Empirical evidence on EBay bidding strategies.

ABSTRACT

EBay has become a global force in electronic commerce. Numerous books and articles have offered advice to potential buyers and sellers as to the most effective strategies for improving their prices. This study analyzes price data from eBay transactions to measure the impact of various buyer and seller strategies on ultimate price. The measurement of seller trustworthiness is studied and its role in price determination is analyzed. It appears that some aspects of the conventional wisdom about eBay tactics are upheld while others do not seem statistically significant. With the growing use of on-line auctions and electronic markets, these factors are likely to grow in importance in the future.

INTRODUCTION

In just a few years, eBay has emerged as a dominant force in e-commerce. Founded in September 1995, eBay has become the world's largest online marketplace. The eBay community includes more than a hundred million registered members from around the world. People spend more time on eBay than any other online site, making it the most popular shopping destination on the Internet. As stated in their Annual Report, eBay's mission is to provide a global trading platform where practically anyone can trade practically anything. On an average day, there are millions of items listed on eBay. The annual revenues for eBay are greater than the gross domestic product of more than half the nations in the world (Hof 2003).

EBay has become a global phenomenon. EBay has local sites that serve Australia, Austria, Belgium, Canada, China, France, Germany, Hong Kong, India, Ireland, Italy, Malaysia, the Netherlands, New Zealand, Poland, the Philippines, Singapore, South Korea, Spain, Sweden, Switzerland, Taiwan, the United Kingdom, and the United States. Furthermore, eBay is a major business success. EBay has a market capitalization of over $58 billion, giving it a valuation greater than General Motors and Ford put together. If the 430,000 individuals who earn all or most of their income selling on the site were eBay employees, it would be the second largest employer in the Fortune 500 list, behind only Wal-Mart (Sellers 2004).

Numerous books and articles have suggested various bidding and selling strategies to improve the transaction prices of desired products. This study analyzes a data set of eBay transactions to determine whether or not these proposed strategies result in better prices. In particular, the effects of sniping (last minute bidding), opening bid manipulation, seller track record, handling/shipping costs substitution, and other commonly observed approaches are analyzed. Prior research in this area has been ambiguous. Most studies have focused on a particular aspect of transactions and have not controlled for other variables. This study develops a multivariate model of auction price determination. There is evidence to suggest that proper use of specific strategies can produce superior net results for both buyers and sellers.

BACKGROUND

The rise of electronic commerce has led to numerous online markets where buyers and sellers have never met. Much of the previous research into online auctions has focused on the role of trust. Trust is an important factor in every marketplace, but even more so in electronic commerce (Ba et al. 1999, Brynjolfsson and Smith 2000). The impersonal and anonymous nature of electronic commerce creates a fertile environment for the manipulation of transactions and potential fraud. Almost all online auction sites address this issue through the use of information exchange about the background and experience of buyers and sellers. The most notable of these systems is EBay's Feedback Forum. Online feedback mechanisms allow buyers and sellers to report their satisfaction with each transaction in a public forum. These mechanisms can serve to build trust among the participants in electronic commerce (Walden 2000, Ba and Pavlou 2002). The fundamental problem is the information asymmetry between buyers and sellers (Akerlof 1970). When bidders view a product at an auction site, they do not have the opportunity to inspect the product and directly observe the quality (Fung and Lee 1999). This unequal access to information can lead to fraud or market failure. Buyers are exposed to additional risk because they must rely on incomplete or potentially distorted information provided by the seller (Lee 1998). The building of trust is most crucial when two factors are present: risk and incomplete/asymmetric information (Swan and Nolan 1985). Both of these factors are more prevalent in electronic commerce than traditional markets. Online feedback mechanisms, also known as reputation systems (Resnick et al. 2000), are using the Internet's bidirectional communication capability to artificially engineer large-scale, word-of-mouth networks in which individuals share opinions and experiences (Dellarocas 2003). Despite the wide use of reputational mechanisms such as EBay's Feedback Forum to promote trust, there has been little empirical evidence as to whether these mechanisms actually induce trust or create any desirable outcomes such as increased bidding or higher final prices (Ba and Pavlou 2002). This study will investigate several measures of trustworthiness and analyze the effect on transaction prices.

The sheer magnitude of the data required for assessing reputations is daunting. At any given moment, eBay will list several million products for sale to over 100 million potential customers. An ongoing trading relationship between buyers and sellers is one way to develop trust (Ba and Pavlou 2002), but the vast majority of eBay transactions are one shot deals. Resnick and Zeckhauser (2002) found that 89% of all buyer-seller pairs conducted just one transaction during the 5 month period under investigation.

