Printer Friendly

The behavior of inexperienced bidders in Internet auctions.

I. INTRODUCTION

As has now been documented in many studies, the Internet auction market has quickly become a thriving marketplace. In the United States, the market is dominated by eBay. In the second quarter of 2007 alone, eBay users posted a total of 559 million listings, resulting in gross merchandise value (the total amount that all goods sold for) of $14.46 billion. (1) This impressive success has occurred despite the fact that bidders face substantial risks because of the nature of the market. Sellers by convention do not send goods to winning bidders until after they have received payment. The seller can then simply pocket the money or send an item of poor quality. A widespread and widely publicized example of this behavior occurred in 2002 when a seller with a history of over 6,000 properly conducted transactions sold hundreds of porcelain collectibles on 4 January 2002 but did not send the winners anything after receiving payments of about $300,000. (2) A consumer who is defrauded by a seller has little recourse because the identity of a seller is known only through an e-mail address, which can be anonymously obtained. (3)

To solve this problem, rather than formally enforcing contracts, eBay has established an innovative reputation system, which allows winning bidders to post ratings of sellers' actions that are publicly viewable (and vice versa). The ratings are classified as positive, neutral, or negative. This information is also presented as a feedback rating that is equal to the number of positive reports received from unique trading partners minus the number of negative reports received from unique trading partners. (4) Potential bidders can use this information to form expectations about how the seller will behave in the future and reward sellers who develop good reputations by bidding more and more often in their auctions.

The seller's benefits to a good reputation can come in two forms: an auction can be more likely to attract bidders and thus result in a sale and bidders can be willing to bid larger amounts. Most studies do find that auctions are more likely to end in a sale if the seller is more reputable; however, the results regarding the amount by which bids increase as the seller's reputation improves have been quite mixed: some find that bid amounts increase substantially as the seller's reputation improves, some find extremely small increases in bid amounts, and some find no increase in bid amounts at all. (5) The prevalence of studies that show very small changes in bidder behavior as the seller's reputation improves has led to an apparent consensus that the returns to reputation in Internet auctions are indeed quite small. For example, in the September 2007 issue of the Journal of Economic Literature, MacLeod (2007) refers to Bajari and Hortacsu's survey and notes that "the results are rather consistent. Mikhail I. Melnik and James Alto (2002), Paul Resnick and Richard Zeckhauser (2002), and Luis Cabral and Hortacsu (2004) all find that there is a positive relationship between price and quality (of the seller's reputation). However, the overall effect appears to be small."

Despite this consensus, recent studies favor the hypothesis that there is a significant increase in bid amounts as the seller's reputation improves. Melnik and Aim (2005) find that the returns to an improved seller reputation increase when there is more uncertainty about product quality. Livingston (2005) shows that sellers are strongly rewarded for the first few reports that they have behaved honestly, but marginal returns to additional reports are severely decreasing. He notes that other studies may have found small increases in bid amounts because they typically assume that the relationship between the price and the number of positive reports received by the seller is linear or log linear. If marginal returns to reputation are severely decreasing, these functional forms may only pick up the small returns that occur after an initial reputation is established.

The primary analysis of this paper examines another reason why past studies may have underestimated the degree to which the seller's reputation is reflected in the bid amounts of at least some bidders. Bidders who have more experience in the Internet auction market may value the seller's reputation more highly than bidders who have little or no experience in the market. This may be reflected in differences in the degree to which the seller's reputation is reflected in the bid amounts of experienced and inexperienced bidders. Bidders receive reports from sellers as well; I follow Garratt, Walker, and Wooders (2004) in using the bidder's feedback rating as a measure of the bidder's experience. Since almost all reports received by bidders are positive, a bidder's feedback rating is almost always identical to the number of positive reports that the bidder has received, and each of these reports represents an auction where the bidder completed a transaction with an ebay seller. I examine bidder behavior in auctions for two products of different values: Taylor Made Firesole Irons, a variety of golf club, which sold for an average of $393.31, and HALO video game discs, which sold for an average of $22.97.

The results for both products are quite clear: "inexperienced" bidders bid just as highly in the auctions of sellers with no transaction history as they do when the seller has a history of hundreds or even thousands of properly conducted transactions. However, across both goods, experienced bidders bid substantially less if the seller has no transaction history than they do when the seller has established a reputation for honesty. For example, using data from the auctions of the Taylor Made Firesole Irons golf clubs, I find that the most experienced bidders (with feedback ratings of over 22) bid $73.21 less (about 19% of the average sale price) in the auctions of sellers who have zero positive reports than in the auctions of sellers who have anywhere from 1 to 31 positive reports.

Experienced and inexperienced bidders also differ in their decision of when to place their final bid. While eBay auctions are similar to traditional English auctions where bidders incrementally raise their bids until all but one bidder stops, there is one key difference: the auctions have a set ending time. Roth and Ockenfels (2002) argue clearly that rational bidders should wait until the final moments of an auction to place their bids. (6) Wilcox (2000) and Ockenfels and Roth (2006) show that bidders with more experience are more likely to wait until the closing moments of an auction to bid. Using the complete bid histories from each sample of auctions, I extend these results by showing that bidders learn the advantages of waiting to bid until near the auction close very quickly: in both samples, bidders who have participated in just a handful of eBay transactions place their final bid over 9 h closer to the end of the auction than bidders who have no transaction history. Also, these bidders with a very small amount of experience are substantially more likely to place their final bid in the closing minutes of the auction than bidders with no feedback profile. Bidders with larger amounts of experience, however, bid no closer to the end and are no more likely to place bids in the final moments than bidders with a small but positive amount of experience.

The paper is organized as follows. Section II describes the data. Section III analyzes whether bidders of different experience levels bid differently depending on the seller's reputation. Section IV argues that the results of Section III are robust to a potential econometric problem. Section V examines whether bidders wait longer to place their bids if they are more experienced. Section VI concludes.

II. DATA

Data were collected from auctions of two different products: Taylor Made Firesole Irons, a variety of golf clubs (hereafter referred to as the golf clubs sample), and "HALO: Combat Evolved," a video game played on Microsoft's Xbox gaming console (hereafter referred to as the video game sample). While these goods have very different market values, they are also similar in important respects: they each are goods that provide entertainment to the consumer and are likely to at least have a private value component. This is important because it means that the auctions are roughly strategically equivalent to private value second-price sealed-bid auctions, and we can regard the amount of the recorded highest bid as being equal to the second-highest bidder's valuation of the good. (7)

The data were collected at different times. There are 861 observations of Taylor Made golf club auctions that ended between October 20, 2000, and August 20, 2001, and 841 observations of auctions of the HALO video game that ended between January 6, 2004, and January 25, 2004. The key independent variable, the second-highest bidder's feedback rating, is not available if fewer than two bids were placed. This forces us to drop 273 observations from the golf clubs sample and 244 observations from the video game sample where fewer than two bids were placed, the consequences of which will be addressed in the next section. Also, in the golf clubs sample, 48 auctions that did receive at least two bids ended using eBay's Buy It Now option, which allows sellers to post a price in conjunction with the auction that a buyer can immediately agree to pay. (8) If a buyer accepts this price, the auction ends immediately. When this occurs, the price is no longer equal to the second-highest bidder's willingness to pay--all we know is that the value of at least one bidder is greater than or equal to the Buy It Now price. (9) These 48 observations are therefore also discarded from the golf clubs sample. Observations where a secret reserve price was used but not met, so the auction did not result in a sale, are not dropped since they still contain all the information we need. (10) This leaves us with 540 observations in the golf clubs sample and 597 observations in the video game sample.

