid premium in Taipei County).

The coefficient on previous auction premium is insignificant. Investors did not follow high previous premium to bid high at the current auction. Regarding the

samples in Taipei City. When the land code is for residential use, bidders tend to bid high. It is also interesting to note that for samples in Taipei County, area has a significantly positive relation with bid premium. As I mentioned above, developers in Taipei County like to obtain land for large scale projects.

I also redefine the bid premium as the ratio of AVERAGE bid price over the reservation price. The average bid price include information of those losing bids. As a robustness check, I use this new definition of premium to run the regression. Results are the same.

7. Conclusions

Empirical results indicate that bidders in Taipei City’s land auctions are generally consistent with the predictions for rational informed bidders. Higher previous premium and higher market return may attract more bidders to bid in Taipei City’s auctions. However, these bidders tend to shave their bids optimally. For Taipei City samples, entry of bids is not related to bid premium. Bidders in Taipei City’s auctions did not overbid by paying too high premium. They tend to spend resources to collect information and shave their bids optimally.

On the contrast, bidders in Taipei County’s land auction show a different behavior. For Taipei County samples, bidders’ entry decisions cannot be predicted by previous auction premium or by market return. Entry of bids, therefore, has a significantly positive impact on current auction’s bid premium. Bidders in Taipei County’s land auctions tend to be uninformed and overbid.

Empirical evidences suggest that investors should acquire information in order to avoid overbidding. Sophisticated professional investors can make use of their professional expertise and extensive resources to evaluate a tendered land. Without information, investors tend to participate in a land auction on the spur of the moment.

Entry of these uninformed investors causes investors to overbid, i.e. high winning price.

In a core metropolitan area, like Taipei City, competition for acquiring a land is intensive. Although intensive competition may induce more investors to participate in a land auction, investors tend to expend resource to collect information and to form an optimal bidding strategy. On the contrast, in a suburb area, like Taipei County, competition is not so intensive. Entry of bids becomes less predictable. Shock of bid entry causes overbidding.

References

Ashenfelter, Orley and Genesove, David. (1992). Testing for Price Anomalies in Real-Estate Auctions, American Economic Review, vol. 82(2), 501-05,

Chiang, Yao-Min, Qian, Yiming, and Sherman, Ann. (2010). Endogenous entry and partial adjustment in IPO auctions: Are institutional investors better informed? Review

of Financial Studies, Vol.23, No.3, pp.1200-1230.

Engelbrecht-Wiggans, R., and E. Katok. (2005). Experiments on Auction Valuation and Endogenous Entry. In J. Morgan (ed.), Behavioral and Experimental Economics, pp. 171–96. Stamford, CT: Elsevier Science.

Giliberto M. and Varaiya, N. (1989). The winner’s curse and bidder competition in acquisitions: Evidence from failed bank auctions. Journal of Finance, 44 (1):59–75.

Hsueh, L, Li, C. and Tseng, H. (2002). The Population Migration in Taiwan, and its Causal Relationship with Labor Market and Housing Market, International Real

Estate Review, 2002, vol. 5, issue 1, pages 61-90

Lusht K.M. (1996). A comparison of prices brought by English auctions and private negotiations. Real Estate Economics, 24(4): pp.517-530.

Pagan, A. (1984). Econometric Issues in the Analysis of Regressions with Generated Regressors. International Economic Review 25:221–47.

Sherman, A. (2005). Global Trends in IPO Methods: Bookbuilding versus Auctions with Endogenous Entry. Journal of Financial Economics 78:615–49.

Table 1 Summary statistics of land auctions

The sample includes five-hundred-ninety-seven land auctions in Taiwan during 2007-2010. Land is tendered either in Taipei City (the core area) or in Taipei County (a suburb). Number of bids is the number of bids submitted by investors in an auction. Bid premium is the winning price relative to the reservation price. Housing price index returns are calculated based on Taipei City Housing Index, and Taipei County Housing Index provided by Sinyi Realtor. Panel B shows by year the mean, median (in parentheses), and standard deviation (in square brackets) of each variable.