This study will analyze the impact of buyer and seller reputation on final transaction prices. Earlier studies have tried to measure the importance of reputation in determining both the probability of sale and the ultimate price in online transactions. The results have been mixed. Most studies show at least a weak positive impact of reputation on price, but some have found the opposite effect (Dellarocas 2003). Part of the problem in measuring the impact of reputation is the need to control for other variables. In this study, reputation will be measured in several ways and other factors affecting price will be incorporated into a multivariate model.

HYPOTHESES AND DATA SET

To study the determinants of prices in eBay transactions, the first step was to define a product class. Prior studies have generally chosen a single product in order to hold constant for the effect of product attributes. Products sold on eBay can vary between disposable items selling for less than one dollar to real estate selling in the tens of millions. The price level plays a significant role in determining the importance of reputation. The impact of feedback profiles on prices has been found to be more important for riskier transactions and more expensive products (Dellarocas 2003, Lee et al. 2000). The product chosen for this study was the DVD of the first season of The Sopranos (HBO, 1999). This product was chosen for several reasons. First, it is readily available from a variety of online and traditional retailers. It is an essentially standardized product, but may still raise quality issues in the mind of the buyer. Approximately half of the transactions studied were for used DVDs, thus raising the possibility of purchasing a damaged disc. Many of the sellers were located abroad. This could create suspicion that the DVDs could be bootlegs or be in a format that is incompatible with North American DVD players. The average price was approximately $28 (including used versions). This is high enough to be treated somewhat seriously by potential buyers, but low enough to not require extensive shopping or inspection. An older DVD (vintage 1999) was chosen to prevent any distortions due to time effects from recent releases.

The following variables and hypotheses were studied:

PRICE. Prices were collected for 147 transactions. The average price for new versions of the DVD was approximately $32. This is considerably below the retail price ($59.98) and the price charged by traditional outlets (Wal-Mart $44.88, Barnes & Noble $43.18). Other online vendors were also more expensive. Amazon.com and Target.com both charged $44.99. The eBay selling price served as the dependent variable in this study.

Reputation. As discussed previously, reputation is crucial in online markets. EBay has recognized this fact and developed several mechanisms for measuring seller trustworthiness. This study used four measures:

(1) SELLER SCORE. At the conclusion of each transaction, buyers are given the opportunity to rate their satisfaction with the seller as (-) negative, (0) neutral, or (+) positive. Buyers provide a rating for approximately 52% of all transactions (Resnick and Zeckhauser 2002). EBay then calculates a Seller Score as the number of positive transactions with unique buyers that a seller has completed. This score measures both the experience and reliability of a seller. In this study, the seller scores ranged between 0 and 9,665. A score of zero means that this was the first eBay sale for that seller. A score of over 9,000 is indicative of a commercial enterprise selling to many different buyers who were satisfied with their experience. The hypothesis in this study is that a high seller score is likely to lead to higher final transaction prices. All other things equal, buyers should be willing to pay a price premium to do business with individuals (or companies) with an established positive track record. This is consistent with the findings of Dewan and Hsu (2002) and Melnik and Alm (2002) among others.

(2) SELLER % POSITIVE. In addition to calculating a Seller Score, eBay also provides information on the total number and distribution of ratings for each seller. From this, it is possible to calculate the percentage of positive ratings that each seller has received. Unlike the seller score, this variable does not depend on the volume of activity, but simply measures the reliability of the seller. This variable should play a similar role to Seller Score, but raises the possibility of overstating the trustworthiness of a seller who has not engaged in many transactions. The hypothesis is that a high percentage of positive ratings will be associated with higher transactions prices.

(3) YEAR OF FIRST TRANSACTION. It is relatively simple to establish an eBay account for the purposes of buying or selling. As a result, there is the constant threat of dealing with disreputable individuals who simply create aliases for fraudulent purposes and disappear into cyberspace after failing to perform. As a check on this, eBay provides background information on buyers and sellers including the year of their first recorded transaction. A long track record of commerce will add credibility to a seller's reputation. In the data set used in this study, seller's ranged between 0 and 10 years of experience. The hypothesis is that buyers will be willing to pay a price premium to deal with well-established sellers having long track records. Conversely, buyers will expect a price discount from sellers who have no established record on eBay.