Table 1 presents summary statistics for the variables used in the first portion of this study, which analyzes how bidders of different experience levels alter their bids depending on the seller's reputation. The unit of observation is a single auction. The dependent variable is the price paid by the winning bidder. On average, the winner paid $393.31 for the golf clubs and $22.97 for the video games.

The first critical explanatory variable is the reported history of the seller. Since a seller who ruins his reputation by cheating can start over with a new identity, I examine the effect of reputation by looking at how bidders reward sellers who gain more positive reports relative to sellers who have yet to establish a trading history. Livingston (2005) shows that the relationship between bid amounts and positive reports is highly nonlinear. The marginal returns to reputation turn out to be severely decreasing because bidders need to see only a few positive reports in order to be convinced of a seller's honest intentions. Once they are convinced, further positive reports have little impact on bid amounts because bidders are already bidding nearly what they would be willing to pay if there were no risk involved in the transaction. In order to capture the nonlinear nature of this relationship, for each good, the sample distribution of the number of positive reports held by the seller in each auction is divided into quartiles, and dummy variables are created that indicate whether an auction falls into each quartile. The first quartile is further divided by splitting off auctions where the seller has zero positive reports. The remainder of the first quartile will still be referred to as the first quartile, though the reader should keep in mind that this group excludes auctions where the seller has zero positive reports. The five categories have somewhat different cutoff points in each sample since the distribution of seller reputations is different for each product. However, in each sample, sellers have few reports in most of the auctions. The auction at the 25th percentile has a seller with only 31 positive reports in the golf clubs sample and only 34 positive reports in the video game sample. The other quartiles cover much broader ranges of positive reports received in each sample.

Negative reports are also included in the empirical analysis. Each specification controls for the fraction of reports that a seller has received that are negative. I do not use the same specification for negative reports as I do for positive reports because there are very few negative reports in the samples. The mean fractions of negative reports are 0.02 and 0.01 in the golf clubs and video game sample, respectively, and the standard deviations of this fraction are only 0.06 and 0.03, respectively, in the two samples.

The other critical explanatory variable is the experience level of the second-highest bidder. Both the seller and the winning bidder can leave reports about how their trading partner behaved, so the bidder's feedback rating can be used as a proxy for bidder experience. This number may not be equal to the number of eBay transactions that a bidder has participated in for several reasons. First, there may have been some auctions in which the bidder was the winner where the seller did not leave a report about the behavior of the bidder. Second, a bidder might be using a different identity than the identity she used in her past eBay transactions. Third, inexperienced bidders might get advice from more experienced eBay users, and experienced buyers might use their identities to buy for inexperienced friends. Finally, the bidder may have received some negative feedback, reducing her feedback rating by 1. (11)

Since the highest recorded bid is equal to the second-highest bidder's expected valuation of the good, the relevant experience level is that of the second-highest bidder. These bidders are divided into five categories: bidders with a feedback rating of 0 make up one category. Bidders who have positive ratings (in absolute value) are then split into quartiles. In the golf clubs sample, bidders in Quartile 1 have a feedback rating of 1-2, bidders in Quartile 2 have a rating of 3-5, bidders in Quartile 3 have a rating of 6-22, and bidders in Quartile 4 have a rating of more than 22. In the video game sample, bidders in Quartile 1 have a feedback rating of 1-5, bidders in Quartile 2 have a feedback rating of 6 16, bidders in Quartile 3 have a feedback rating of 17-44, and bidders in Quartile 4 have a feedback rating of more than 44. The second-highest bidders in the golf clubs sample are more inexperienced than their counterparts in the video game sample. For example, 28% of the second-highest bidders in the golf clubs auctions, but only 5% of the second-highest bidders in the video game auctions, have a feedback rating of 0.

The price regressions also control for several auction-specific characteristics. The regressions on each sample control for a subset of the variables listed in Table 1. The golf clubs can have several differences that have a substantial impact on the value of the good, so regressions using that sample control for the retail price of the good. Sellers in both samples charge varying amounts for shipping. (12) Both products can be new or used, so a dummy variable indicating whether the good is new is included in all specifications. Other controls common to regressions on both samples include a dummy variable indicating whether the seller accepts credit cards as a form of payment, a series of dummy variables indicating the duration of the auction (either 3, 5, 7, or 10 d in the golf clubs sample; auctions could also run for only 1 d in the video game sample), and a dummy variable indicating whether the seller used a hidden reserve price. (13) The golf club regressions include dummy variables indicating whether the clubs are left-handed, designed for women, and designed for seniors. Finally, the video game regressions occurred a few years later than the golf club auctions, and several new features were added to eBay auctions in that period. Sellers can be classified as "Power Sellers" if they sell a high enough volume and maintain at least 98% positive feedback, they can be labeled as "ID Verify" if they pay for a service that confirms their identity by crosschecking their contact information using consumer and business databases, and they can display a link by their eBay ID to their "About Me" page, where they can give potential bidders more information about themselves. Each of these features may reassure bidders that the seller is more trustworthy, encouraging them to bid larger amounts.

Table 2 presents summary statistics for the variables used in the second portion of this study, which examines whether more experienced bidders place their final bids closer to the end of the auction and how much experience bidders gain before they time the placement of their bids more strategically. The unit of observation is a bidder's final bid in an auction. Attention is again restricted to auctions that did not end prematurely when a buyer accepted the Buy It Now price since the buyer changes the end time in such auctions and the winning bid by definition occurs at the moment the auction ends. In these auctions, there were 4,377 unique bidders in the golf clubs sample and 4,394 unique bidders in the video game sample.

The dependent variables are the number of minutes left in the auction when the bidder's final bid was placed and dummy variables indicating whether a bidder's final bid was placed in the last 30 s, 1 min, 5 min, and 10 min of the auction. While the average bid was placed quite a long time before the end of the auction (over 3,500 min in the golf clubs sample and over 1,400 min in the video game sample), a significant portion was placed near the end of the auction. Three percent of the bids in the golf clubs sample and 5% of the bids in the video game sample were placed with 30 s or less remaining in the auction, 4% in the golf clubs sample and 8% in the video game sample were placed in the final minute, 8% in the golf clubs sample and 12% in the video game sample were placed in the final 5 min, and 10% in the golf clubs sample and 14% in the video game sample were placed in the final 10 min.

The key explanatory variable is the bidder's feedback rating. Ockenfels and Roth (2006) use a linear specification to study the relationship between bidder experience and the probability that a bid is placed in the closing moments of an auction. The same specification is examined here, but I also allow for the possibility that the effect of experience is nonlinear by breaking the bidder ratings into five categories as before: bidders with a rating of 0 make up the first category. The remaining bidders who have positive ratings (in absolute value) are again split into quartiles. As a whole, the bidders in the golf clubs sample are fairly inexperienced: 31% of the bidders have a feedback rating of 0. Among bidders with experience, the first quartile includes bidders with a feedback rating of 1 or 2. The second quartile includes bidders with a rating of 3-5. The third quartile includes bidders with a rating of 6-21. The fourth quartile includes bidders with a rating above 21. The bidders in the video game sample are somewhat more experienced. Seven percent of the bidders have a rating of 0. Among bidders with experience, the first quartile includes bidders with a feedback rating of 1-5. The second quartile includes bidders with a rating of 6-16. The third quartile includes bidders with a rating of 17-49. The fourth quartile includes bidders with a rating above 49. Ockenfels and Roth also control for the number of bidders participating in the auction. Over nine bidders participated in an auction on average in the golf clubs sample, and over eight bidders participated in an auction on average in the video game sample.