Panel A,

Variable N Mean Median Std Dev Minimum Maximum

Number of bids 295 6.91 4.00 7.95 1.00 52.00

Bid premium 295 1.50 1.25 1.84 1.00 32.09

Land area (1,000m²) 597 373.23 112.74 832.14 1.00 8,812.00 Land area + floor area (1,000m²) 594 431.24 173.79 837.12 1.00 8,812.00

Housing price index return (%) 597 2.22 2.09 4.04 -4.83 8.31

Panel B,

Year 2007 2008 2009 2010

Mean of number of bids 6.26 6.29 7.19 8.72

(3.00) (4.00) (3.00) (3.00) [7.10] [6.70] [8.75] [9.67]

Mean of bid premium 2.73 1.42 1.35 1.50

(1.28) (1.28) (1.20) (1.27)

[6.42] [0.42] [0.42] [0.61]

Mean of land area (1,000m²) 265.87 499.01 307.55 262.06 Mean of (land area + floor area) (1,000m²) 348.89 537.80 372.87 335.72 Mean of housing index return (%) -0.33 0.22 4.47 2.74

Number of successful auctions 23 119 124 29

Number of total auctions 61 229 261 46

Table 2 Summary statistics of bids by locations

The sample includes five-hundred-ninety-seven land auctions in Taiwan during 2007-2010. Land is tendered either in Taipei City (the core area) or in Taipei County (a suburb). Number of bids is the number of bids submitted by investors in an auction. Bid premium is the winning price relative to the reservation price. Housing price index returns are calculated based on Taipei City Housing Index, and Taipei County Housing Index provided by Sinyi Realtor. I use a paired t-test for differences in means, and the Wilcoxon signed rank test for differences in medians between groups. ***, **, and * denote that the difference is significant at the 1%, 5%, and 10% levels, respectively.

Variables Taipei City Taipei County Difference

Number of bids 7.12 6.41 0.71

(4.00) (3.00) 1.00

Bid premium 1.58 1.32 0.25

(1.26) (1.17) 0.09**

Land area (1,000m²) 314.19 486.97 -172.78**

(116.00) (95.42) 20.58

Land area + floor area (1,000m²) 376.09 539.06 -162.97**

(191.00) (171.13) 19.87

Housing price index return (%) 2.72 1.24 1.18***

(4.23) (0.78) 3.45***

Number of successful auctions 205 90 115

Number of total auctions 393 204 189

Table 3 Logistics regression of auction success

The dependent variable is a dummy equal to one if a land auction is successful. Previous auction premium is the premium of last auction in the same location. Land area is measured in one thousand square meters. Building dummy equals 1 if there is building attached to the land and 0 otherwise.

Residential dummy equals 1 if the land use code is for residential use and 0 otherwise. Housing price index returns are calculated based on Taipei City Housing Index, and Taipei County Housing Index provided by Sinyi Realtor. Taipei dummy equals 1 if the auctioned land is in Taipei City and 0 otherwise. t-statistics are adjusted using White’s correction for heteroskedasticity. כ, ככ, and כככ denote significance at the 1%, 5%, and 10% levels, respectively.

All Samples Taipei City Taipei County Estimates χ2 Estimates χ2 Estimates χ2 Previous auction premium 0.6241 6.68*** 0.6977 4.61 ** 0.5311 2.09

Land area 0.1750 2.50 0.1468 0.76 0.1779 1.44

Building dummy 0.2276 1.53 0.5144 5.23 ** -0.3883 1.38 Residential dummy 0.8465 17.12*** 0.8564 10.79 *** 0.9861 7.43 ***

Housing price index return 0.1105 23.20*** 0.1233 21.25 *** 0.0617 1.65 Taipei city dummy -0.1116 1.43

Intercept -1.9075 23.62*** -2.0642 15.56 *** -1.7263 7.82 ***

-2 Log likelihood 751.973 *** 483.95*** 259.866 **

Pseudo R² 7.8089 7.3427 11.2145

N 583 383 200

Table 4 Entry regression

The dependent variable is the natural logarithm of the number of bids in an auction. Previous auction premium is the premium of last auction in the same location. Land area is measured in one thousand square meters. Building dummy equals 1 if there is building attached to the land and 0 otherwise.

Residential dummy equals 1 if the land use code is for residential use and 0 otherwise. Housing price index returns are calculated based on Taipei City Housing Index, and Taipei County Housing Index provided by Sinyi Realtor. Taipei dummy equals 1 if the auctioned land is in Taipei City and 0 otherwise. t-statistics are adjusted using White’s correction for heteroskedasticity. כ, ככ, and כככ denote significance at the 1%, 5%, and 10% levels, respectively.