(4) NUMBER OF TRANSACTIONS. In addition to the length of time that a seller has participated on eBay, buyers may also prefer to deal with sellers who have engaged in many transactions. EBay provides background information on the cumulative number of deals that a seller has participated in. The hypothesis for this variable is similar to the other three measures of seller trustworthiness. Buyers will be willing to pay a price premium to deal with sellers who have an extensive track record on eBay. Conversely, they will expect a price discount if they are dealing with an unknown quantity.

NUMBER OF BIDS. The conventional wisdom on eBay is that sellers should make every effort to attract attention to their auctions and create a critical mass of bidders. As will be discussed later, there are a variety of ways to attempt to do this. The objective is to ignite a bidding war that encourages bidders to offer the highest prices they are willing to pay. Data was collected on the total number of bids for each of the 147 auctions in this study. This number varied between 1 and 28. The hypothesis is that auctions with a larger number of bids have succeeded in attracting the attention of many possible buyers and will lead to a higher final price. However, there is a counter strategy from buyers that could mitigate this effect. The most common pattern for bidding on eBay is to have a relatively low starting bid that is followed by gradually increased prices up until the final moments of the auction. The last few minutes is typically when the real action starts. The bids generally escalate rapidly until the final few seconds. Some buyers will try to derail this process by making a somewhat high preemptive bid in the early period of the auction. This can discourage other potential bidders from even considering getting involved as they might assume the product is getting expensive early on. The bargain hunters will move on to other auctions for similar products and opt out of becoming participants in a bidding war.

SHIPPING COST. After winning an auction, the buyer must pay for the product to be shipped. The seller can set any shipping cost desired. In some cases, free shipping is offered as an enticement to attract attention. In other cases, sellers will treat the shipping as a profit center and charge an amount well in excess of the actual cost. For example, the shipping costs charged for the auctions in this data set varied between 0 and $15.27 (with a mean of $5.98). In an efficient market, neither of these seller strategies should have much impact. A rational buyer should only care about the total price paid (sales price plus shipping cost) and would be indifferent between the split between them. Previous research has suggested that buyers might be fooled by this tactic (Hossain and Morgan, 2006). A low sales price might encourage buyers to feel they have found a bargain, even if the savings is immediately spent on higher shipping. The hypothesis in this study is that buyers are rational and will be indifferent to artificially high or low shipping costs.

NEW. This study analyzed both new DVDs and resales. As such, it was important to control for the condition of the product. A dummy variable of (0,1) was used to distinguish between new and used discs. About half of the transactions were for new DVDs. The coefficient on this variable will indicate the price premium for new discs over used ones.

FOREIGN. About 15% of the transactions in this study originated from abroad. This raises another dimension of trust: whether or not the DVD is a legitimate copy. Some of the foreign DVD sellers on eBay are notorious bootleggers. Many buyers will shun these auctions in principle while others may be willing to buy from abroad if the price is right. The foreign discs also have the reputation of having format problems. DVD formats vary around the world. Many of the foreign sellers are located in China and while they may describe their products as compatible with North American DVD players, it isn't always the case. The hypothesis in this study is that buyers will expect a price discount for DVDs that originate outside of the United States.

BUYER EXPERIENCE. Just as the seller track record can affect price, so can the experience of the buyer. Sophisticated buyers are more likely to employ tactics that lead to lower final prices, such as making last minute bids (sniping) or doing extensive comparison shopping. EBay provides bidder background information including the number of transactions each buyer has completed in the past. This variable was used as a measure of buyer experience and sophistication. In this data set, buyer experience varied between 0 and 806 transactions, with a mean of 83. The hypothesis is that auctions won by experienced buyers are more likely to have lower final prices.

LENGTH OF AUCTION. As part of establishing an auction, the seller can choose how many days to list the product. The conventional wisdom is that auctions must be long enough to allow as many potential buyers as possible to browse across your item. However, there is a risk that impulse buyers will shun auctions that will not close for many days. Since much of the bidding activity takes place in the final hours of an auction, it is not clear that additional days of listings will contribute much to the final price. In this study, the length of auctions varied between 1 and 10 days. The hypothesis is that longer auctions will provide a greater opportunity to attract buyers and will result in higher final prices.

STARTING BID. Sellers are also given the opportunity to specify a starting bid. On their web site, eBay encourages sellers to set a low starting bid as a way to attract attention. They warn that asking for a high starting bid can choke off interest in the product and lead to lower final prices. However, it is not clear that buyers who are attracted to artificially low prices are likely to end up bidding at higher levels. These bargain hunters may simply fade away as soon as prices approach competitive levels. On the other hand, once a specific auction has entered the evoked set of possibilities that buyers are actively monitoring, there is the possibility that they will become vested in the transaction and contribute to a bidding war. In this data set, the starting bids set by sellers varied between $.01 and $40. The hypothesis is that a low starting bid will attract the attention of more potential buyers and ultimately lead to a higher final price.