III. MARKET EXPERIENCE AND THE SELLER'S REPUTATION

The primary analysis of the paper examines whether bidders who have different amounts of experience transacting using eBay auctions weigh the seller's reputation differently. Since eBay auctions are strategically similar to sealed-bid second-price auctions, the recorded price is equal to the second-highest bidder's valuation of the good, and the relevant experience level is that of the second-highest bidder.

As discussed in Section II, we are not able to use observations where fewer than two bids were placed. This presents an obvious sample selection problem, which will be addressed in the next section. For now, it is left in the background. The relationship between a seller's reputation and the willingness to pay of the second-highest bidder is estimated using ordinary least squares (OLS). The model has the following form:

(1) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

where [P.sub.i] is the price paid by the winning bidder in auction i; EXP is a vector of dummy variables [EXP.sup.Q1], [EXP.sup.Q2], [EXP.sup.Q3], and [EXP.sup.Q4], which indicate the feedback rating quartile of the second-highest bidder, leaving bidders with a rating of 0 (indicated by [EXP.sup.0]) as the omitted category; POS is a vector of dummy variables that indicate the quartile of the number of positive reports held by the seller, leaving sellers with zero positive reports as the omitted category; X is a vector of the other auction characteristics discussed in Section II; and [[epsilon].sub.i] is normally distributed with mean 0, but the variances are allowed to be correlated; robust standard errors are calculated.

The model is estimated separately for each good since there are different auction characteristics that need to be controlled for in each sample. Consider first the results for the more expensive good, the golf clubs. Inexperienced bidders bid more on average than experienced bidders, and they bid the same amount regardless of the reputation of the seller. Experienced bidders, on the other hand, bid substantially less if the seller has little or no reported history of honest behavior. Once the seller has a fully established reputation, there is little difference between the bids of inexperienced and experienced bidders.

Table 3 reports a specification that controls for the seller's reputation but not the bidder's experience. Bidders as a whole do bid larger amounts when the seller begins to establish a reputation, but there are severely decreasing returns to positive reports. Sellers in the first quartile, who have 1-31 positive reports, receive prices that are $24.06 higher than sellers who have 0 positive reports. Sellers in the fourth quartile, who have more than 836 positive reports, receive prices that are $29.81 higher than sellers who have 0 positive reports. So, 1-31 positive reports increase the price the second-highest bidder is willing to pay by about 6.1%, of the average price of the good, but the seller needs to receive around 800 additional reports for the price to increase by an additional 1.5%. Also, while the difference between the price in auctions where the seller has 1-31 reports (Quartile 1) and the price in auctions where the seller has 32-205 reports (Quartile 2) is statistically significant at the 10% level, the price differences between auctions in any other pair of quartiles are not statistically significant, suggesting that returns to positive reports are in fact severely decreasing. There are at least two explanations for these findings. Bidders might bid more as the seller's reputation improves because they think that it is more likely that the seller will deliver what they promise. Alternatively, sellers with more positive reports (and therefore more experience) might have learned how to list their auctions more effectively. For example, they might set reserve prices close to their optimal level, and they might market the item more effectively in their description of the good. (14)

The specification in Table 4 adds controls for the experience of the bidders. Bidders are divided into five categories: those who have a feedback rating of 0 (the omitted category) and quartiles of bidders who have positive feedback ratings (in absolute value): those who have a rating of 1-2, those who have a rating of 3-5, those who have a rating of 6-22, and those who have a rating of more than 22. Column 1 reports the coefficient estimates for the bidder experience quartile indicators. Columns 2-6 report the coefficient estimates for the interactions between the bidder experience categories and the seller positive report quartiles, so each column shows how a different group of bidders reacts to a seller's reputation. The results are clear: inexperienced bidders bid relatively high amounts, and they bid just as much in the auctions of new sellers as they do in the auctions of sellers who have an established reputation for honesty. Bidders with ratings of 0 and bidders in Quartiles 1 and 2, who have ratings of only 1-5, bid the same amounts on average. None of these three groups raise their bids as the seller gains more positive reports.

As bidders gain experience, they become less willing to bid high amounts unless the seller has at least begun to gain positive reports. The point estimate suggests that bidders in Quartile 3, with a rating of 6-22, bid $41.49 less than bidders with a rating of 0, an amount that is about 11% of the average price, though this estimate is not statistically significant. The most experienced bidders, who have ratings over 22, bid $69.69 less than bidders with a rating of 0. Each of these more experienced groups then raises their bids if the seller has a track record of acting honestly. Bidders in Quartile 3 bid $35.12 more if the seller has 1-31 positive reports, and $33.70 more if the seller is in the top quartile of positive reports, than they do if the seller has 0 reports (though these estimates are not statistically significant). The most experienced bidders in Quartile 4, with ratings of more than 22, bid $73.21 more if the seller has 1-31 positive reports and $79.88 more if the seller has more than 836 reports than they do in the auctions of sellers who have yet to gain any positive reports. While each of these groups rewards sellers who have a history of acting honestly, there are severely decreasing marginal returns to positive reports. In fact, there is no statistical difference between the bids of the most experienced bidders in the auctions of sellers who have only 1-31 reports and their bids in the auctions of the sellers in any other positive report quartile.

The results are similar for the sample of auctions of the HALO: Combat Evolved video game despite the fact that this product is worth substantially less than the golf clubs selling for an average of only $22.97. The regressions are again estimated by OLS. The specification reported in Table 5 shows that bidders do substantially increase their bids when the seller receives a few positive reports, but further reports once again have little impact since bidders are already convinced of the seller's honest intentions. Sellers who have 1-34 reports receive bids that are $2.13 higher than sellers who have 0 reports, an amount that is about 9% of the average price in the sample, though this estimate narrowly misses being statistically significant at the 90% level. There is no statistical difference between the bids received by the Quartile 1 sellers and sellers in the other positive report quartiles, however, suggesting that all the gains to reputation accrue to the first few reports.

Table 6 investigates how bidders of different experience levels react to a seller's reputation. The pattern of results is the same as with the golf clubs--inexperienced bidders bid similar amounts regardless of the seller's reputation, while more experienced bidders are willing to bid their full value only if the seller has received positive reports. Column 2 reports the results for bidders with a feedback rating of 0. It is difficult to evaluate whether these bidders adjust their bids as the seller establishes a reputation. There are only 30 observations where the second-highest bidder had a feedback rating of 0, and among these, there are no observations where the seller had zero positive reports. Thus, we can only compare bids when the seller has at least some positive feedback, and even then, the hypothesis tests have little power. The highest positive report quartile is used as the reference group for the bidders with a rating of 0. For the most part, the results indicate that these bidders do not change their bids as the seller's reputation improves--while sellers in the second quartile receive higher bids than the other sellers, these new bidders do not bid any more or less when the seller has 1-34 positive reports than they do when the seller has more than 1,422 positive reports. However, since there are so few observations where the bidder has a rating of 0, the conclusions of these tests are not very reliable.

The other relatively inexperienced bidders (who have a rating of 1-5) do not increase their bids as the seller's reputation improves--in fact, as reported in Column 3, there is some evidence that they decrease their bids when the seller's reputation improves. However, the point estimates do suggest that some bidders with a rating of 6-16 are starting to bid less when the seller has yet to establish a reputation though not statistically significant, the estimates suggest that they bid $2.25 less than bidders with a rating of 0, as reported in Column 1, and they raise their bids somewhat if the seller has at least a few positive reports, as reported in Column 4.