All Samples Taipei City Taipei County Variable Estimatest-value Estimatest-value Estimatest-value Previous auction premium 0.1483 1.27 0.3201 1.95* 0.0087 0.07 Housing price index return 2.3535 1.70* 2.2643 1.56 3.5754 0.90 Land area 0.1861 2.01** 0.3181 3.02*** 0.1015 1.10 Taipei dummy -0.0265 -0.23

Building dummy 0.3627 3.36*** 0.4627 3.86*** 0.0875 0.38 Residential dummy 0.4014 3.56*** 0.3479 2.72*** 0.4988 2.26**

Intercept 0.8313 4.08*** 0.5144 1.89* 1.0493 4.12***

R² 0.1007 0.1319 0.0821

N 292 205 87

Table 5 Premium regression

The dependent variable is the bid premium for each auction, calculated as the winning price divided by the reservation price. Previous auction premium is the premium of last auction in the same location.

Entry of bids is the residuals from the entry regressions. Land area is measured in one thousand square meters. Building dummy equals 1 if there is building attached to the land and 0 otherwise. Residential dummy equals 1 if the land use code is for residential use and 0 otherwise. Housing price index returns are calculated based on Taipei City Housing Index, and Taipei County Housing Index provided by Sinyi Realtor. Taipei city dummy equals 1 if the auctioned land is in Taipei City and 0 otherwise. כ, ככ, and כככ denote significance at the 1%, 5%, and 10% levels, respectively. Unstandardized parameter estimates retain scaling information of variables and can only be interpreted with reference to the scales of the variables. Standardized parameter estimates are transformations of unstandardized estimates that remove scaling and can be used for informal comparisons of parameters throughout the model.

Standardized path coefficients with absolute values less than .10 may indicate a “small” effect. Values around .30 indicate a “medium” effect. Values greater than .50 indicate a “large” effect.

Panel a, all sample

Log of number of bids Bidding premium Variables Standardized

estimate estimate t-value Standardized

estimate estimate t-value

Log of number of bids 0.1180 0.2451 1.9849 **

Previous auction premium 0.0649 0.1483 1.1626 0.0331 0.1571 0.5833 Housing price index return 0.1018 0.0235 1.7606* -0.1031 -0.0495 -1.7486 * Land area 0.1705 0.1861 2.9659*** 0.1979 0.4486 3.3441 ***

Taipei city dummy -0.0136 -0.0265 -0.2317 0.0852 0.3443 1.4302 Building dummy 0.2036 0.3627 3.4493*** 0.0814 0.3012 1.3326 Residential dummy 0.1801 0.4014 3.2340*** 0.0388 0.1799 0.6758

Table 5 Premium regression (continued)

Panel B, Taipei City

Log of number of bids Bidding premium Variables Standardized

estimate estimate t-value Standardized

estimate estimate t-value

Log of number of bids 0.0480 0.1180 0.6854

Housing price index return 0.1038 0.0226 1.5669 -0.0895 -0.0480 -1.3418 Previous auction premium 0.1122 0.3201 1.7070* -0.0620 0.4348 0.9360 Land area 0.2059 0.3181 3.0784*** 0.3288 1.2485 4.8031***

Building dummy 0.2588 0.4627 3.8636*** 0.1303 0.5727 1.8773*

Residential dummy 0.1565 0.3479 2.3825*** 0.0315 0.1719 0.4721

Panel C, Taipei County

Log of number of bids Bidding premium Variables Standardized

estimate estimate t-value Standardized

estimate estimate t-value

Log of number of bids 0.7607 0.335310.1682***

Housing price index return 0.1107 0.0358 1.0354 -0.0617-0.00878 -0.8267 Previous auction premium 0.0052 0.0087 0.0494 -0.0218 -0.0161 -0.3010 Land area 0.1360 0.1015 1.2522 -0.0482 -0.0159 -0.6339 Building dummy 0.0446 0.0875 0.4112 -0.1733 -0.1498 -2.2994***

Residential dummy 0.2230 0.4988 2.0809** -0.1189 -0.1172 -1.5605