SNIPE. Perhaps no other phenomenon on eBay has received as much attention as "sniping"-the practice of waiting until the very last seconds of an auction and bidding slightly above the going price to steal away an item. There are even specialized software programs and web sites to automatically do this for buyers. Such behavior can incur the wrath of other bidders, but does it lead to better prices? In this study, a dummy variable was assigned a value of 0 or 1 depending on whether or not the winning bid was submitted in the final two minutes of the auction. Approximately 37% of the auctions were "sniped" in this data set. The hypothesis is that an auction that received a winning bid during the final two minutes will have a lower final price.

PICTURES. EBay encourages sellers to do everything they can to attract attention. They offer additional promotional opportunities, such as including photos of the product as part of the listing. In this study, the product is relatively standardized, so the inclusion of photos might not have a strong effect. However, anything which encourages more buyers to visit or monitor a particular auction can contribute to a higher closing price. Prior studies have shown a relationship between the inclusion of auction pictures and the creation of a herd instinct that can result in higher prices (Stern and Stafford, 2006). The hypothesis is that auctions that included photos of the product are likely to exhibit higher final prices.

RESULTS

The preceding variables were analyzed using multivariate regression. Some typical results are shown in Table 1. Approximately 43% of the variation in the dependant variable (PRICE) can be explained by the six included independent variables (NUMBER OF BIDS, SHIP COST, NEW, STARTING BID, SNIPE, SELLER % POSITIVE). All of these variables were significant at the 95% confidence level. Most of the findings are consistent with the hypotheses described previously.

The results from the variables relating to seller trustworthiness are of particular interest. Four different measures of trust were used (SELLER % POSITIVE, SELLER SCORE, START YEAR, # OF TRANSACTIONS). Of these, the percentage of positive ratings received by sellers was by far the most significant. Each of these measures addresses a slightly different aspect of trust. Of the four, SELLER % POSITIVE is the only one that is sensitive to the proportion of negative ratings that a seller has received. The other variables are more dependent on volume and duration of seller experience. This means that sellers must be very careful not to alienate buyers during their dealings. Any blemishes on their record that lower their percentage of positive scores will lead to lower selling prices.

The importance of attracting a large number of bids was also demonstrated. There was a strong relationship between the number of bids and the final price. This supports the hypothesis that sellers need to attract the attention of buyers and to encourage them to submit bids, even if they are low. The herding behavior identified in other studies seems to play an important role in price determination. This may also explain the findings on starting bids. It appears to be helpful to have a lower starting bid as a means of attracting attention to the auction.

The findings on shipping cost are somewhat surprising. If buyers are rational, there should be a one-to-one discount between higher shipping costs and lower selling prices. This isn't quite the case. The coefficient on shipping costs is -.937, indicating that an additional $1.00 of shipping cost will only reduce the sales price by 94 cents. In other words, the seller would be wise to offer products at artificially low selling prices and to recoup the difference in inflated shipping costs. There is approximately a 6% profit gain from shifting the revenue from sales price to shipping cost. This may explain why it is common to see sellers employing this tactic on eBay, despite the readily observable nature of this strategy.

Not surprisingly, new DVDs sell for more than used ones. However, the difference is not very big. On average, new DVDs only command a $2.20 premium over used ones. This might be due to the fact that the new DVDs sold at such low prices on eBay relative to retail and traditional outlets. The average sales price of the new DVDs in this study was 47% below the retail price and 29% below the price charged by bricks and mortar outlets.

It also appears that sniping does work. In fact, it creates a significant discount for the buyer when used. The auctions which had a winning bid during the final two minutes would appear to provide a $2.22 reduction in the final price. This is a very significant percentage of the entire product price. Buyers are well advised to control their impulses to bid until the very end of an auction. Patience pays off.

There were several non-findings in this study. Some of the variables tested did not prove significant. Surprisingly, there was no discount for foreign DVDs when shipping costs are controlled for. Buyer experience did not play a direct role, although it may have been picked up by the sniping variable. Several of the reputation variables seemed weakly related to price. In general, it appeared that buyers were more sensitive to reliability (as measured by percentage of positive ratings) than to seller experience or track record. Length of auction and inclusion of pictures were not significant.