The most experienced bidders bid less on average than bidders with no experience. However, this difference disappears as the seller gains positive reports. Column 1 shows that bidders in Quartile 3 bid $7.08 less than bidders with a rating of 0, but they raise their bids by $7.03 when the seller has 1-34 positive reports relative to when the seller has zero positive reports, as shown in Column 5. Similarly, bidders in Quartile 4 bid $4.61 less than bidders with a rating of 0, as shown in Column 1, but they raise their bids by $4.62 when the seller has 1-34 positive reports relative to when the seller has zero positive reports, as shown in Column 6. However, their bids do not increase any further as the seller gains more positive reports. Again, for the bidders in Quartiles 3 and 4, there is no statistical difference between the bids received by sellers in the first quartile and sellers in the higher quartiles of positive reports received.

Overall, the pattern of the results is clear: inexperienced bidders bid the same relatively high amount regardless of the seller's reputation, while experienced bidders only bid higher amounts once the seller receives at least a few positive reports. There are several possible explanations why this might occur. First, inexperienced bidders might not understand the potential risks involved in Internet auctions. In general, in a second-price auction format, bidders should bid their expected value for the good, which takes into account their assessment of the probability that the seller will fail to deliver the item or there will be something wrong with the product. Inexperienced bidders may have incorrect prior beliefs about the probability of being cheated (perhaps because they do not believe that there are very many dishonest sellers out there). If so, they would bid what the item is worth to them regardless of whether the seller has proven himself to be honest. Experienced bidders, who may have become more aware of the potential risks after having bad experiences with low-reputation sellers (or hearing about the bad experiences of others as they get more involved in the market), would only bid what the item is worth after the seller has proven himself to be worthy of the bidders' trust. Once there is little doubt in bidders' minds about the security of the transaction, both types of bidders will pay what the item is worth to them and the price that prevails in the market will be similar regardless of the ratio of inexperienced to experienced bidders.

Second, inexperienced bidders may actually be aware of the risks. However, since they only occasionally make purchases on eBay, they might decide that these risks are outweighed by the costs of gathering information about the seller's reputation. Their decision to ignore the seller's reputation and bid the same amount regardless of the seller's profile therefore might be quite rational.

Third, eBay is a very active market where many auctions of the same good are often running at the same time, and if a bidder fails to win one auction, another auction of the same good will likely be available shortly thereafter. If experienced bidders are more aware of the thickness of the market, they may be willing to wait for a different auction of the same good from a more reputable seller since they know that the wait will be a short one. Experienced bidders would accordingly bid much less in the auctions of low-reputation sellers since they know that better opportunities will be coming soon. These bidders might even be resellers who patiently wait to win auctions at low prices in order to preserve a potential profit margin. Inexperienced bidders might bid more aggressively regardless of the seller's reputation since they may be unaware that other opportunities to obtain the item they want will be available soon.

Fourth, bidders with low feedback ratings may actually be alternate identities of the seller who are submitting shill bid--bids that are intended solely to raise the price in the auction. Kauffman and Wood (2000) present evidence that such behavior may occur in collectable coin auctions. They identify bids where the bidder could have placed a bid of the same amount or of a lower amount in a different auction of the same item that was closer to the auction's end and find such bids in about 5% of the auctions in their sample. If inexperienced bidders are shill bidders, they may have inflated bids as a result.

Fifth, as noted above, Roth and Ockenfels (2002) argue that inexperienced bidders tend to continually raise their bids to maintain their status as high bidder because they do not understand the incentives present on eBay that encourage bidders to place last-minute bids. As a result, inexperienced bidders may pay more when they win an auction.

Finally, sellers who have few positive reports are also likely to be inexperienced, and such sellers may be less likely to accept payment using Paypal. Paypal is a service that makes it easy for sellers to accept payments by credit card and for bidders to send such payments to the seller. If so, the transaction cost of trading with inexperienced sellers will be higher, so we would expect sellers with few positive reports to receive lower bids. This effect may be stronger for experienced bidders, who may account for such costs more carefully than inexperienced bidders.

IV. ECONOMETRIC CONCERNS

A potential selection problem with the above results exists since the analysis uses only observations where at least two bids were placed, and bidders choose whether to place bids, thus endogenously determining whether an observation makes it into the usable sample. To address this issue, the equations are reestimated using a Heckman sample selection model, as in Livingston (2005). Ideally, the selection equation in such a model would control for the bidder's experience level because we might expect that inexperienced bidders would be less reluctant to place bids in the auctions of sellers with weaker reputations. However, this is not possible since information about the second-highest bidder's experience level is not available when only one bid or zero bids are placed. Instead, the selection equation includes the positive report quartile indicators, the set of auction characteristics controlled for in our original estimations, and controls for the minimum bid allowed by the seller. (15)

For the sake of brevity, the mathematics of the model and the results of the estimations are omitted here, but they are available from the author by request. For both goods, the qualitative results are unchanged, and the point estimates are very similar to the estimates of the model reported in the previous section. In each case, the estimation of the selection equation shows that auctions in successively higher quartiles of the minimum bid-to-retail price ratio are less likely to receive at least two bids, and the estimation of the bid amount equation shows that the degree to which bidders of different experience levels react to the seller's reputation is very similar to what was estimated when using only the selected sample.

V. MARKET EXPERIENCE AND STRATEGIC TIMING OF BID PLACEMENT

As noted above, Roth and Ockenfels (2002) argue convincingly that rational bidders may wait to place their final bid until near the end of an auction, in an effort to pay lower prices. Wilcox (2000) and Ockenfels and Roth (2006) both show that more experienced bidders do in fact wait longer to submit their final bids. In this section, I examine the full history of bids in each sample. The results are very similar to what is found in previous studies: experienced bidders wait substantially longer to submit their bids than inexperienced bidders. This result is extended by examining how much experience bidders need to gain before they learn of the benefits to waiting to submit their bids. This lesson appears to be learned very quickly: bidders with a small amount of experience already submit their bids much closer to the end of an auction than bidders with no experience, but there is little change in the timing of bid placement as the bidders gain additional experience.

This hypothesis is tested in two ways. In both sets of tests, each bid placed in each auction is treated as a separate observation. The first test regresses the number of minutes left in the auction when the bidder's final bid was submitted against the bidder's feedback rating and the number of bidders who participated in the auction. Since bids from the same auction will be correlated, the specification allows for the presence of auction-level random effects and calculates standard errors that are clustered by auction. Two specifications are estimated. Specification 1 assumes that the relationship between the number of minutes left in the auction and the bidder's feedback rating is linear:

(2) [MINLEFT.sub.ij] = [[alpha].sub.ij] + [BEXP.sub.ij] [[beta].sub.1] + [BIDDERS.sub.i] [[beta].sub.2] + ([c.sub.i] + [u.sub.ij]),

where [MINLEFT.sub.ij] is the number of minutes left in auction i when bidder j submitted her final bid, [BEXP.sub.ij] is the feedback rating of bidder j in auction i, [BIDDERS.sub.i] is the number of bidders who placed a bid in auction i, and the error term, [c.sub.i] + [u.sub.ij], includes an auction random-effects component. Specification 2 instead breaks the bidders into categories by their feedback ratings as before: bidders who have a rating of 0 make up the first category, and the remaining bidders who have positive ratings (in absolute value) are split into quartiles. Bidders who have a rating of 0 are the reference group:

(3) [MINLEFT.sub.ij] = [[alpha].sub.ij] + [BEXP1.sub.ij] [[beta].sub.1] + [BEXP2.sub.ij] [[beta].sub.2] + [BEXP3.sub.ij] [[beta].sub.3] + [BEXP4.sub.ij] [[beta].sub.4] + [BIDDERS.sub.i] [[beta].sub.5] + ([c.sub.i] + [u.sub.ij]),

where [BEXP1.sub.ij] through [BEXP4.sub.ij] are dummy variables indicating the quartile of bidder experience as measured by their feedback rating.

The results of the estimation of Specification 1 are reported in Columns 1 and 3 of Table 7 for the golf clubs sample and the video game sample, respectively. The results from the golf club sample are very similar to the findings of Ockenfels and Roth (2006): if the bidder's feedback rating increases by 1, that bidder waits an additional 0.92 min to submit her bid. No linear relationship is found between the bidder's feedback rating and the time they place their final bid in the video game sample, however.

The results of the estimation of Specification 2 are reported in Columns 2 and 4 for the golf clubs sample and the video game sample, respectively. They add an interesting wrinkle to the story: in both samples, bidders whose feedback rating is in just the first quartile (a rating of 1-2 in the golf club sample and a rating of 1-5 in the video game sample) submit their bids over 9 h closer to the end of the auction than bidders who have a feedback rating of 0. So, bidders learn that it is beneficial to wait much longer to submit their final bids closer to the end of the auction after participating in only a handful of eBay auctions. There is no statistical difference, however, between the time left in the auction when bidders in Quartile 1 place their final bids and the time left when bidders with higher ratings (and more experience) place their bids in either sample.

Ockenfels and Roth (2006) examine the relationship between the probability that a bid is placed with only 10 min left in an auction and the bidder's experience level, while the working paper version of their study also looked at the probability that the bidder's final bid was placed with smaller amounts of time left in the auction. Wilcox (2000) also examines the relationship between the probability that a bid is placed with 1 min left in an auction and the bidder's experience level. Following their work, I examine the relationship between the different groups of bidders categorized by their feedback ratings and the probability that they placed their final bid within the final 30 s, 1 min, 5 min, and 10 min of the auction. Probits of the following form are estimated:

Pr([TIMELEFT.sub.ij] = 1) = + [BEXP1.sub.ij] [[beta].sub.1] + [BEXP2.sub.ij] [[beta].sub.2] + [BEXP3.sub.ij] [[beta].sub.3] + [BEXP4.sub.ij] [[beta].sub.4] + [BIDDERS.sub.i] [[beta].sub.5] + [[epsilon].sub.ij], (4)

where TIMELEFT is a dummy variable that equals 1 if 30 s, 1 min, 5 min, or 10 min is left until the end of auction i when bidder j submits her final bid, as appropriate, and the other variables are as previously defined. Again, standard errors are clustered by auction. Table 8 reports the estimated marginal effects of each covariate for regressions using each of the four dependent variables. Columns l, 2, 3, and 4 examine the probability that the final bid is placed within the final 30 s, 1 min, 5 min, and 10 min of the auction, respectively, for the golf clubs sample. Columns 5-8 do the same for the video game sample. In the golf clubs sample, in each case, bidders in Quartile 1 are significantly more likely to wait until the specified time to place their final bid, but bidders who have additional experience are no more likely to wait until the specified time to submit their final bid. Consider first the results reported in Column 1. Bidders who have a rating of 1 or 2 are 2.1 percentage points more likely to place their final bid in the last 30 s than bidders who have a rating of 0. This effect is very large relative to the sample mean: only 3% of the bidders in the sample place their bid in the final 30 s of the auction. However, there is no statistical difference between the probability that bidders who have a rating of 1 or 2 will bid in the last 30 s and the probability that bidders in the higher quartiles will bid in the last 30 s. The same pattern holds for each time interval: bidders with a positive feedback rating are statistically significantly more likely to place their final bid in the last 1 min, last 5 min, and last 10 min of an auction, and each of the marginal effects is quite large relative to the sample mean, but there is no statistical difference in these probabilities across the quartiles of positive reports.

The results from the video game sample are somewhat different. For each dependent variable, although the point estimates are all positive, the bidders in the first quartile of experience are not statistically significantly more likely to place their final bid in the closing moments than bidders with a feedback rating of 0. Bidders in higher quartiles, however, are substantially more likely to place their bids in the closing moments, and there is no statistical difference in this probability among the bidders in experience Quartiles 2-4 in any specification.

VI. CONCLUSIONS

This study has shown that inexperienced bidders, as measured by their feedback rating, behave differently than more experienced bidders. For two products of different average values, inexperienced bidders bid just as much in the auctions of sellers who have yet to establish a reputation as they do in the auctions of sellers who have a long history of honest trades. More experienced bidders, however, shade their bids substantially if the seller has a weak reputation, and they raise their bids as the seller's reputation improves. Although there are many explanations for these findings, one possibility is that bidders learn from their experiences and discover ways to behave that increase their expected payoffs. Inexperienced bidders may not bid less when the seller has yet to establish a reputation because they have not considered the risks involved. The fact that they are expected to send payment to sellers before the seller sends them their winnings puts them in a risky position. In deciding how much to bid, bidders may learn from any bad experiences they might have to take into account the probability that the seller will fail to deliver what they promise.

Inexperienced bidders also differ in their choices of when to place their bids. Roth and Ockenfels (2002) argue that rational bidders will wait until near the end of auctions to submit their final bid. The data show that experienced bidders wait much longer to submit their bids than bidders who have no experience and experienced bidders are also much more likely to submit their final bid in the closing moments of the auction. However, bidders quickly learn about the benefits of patience-bidders who have only been involved in a handful of eBay transactions bid much later and are more likely to wait until the end to place their bids than bidders with a feedback rating of 0, but bidders with more experience bid no later and are no more likely to wait until the end than bidders who have a small but positive amount of experience.

ABBREVIATION

OLS: Ordinary Least Squares

doi: 10.1111/j.1465-7295.2008.00128.x

REFERENCES

Ba, S., and P. Pavlou. "Evidence on the Effect of Trust Building Technology in Electronic Markets: Price Premiums and Buyer Behavior." MIS Quarterly, 26, 2002, 243-68.

Bajari, P., and A. Hortacsu. "Economic Insights from Internet Auctions." Journal of Economic Literature, 42, 2004, 457-86.

Dellarocas, C., and C. A. Wood. "The Sound of Silence in Online Feedback: Estimating Trading Risks in the Presence of Reporting Bias." Management Science, 54, 2008, 460-76.

Eaton, D. H. "Valuing Information: Evidence from Guitar Auctions on eBay." Journal of Applied Economics and Policy, 24, 2005, 1-19.

Garratt, R., M. Walker, and J. Wooders. "Behavior in Second-Price Auctions by Highly Experienced eBay Buyers and Sellers." Working Paper, University of California, Santa Barbara, 2004.

Houser, D., and J. Wooders. "Reputation in Auctions: Theory, and Evidence from eBay." Journal of Economics & Management Strategy, 15, 2006, 353-69.

Kauffman, R. J., and C. A. Wood. "Running Up the Bid: Modeling Seller Opportunism in Internet Auctions," in Proceedings of the 2000 Americas Conference on Information Systems, edited by M. Chung. Long Beach, CA: Association for Information Systems, 2000, 929-35.

--. "Doing Their Bidding: An Empirical Examination of Factors That Affect a Buyer's Utility in Internet Auctions." Information Technology and Management, 7, 2005, 171-90.

Klein, T. J., C. Lambertz, G. Spagnolo, and K. O. Stahl. "Last Minute Feedback." Manuscript, University of Mannheim, January 2007.