CONCLUSION

EBay has become a major force in e-commerce. They have developed a brilliant business model that has become part of the lives of millions of consumers. This study has analyzed the factors which determine final transaction prices in the eBay marketplace. In general, many elements of the conventional wisdom were validated. There are specific tactics that sellers can employ to improve the prices they receive. Likewise, buyers can affect final prices through the use of sniping. There are several ways in which future research could extend the results of this study. A compelling theoretical model of the online bidding process has still not been developed. This study controlled for product attributes by examining a single product. Future cross-sectional studies will enhance our understanding of the interplay between product complexity and the bidding process. The role of trust and reputation needs further investigation. The emergence of online markets has created a paradigm shift in how buyers and sellers assess each others attractiveness as partners. EBay's Feedback Forum was an innovative first step down this road, but there is certainly room for better tools to improve the efficiency of online commerce. There are now over 1,600 online auction sites (Brint, 2003) that are seeking better mechanisms to overcome buyer and seller reluctance. Future research will improve our understanding of the dynamics of online marketplaces.

REFERENCES

Akerlof G. (1970). The Market for Lemons: Quality Under Uncertainty and the Market Mechanism. Quarterly Journal of Economics, (84), 488-500.

Ba, S., Whinston, A. B. & H. Zhang (1999). Building Trust in the Electronic Market Using an Economic Incentive Mechanism. Proceedings of the 1999 International Conference on Information Systems, Charlotte, NC.

Ba, S. & M. Pavlou (2002). Evidence on the Effect of Trust Building Technology in Electronic Markets: Price Premiums and Buyer Behavior. MIS Quarterly (26:3), 243-268.

Brint, A. (2003). Investigating Buyer and Seller Strategies in Online Auctions. Journal of the Operational Research Society, (54), 1177-1188.

Brynjolfsson, E. & M. Smith (2000). Frictionless Commerce? A Comparison of Internet and Conventional Retailers, Management Science (46:4), 563-585.

Dellarocas, C. (2003). The Digitization of Word of Mouth: Promise and Challenges of Online Feedback Mechanisms. Management Science (49:10), 1407-1424.

Dewan, S. & V. Hsu (2002). Adverse Selection in Reputation-Based Electronic Markets: Evidence from Online Stamp Auctions. Working Paper, University of California, Irvine.

Fung R. & M. Lee (1999). EC-Trust: Exploring the Antecedent Factors. Proceedings of the Fifth Americas Conference on Information Systems, Milwaukee, WI, 517-519.

Hof, R. D.(August 25, 2003). The eBay Economy. Business Week , 125-128.

Hossain, T. & J. Morgan (2006). Advances in Economic Analysis and Policy.

Lee H. G. (1998). Do Electronic Marketplaces Lower the Price of Goods? Communications of the ACM (41:1) 73-80.

Lee, Z., Im, I & J. Lee (2000). The Effect of Negative Buyer Feedback on Prices in Internet Auction Markets. Proceedings of the ICIS, 286-287.

Melnik, M. I. & J. Alm (2002). Does a Seller's Reputation Matter? Evidence from EBay Auctions. Journal of Industrial Economics, (50:3) 337-349.

Resnick, P., Zeckhauser R., Friedman, E. & K. Kuwabara, (2000). Reputation Systems. Communications of the ACM, (43:12) 45-48.

Resnick, P. & R. Zeckhauser (2002). Trust Among Strangers in Internet Transactions: Empirical Analysis of EBay's Reputation System. The Economics of the Internet and E-Commerce. Advances in Applied Microeconomics, (11) JAI Press.

Sellers, P. (May, 2004). EBay's Secret, Fortune, (150:8) 160-1267.

Stern, B. & M. Stafford (2006). Individual and Social Determinants of Winning Bids in Online Auctions. Journal of Consumer Behaviour, (5:1), 43-56.

Swan, J. E. & J. Nolan (1985). Gaining Customer Trust: A Conceptual Guide for the Salesperson. Journal of Personal Selling and Sales Management (5), 39-48.

Walden, E. (2000). Some Value Propositions of Online Communities. Electronic Markets, 10, 4.

Michael Hergert, San Diego State University 

Table 1: Determinants of Auction Prices

Dependent Variable: PRICE

Independent Variables      Coefficient   t Statistic   Significance
CONSTANT                     14.460         3.57           .001
NUMBER OF BIDS                 .836         7.31           .000
SHIP COST                     -9.37         4.59           .000
NEW                           2.198         2.34           .021
STARTING BID                  -.442         6.71           .000
SNIPE                        -2.220         2.30           .023
SELLER % POSITIVE              .068         1.97           .050
Adjusted R Squared: .43%