Livingston, J. A. "How Valuable is a Good Reputation? A Sample Selection Model of Internet Auctions." Review of Economics and Statistics, 87, 2005, 453-65.

Lucking-Reiley, D., D. Bryan, N. Prasad, and D. Reeves. "Pennies from eBay: The Determinants of Price in Online Auctions." Journal of Industrial Economics, 55, 2007, 223-33.

MacLeod, W. B. "Reputations, Relationships, and Contract Enforcement." Journal of Economic Literature, 45, 2007, 595-628.

McDonald, C. G., and V. C. Slawson, Jr. "Reputation in an Internet Auction Market." Economic Inquiry, 40, 2002, 633-50.

Melnik, M. I., and J. Alto. "Does a Seller's ECommerce Reputation Matter? Evidence from EBay Auctions." Journal of Industrial Economics, 50, 2002, 337-50.

--. "Seller Reputation, Information Signals, and Prices for Heterogeneous Coins on eBay." Southern Economic Journal, 72, 2005, 305-28.

Ockenfels, A., and A. E. Roth. "Late and Multiple Bidding in Second Price Internet Auctions: Theory and Evidence Concerning Different Rules for Ending an Auction." Games and Economic Behavior, 55, 2006, 297-320.

Resnick, P., and R. Zeckhauser. "Trust among Strangers in Internet Transactions: Empirical Analysis of eBay's Reputation System," in The Economics of the Internet and E-Commerce (Advances in Applied Microeconomics. Vol. 11), edited by M. Baye. Oxford: JAI Press, 2002.

Reynolds, S. S., and J. C. Wooders. "Auctions with a Buy Price." University of Arizona Working Paper No. 03-01, 2003.

Roth, A. E., and A. Ockenfels. "Last Minute Bidding and the Rules for Ending Second-Price Auctions: Theory and Evidence from a Natural Experiment on the Internet." American Economic Review, 92, 2002, 1093-103.

Wilcox, R. T. "Experts and Amateurs: The Role of Experience in Internet Auctions." Marketing Letters, 11, 2000, 363-74.

(1.) Figures are cited from eBay's 2007 second-quarter earnings release, available at http://investor.ebay.com/ financial_releases.crm.

(2.) Rumors are that the seller had large gambling debts to pay off. See the discussion at www.slashdot.org: http:// slashdot.org/comments.pl?sid=28414&cid= 3054010; for the full discussion of this incident, see http://slashdot.org/ article.pl?sid=12/12/22/1941255&mode=thread&tid=98.

(3.) Sellers are also required to provide a credit card number to confirm their identity, but it is possible for a malicious seller to obtain fraudulent credit card.

(4.) See Bajari and Hortacsu (2004) for a full description of the eBay marketplace and its reputation system.

(5.) Ba and Pavlou (2002), Eaton (2005), Houser and Wooders (2006), Lucking-Reiley et al. (2007), McDonald and Slawson (2002), Melnik and Aim (2002), and Resnick and Zeckhauser (2002) are among the many papers that have studied this issue; Bajari and Hortacsu (2004) provide a survey of many of these results.

(6.) Waiting until the closing moments can be a best response to several strategies. For example, Roth and Ockenfels (2002) note, "inexperienced bidders might make an analogy with first-price 'English' auctions, and be prepared to continually raise their bids to maintain their status as high bidder.... Bidding very near the end of the auction would not give the incremental bidder sufficient time to respond, and so a sniper competing with an incremental bidder might win the auction at the incremental bidder's initial, low bid. In contrast, bidding one's true value early in the auction ... would win the auction only if one's true value were higher than the incremental bidder's, and in that case would have to pay the incremental bidder's true value."

(7.) See Livingston (2005) for a discussion of why this is an appropriate assumption.

(8.) The Buy It Now option disappears as soon as a bid in most cases. So, in most cases where a buyer was able to accept the Buy It Now option, only one bid was placed and the observation already had to be dropped. This is the case for all Buy It Now auctions in the video game sample, so no additional observations had to be dropped from that sample. However, if the seller uses a secret reserve price, the Buy It Now option does not disappear until the highest bid exceeds that price. This was the case with the 48 auctions from the golf clubs sample where two or more bids were placed despite the fact that a buyer ended the auction early by accepting the Buy It Now price.

(9.) Reynolds and Wooders (2003) show that when a Buy It Now price is available, the bidders' strategies become more complicated. They place a bid if their value is below some cutoff value c* which depends on their risk preferences. Thanks to an anonymous referee for pointing this out.

(10.) Since the focus of the study is to examine how differences in the experience level of bidders affect their bidding behavior, and not to examine the impact of bidder experience on the seller revenue conditional on a sale occurring, there is no need to drop any remaining observations where the auction did not result in a sale. Doing so would discard observations where we do have information about the second-highest bidder's experience level, and the winning bid is representative of that bidder's willingness to pay, when there is no need to do so. If such observations are dropped, however, the qualitative results do not change in either sample.

(11.) This is unlikely to be a large problem since negative feedback is left very rarely on ebay. Resnick and Zeckhauser (2002) and Kauffman and Wood (2005) estimate that more than 99% of all feedback left on ebay is positive. Bidders are in particular highly unlikely to receive negative feedback. Sellers who leave negative feedback for bidders risk receiving negative feedback in retaliation, and since negative feedback is much more damaging to sellers than to buyers, sellers almost never leave bidders negative feedback in practice. Recent studies by Dellarocas and Wood (2008) and Klein et al. (2007) present empirical evidence that fear of retaliation does in fact lead trading partners to avoid leaving negative feedback for each other.

(12.) Sellers usually choose a fixed shipping price that the bidder must agree to before placing a bid, but occasionally they require bidders to pay "actual shipping charges," which are not specified. In this case, shipping charges are taken to be the median of the fixed price charged in the rest of the sample.

(13.) Livingston (2005) argues that eBay's publicly known reserve price, the minimum bid allowed by the seller, should affect the bidder's decision of whether to place a bid but not the decision of how much to bid. Accordingly, the minimum bid is not included as a regressor, but doing so has no impact on the qualitative results.

(14.) Thanks to an anonymous referee for pointing out this possible explanation.

(15.) As previously noted, Livingston (2005) argues that the minimum bid should affect the decision of whether to place a bid but not the decision of how much to bid. It is therefore an ideal candidate for an exclusion restriction in a sample selection model.

JEFFREY A. LIVINGSTON *

* Thanks to two anonymous referees, Dhaval Dave, Patrick Scholten, Saumyanil Deb, and attendees at seminars at Bentley College and the Southern Economic Association conference who all provided many helpful comments and suggestions. Bidisha Ghosh and Yunlei Tu provided excellent research assistance. Of course, all remaining errors are my own.

Livingston: Assistant Professor, Department of Economics, Bentley College, 175 Forest Street, Waltham, MA 02474. Phone 1-781-891-2538; Fax 1-781-891-2896: E-mail jlivingston@bentley.edu
TABLE 1
Summary Statistics

 Mean (Standard Deviation)

 Taylor Made HALO: Combat
 Firesole Irons Evolved Video
 Golf Clubs Game for Xbox
Variable (N = 540) (N = 597)

Dependent variable
 Price: second-highest 393.31 (81.61) 22.97 (3.59)
 valuation
Reported history of seller
 =1 if seller has 0 positive 0.06 (0.25) 0.01 (0.12)
 reports
 =1 if positive reports are in 0.19 (0.39) 0.24 (0.43)
 Quartile l
 =1 if positive reports are in 0.25 (0.43) 0.25 (0.43)
 Quartile 2
 =1 if positive reports are in 0.25 (0.43) 0.27 (0.45)
 Quartile 3
 =1 if positive reports are in 0.25 (0.43) 0.22 (0.42)
 Quartile 4
% of reports that were negative 0.02 (0.06) 0.01 (0.03)
Bidder experience
 =1 if bidder has 0 feedback 0.28 (0.45) 0.05 (0.22)
 rating
 =1 if bidder rating is in 0.27 (0.45) 0.24 (0.43)
 Quartile 1
 =1 if bidder rating is in 0.12 (0.33) 0.25 (0.43)
 Quartile 2
 =1 if bidder rating is in 0.15 (0.36) 0.23 (0.42)
 Quartile 3
 =1 if bidder rating is in 0.17 (0.38) 0.23 (0.42)
 Quartile 4
Controls for auction, item,
 or market heterogeneity
 Retail price 850.55 (82.68)
 Shipping charges 15.11 (3.93) 5.11 (1.89)
 =1 if new (not used) 0.39 (0.49) 0.56 (0.50)
 =1 if payment by credit 0.48 (0.50) 0.96 (0.18)
 card accepted
 Duration of auction (in d) 6.05 (1.85) 4.74 (2.42)
 =1 if secret reserve price 0.55 (0.50) 0.04 (0.20)
 used
 =l if left-handed 0.02 (0.13)
 =1 if designed for women 0.02 (0.13)
 =1 if designed for seniors 0.02 (0.14)
 =1 if seller is a "Power 0.28 (0.45)
 Seller"
 =1 if seller uses "ID 0.03 (0.17)
 Verify"
 =1 if seller uses "About Me" 0.13 (0.33)

TABLE 2
Summary Statistics: Full Bid History of Auctions

 Mean (Standard Deviation)
 Taylor Made Firesole HALO: Combat Evolved
 Irons Golf Clubs Video Game for Xbox
Variable (N = 4,377) (N = 4,394)

Dependent variables
 Minutes left in 3,704.94 (3,575.64) 1,444.19 (2,383.07)
 auction when
 final bid was
 placed
 30 s left in 0.03 (0.17) 0.05 (0.23)
 auction when
 final bid was
 placed
 1 min left when 0.04 (0.20) 0.08 (0.26)
 final bid was
 placed
 5 min left when 0.08 (0.28) 0.12 (0.32)
 final bid was
 placed
 10 min left when 0.10 (0.30) 0.14 (0.35)
 final bid was
 placed
Bidder experience
 Bidder feedback 21.95 (100.65) 58.75 (191.53)
 rating
 Bidder has 0 0.31 (0.46) 0.07 (0.25)
 feedback rating
 Bidder rating is 0.24 (0.43) 0.24 (0.42)
 in Quartile 1
 Bidder rating is 0.11 (0.31) 0.23 (0.42)
 in Quartile 2
 Bidder rating is 0.17 (0.38) 0.23 (0.42)
 in Quartile 3
 Bidder rating is 0.17 (0.37) 0.23 (0.42)
 in Quartile 4
Number of bidders 9.90 (4.19) 8.29 (2.64)
 participating in
 the auction

TABLE 3
Reputation Reaction by All Bidders
Taylor Made Firesole Irons

Dependent Variable: Price
(Second-Highest Valuation) (1)

Seller positive reports
 1-31 (Quartile 1) 24.06 (1.72) *
 32-205 (Quartile 2) 36.68 (2.63) ***
 206-836 (Quartile 3) 32.10 (2.33) **
 >836 (Quartile 4) 29.81 (2.16) **
% of reports that are negative 12.23 (0.41)
Left-handed 7.09 (0.27)
Designed for women 47.79 (1.92) *
Designed for seniors -2.42 (0.18)
New (not used) 86.78 (15.67) ***
Retail price 0.42 (13.33) ***
Shipping charges -1.49 (2.27) **
Auction ran for 5 d 3.19 (0.41)
Auction ran for 7 d 0.58 (0.07)
Auction ran for 10 d 24.22 (2.04) **
Payment by credit card accepted -4.98 (1.02)
Secret reserve price used -17.91 (3.16) ***
Constant 2.49 (0.08)
Observations 540
[R.sup.2] .584

Notes: Robust t statistics in parentheses.

* Significant at 10%; ** significant at 5%; *** significant
at 1%.

TABLE 4
Reputation Reaction by Bidders of Different Experience
Levels, Taylor Made Firesole Irons (a,b)

 Seller Positive
Bidder rating (1) Reports

1-2 (Quartile 1) -2.04 (0.10) 1-31 (Quartile 1)
3-5 (Quartile 2) -9.99 (0.36) 32-205 (Quartile 2)
6-22 (Quartile 3) -41.49 (1.19) 206-836 (Quartile 3)
>22 (Quartile 4) -69.69 (2.64) *** >836 (Quartile 4)
Observations 540
[R.sup.2] .603

 Bidder Rating

 0 1-2 (Quartile 1)

Bidder rating (2) (3)

1-2 (Quartile 1) 3.35 (0.19) 2.35 (0.13)
3-5 (Quartile 2) 31.76 (1.59) 15.96 (0.93)
6-22 (Quartile 3) 13.74 (0.80) 7.10 (0.44)
>22 (Quartile 4) 1.22 (0.07) 14.95 (0.87)
Observations
[R.sup.2]

 Bidder Rating

 3-5 (Quartile 2) 6-22 (Quartile 3)

Bidder rating (4) (5)

1-2 (Quartile 1) -0.40 (0.02) 35.12 (1.02)
3-5 (Quartile 2) 20.68 (0.77) 45.69 (1.31)
6-22 (Quartile 3) 37.78 (1.44) 20.08 (0.54)
>22 (Quartile 4) 8.97 (0.34) 33.70 (1.04)
Observations
[R.sup.2]

 Bidder Rating

 >22 (Quartile 4)

Bidder rating (6)

1-2 (Quartile 1) 73.21 (3.08) ***
3-5 (Quartile 2) 65.91 (2.77) ***
6-22 (Quartile 3) 70.20 (2.92) ***
>22 (Quartile 4) 79.88 (3.40) ***
Observations
[R.sup.2]

Notes: Robust t statistics in parentheses.

(a) Results of estimation of Equation (1) are displayed. Column 1
reports the bidder feedback quartile indicator estimates. Columns
2-6 report the estimates of the interactions between the bidder
feedback quartile indicators and the seller positive report quartile
indicators.

(b) Specification includes the same auction-specific characteristics
as controlled for in Table 3. These results are not reported but can
be obtained from the author by request.

*** Significant at 1%.

TABLE 5
Reputation Reaction by All Bidders,
HALO Video Game

Dependent Variable: Price
(Second-Highest Valuation) (t)

Seller positive reports
 1-34 (Quartile 1) 2.13 (1.64)
 35-144 (Quartile 2) 2.72 (2.11) **
 145-1,422 (Quartile 3) 2.59 (1.97) **
 >1,422 (Quartile 4) 1.85 (1.36)
% of reports that are negative -7.04 (1.44)
Power Seller 0.39 (0.79)
ID Verify -2.18 (2.48) **
About Me 0.23 (0.50)
Shipping charges -0.23 (2.85) ***
Auction ran for 1 d -0.60 (1.23)
Auction ran for 5 d 0.25 (0.56)
Auction ran for 7 d -0.00 (0.01)
Auction ran for 10 d 0.13 (0.15)
Payment by credit card accepted 2.33 (2.40) **
Secret reserve price used -0.01 (0.01)
Constant 19.65 (12.37) ***
Observations 597
[R.sup.2] .057

Notes: Robust t statistics in parentheses.

** Significant at 5%; *** significant at 1%.

TABLE 6
Reputation Reaction by Bidders of Different Experience Levels, HALO
Video Game (a,b)

Bidder rating (1) Seller Positive Reports

1-5 (Quartile l) 4.67 (1.58) 1-34 (Quartile 1)
6-16 (Quartile 2) -2.25 (0.77) 35-144 (Quartile 2)
17-44 (Quartile 3) -7.08 (1.81) * 145-1,422 (Quartile 3)
>44 (Quartile 4) -4.61 (1.82) * >1,422 (Quartile 4)
Observations 597
[R.sup.2] .094

 Bidder Rating

 0 (c) 1-5 (Quartile 1)

Bidder rating (2) (3)

1-5 (Quartile l) 0.59 (0.27) -3.92 (1.50)
6-16 (Quartile 2) 3.25 (1.73) * -3.61 (1.38)
17-44 (Quartile 3) -0.04 (0.02) -4.50 (1.72) *
>44 (Quartile 4) -4.47 (1.68) *
Observations
[R.sup.2]

 Bidder Rating

 6-16 (Quartile 2) 17-44 (Quartile 3)

Bidder rating (4) (5)

1-5 (Quartile l) 1.85 (0.72) 7.03 (1.93) *
6-16 (Quartile 2) 2.63 (1.03) 7.31 (1.99) **
17-44 (Quartile 3) 2.39 (0.92) 8.24 (2.25) **
>44 (Quartile 4) 2.04 (0.78) 6.33 (1.71) *
Observations
[R.sup.2]

 Bidder Rating

 >44 (Quartile 4)

Bidder rating (6)

1-5 (Quartile l) 4.62 (2.17) **
6-16 (Quartile 2) 5.02 (2.36) **
17-44 (Quartile 3) 5.30 (2.51) **
>44 (Quartile 4) 4.46 (2.06) **
Observations
[R.sup.2]

Notes: Robust t statistics in parentheses

(a) Results of estimation of Equation (1) are displayed. Column 1
reports the bidder feedback quartile indicator estimates. Columns
2-6 report the estimates of the interactions between the bidder
feedback quartile indicators and the seller positive report quartile
indicators.

(b) Specification includes the same auction-specific characteristics
as controlled for in Table 5. These results are not reported but can
be obtained from the author by request.

(c) No observations exist where the bidder had a feedback rating of
0 and the seller had no positive reports, so the interaction between
the 0 rating dummy and the fourth positive report quartile is used
as the reference category for bidders with a rating of 0.

* Significant at 10%; ** significant at 5%.

TABLE 7
Effect of Bidder Experience on Timing of Bid Placement (a,b)

Dependent Variable: Taylor Made Firesole Irons
Minutes Left in
Auction When Final
Bid Was Placed
Bidder feedback
rating (1) (2)

Raw rating -0.92 (2.22) **
Rating is in -567.84 (4.23) ***
 Quartile 1
Rating is in -583.50 (3.37) ***
 Quartile 2
Rating is in -278.48 (1.94) *
 Quartile 3
Rating is in -515.36 (3.54) ***
 Quartile 4
No. of bidders 200.28 (9.11) *** 198.37 (9.05) ***
Constant 1,741.17 (9.02) *** 2,074.66 (10.15) ***
Observations 4,377 4,377
[R.sup.2] .040 .044

Dependent Variable: HALO Video Game
Minutes Left in
Auction When Final
Bid Was Placed
Bidder feedback
rating (3) (4)

Raw rating 0.20 (0.75)
Rating is in -591.06 (3.10) ***
 Quartile 1
Rating is in -631.77 (3.38) ***
 Quartile 2
Rating is in -750.13 (3.96) ***
 Quartile 3
Rating is in -681.09 (3.57) ***
 Quartile 4
No. of bidders 33.09 (1.74) * 30.82 (1.62)
Constant 1,157.86 (7.31) *** 1,816.85 (7.62) ***
Observations 4,394 4,394
[R.sup.2] .001 .007

Notes: Robust z statistics in parentheses.

(a) Level of observation: a bid placed in an auction.

(b) Random-effects regressions with standard errors clustered by
auction.

* Significant at 10%; ** significant at 5%; *** significant at 1%.

TABLE 8
Probability That Final Bid Is Placed at the End of the Auction (a,b)

Dependent Taylor Made Firesole Irons
Variable = 1 if
Bid Is Placed 30s 1 min
in Final (1) (2)

Bidder rating
 Rating is in Quartile 1 0.021 (2.77) *** 0.028 (3.12) ***
 Rating is in Quartile 2 0.015 (1.54) 0.018 (1.47)
 Rating is in Quartile 3 0.023 (2.53) ** 0.026 (2.50) **
 Rating is in Quartile 4 0.043 (4.71) *** 0.046 (4.39) ***
No. of bidders -0.003 (4.29) *** -0.004 (4.39) ***
Observations 4,377 4,377
Pseudo [R.sup.2] .049 .036

Dependent Taylor Made Firesole Irons
Variable = 1 if
Bid Is Placed 5 min 10 min
in Final (3) (4)

Bidder rating
 Rating is in Quartile 1 0.052 (4.03) *** 0.051 (3.86) ***
 Rating is in Quartile 2 0.047 (2.88) *** 0.037 (2.21) **
 Rating is in Quartile 3 0.028 (1.97) ** 0.018 (1.20)
 Rating is in Quartile 4 0.053 (3.60) *** 0.050 (3.37) ***
No. of bidders -0.009 (6.28) *** -0.010 (6.17) ***
Observations 4,377 4,377
Pseudo [R.sup.2] .045 .039

Dependent HALO Video Game
Variable = 1 if
Bid Is Placed 30s 1 min
in Final (5) (6)

Bidder rating
 Rating is in Quartile 1 0.027 (1.39) 0.035 (1.56)
 Rating is in Quartile 2 0.050 (2.39) ** 0.055 (2.41) **
 Rating is in Quartile 3 0.051 (2.40) ** 0.060 (2.59) ***
 Rating is in Quartile 4 0.044 (2.09) ** 0.059 (2.56) **
No. of bidders -0.001 (0.91) -0.001 (0.96)
Observations 4,394 4,394
Pseudo [R.sup.2] .006 .005

Dependent HALO Video Game
Variable = 1 if
Bid Is Placed 5 min 10 min
in Final (7) (8)

Bidder rating
 Rating is in Quartile 1 0.037 (1.40) 0.027 (0.99)
 Rating is in Quartile 2 0.057 (2.19) ** 0.055 (2.15) **
 Rating is in Quartile 3 0.074 (2.81) *** 0.071 (2.68) ***
 Rating is in Quartile 4 0.053 (1.96) * 0.046 (1.73) *
No. of bidders -0.004 (1.83) * -0.005 (2.18) **
Observations 4,394 4,394
Pseudo [R.sup.2] .005 .005

Notes: Robust = statistics in parentheses.

(a) Level of observation: a bid placed in an auction.

(b) Probits with standard errors clustered by auction.

* Significant at 10%; ** significant at 5%; *** significant at 1%.
COPYRIGHT 2010 Western Economic Association International
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2010 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Livingston, Jeffrey A.
Publication:Economic Inquiry
Geographic Code:1USA
Date:Apr 1, 2010
Words:11526
Previous Article:The effect of prayer on God's attitude toward mankind.
Next Article:Centralized and decentralized management of local common pool resources in the developing world: experimental evidence from fishing communities in...
Topics:

Terms of use | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters