«= = Working Paper = = Online Auctions and Multichannel Retailing Jason Kuruzovich Lally School of Management and Technology Rensselaer Polytechnic ...»
Online Auctions and Multichannel Retailing
Lally School of Management and Technology
Rensselaer Polytechnic Institute
Assistant Professor of Technology and Operations
Stephen M. Ross School of Business
University of Michigan
Ross School of Business Working Paper
Working Paper No. 1176
This work cannot be used without the author's permission.
This paper can be downloaded without charge from the
Social Sciences Research Network Electronic Paper Collection:
Key words: Online auctions, multichannel retailing, electronic commerce, channel management
1. Introduction In many markets that involve unique items—i.e., used cars, used books, real estate, and antiques—sellers often use several selling channels in parallel. For example, a seller may try to sell an item through her brick and mortar retail location while simultaneously listing it in an online auction site such as eBay, her own website, and other popular third-party websites which facilitate interactions between buyers and sellers (i.e., Cars.com or the Amazon Marketplace). This multichannel strategy is facilitated by the Internet, which decouples the information component of transactions from the logistics component (Van Heck and Ribbers 1997). The Internet enables retailers to reach out to a broader group of customers through a variety of online channels, while offering the product to local customers through a traditional offline location. Given the ease by which even very small retailers can establish online presences utilizing online marketplaces such as Amazon or eBay, it is increasingly important to develop an understanding of sellers’ practices in multichannel contexts.
Online auctions have gained enormous popularity. The eBay marketplace alone facilitates over 64 billion dollars in merchandise per year and involves over 95 million active users (eBay 2011). Online auctions like those conducted on eBay attract heterogeneous sellers, and important auction outcomes, such as when sales occur and the sale prices, can be easily observed. Past research has found that seller’s rating, the existence of a buy-now price, and the starting bid—i.e., seller and auction characteristics observable by bidders—are associated with both the auction sale price and the likelihood of sale (Ba and Pavlou 2002; Livingston 2005; Ottaway et al. 2003). Recent research has begun to identify ways in which contextual characteristics of the auction channel not directly observed by bidders, such as the existence of concurrent or prior auctions for the same or similar item(s), can influence auction outcomes (Arora et al.
2002; Bapna et al. 2003; Kuruzovich et al. 2010). However, it is yet unknown how auction outcomes might be influenced by the seller’s simultaneous use of non-auction selling channels.
To address this gap in the literature, this paper utilizes both analytical and empirical approaches to show that seller characteristics related to the demand in non-auction channels influence the auction channel outcomes. Specifically, we use the theoretical foundation of search theory (Diamond 1985) to analytically model how seller characteristics that affect the volume or the distribution of offers received (i.e., the demand) in non-auction channels influence the probability a given auction ends in a sale (the auction-sale probability), the probability a given item the seller lists in the auction channel is sold through the auction channel (the item-sale probability), and the expected sale price (sale price) when an auction ends in a sale.
Search theory (Ashenfelter et al. 2003; Genesove 1995; Genesove and Mayer 2001) predicts that sellers with better ability to make a sale in the non-auction channels would set higher reserve prices in both the auction channel and the non-auction channels, resulting in two opposing effects on the probability an auction ends in a sale and the probability a given item sells via the auction channel.
Specifically, an increase in the non-auction reserve price would increase both probabilities, while an increase in the auction reserve price would decrease them. Thus, the combined effect on the auction channel outcomes is not clear in advance. Our analytical model allows us to derive predictions regarding the nature of the combined effect. We show that the relationship between such seller characteristics and the auction outcomes depends upon whether the seller manages the auction and non-auction channels jointly (i.e., setting reserve prices in the different channels to maximize combined profit) or separately.
We then empirically test our analytical model’s predictions with specific seller characteristics related to the quality of the seller’s retail location and her electronic commerce capabilities in non-auction online channels. The empirical analyses use outcomes from 43,461 online auctions conducted on eBay Motors by 296 multichannel retailers for 21,630 unique vehicles. The analysis confirms that a seller’s retail location and her electronic commerce capabilities significantly impact the above three auction channel outcomes. Specifically, a better retail location or improved electronic capabilities in non-auction channels lead to a lower likelihood of the reserve price being met and a sale occurring in an individual auction (auction-sale), a lower likelihood of a vehicle being sold through the online auction channel (item-sale), but a higher price in the auction channel when a sale occurs (sale price). Overall, our results suggest that sellers engage in the joint channel management strategy examined in the analytical model and may raise their online auction reserve price in response to better sales opportunities in other channels.
This research contributes to the understanding of how sellers use online auctions alongside other channels, an area in which there has been little past research. Consideration of all three channel outcomes (auction-sale, item-sale, and sale price) in the analytical and empirical treatments enables a more complete understanding of sellers’ strategies and the cross-channel effects in a multichannel context.
While prior work examines consumer behaviors in multichannel contexts (Ariely and Simonson 2003;
Etzion et al. 2006; Kumar and Venkatesan 2005; Venkatesan et al. 2007) and seller behaviors in online and offline channels (Brynjolfsson and Smith 2000; Forman et al. 2009; Overby and Jap 2009), this is the first paper to present a theoretically-grounded treatment of how sellers’ characteristics related to demand in the non-auction channels affect their online auctions’ outcomes. Thus, this research provides insights into how retailers use auctions as part of a multichannel strategy, answering calls for more research on online auctions (Pinker et al. 2003) and multichannel seller strategies (Neslin and Shankar 2009).
The rest of the paper proceeds as follows. We first review how search theory has been used to study outcomes of the auction channel and present the theoretical framework used in the paper. Next, we present the analytical model that explores the relationship between characteristics of the demand in nonauction channels and auction channel outcomes. The predictions from the analytical model are then tested using data from eBay Motors. Finally, implications for theory and practice are discussed.
2. Theory This section first reviews how search theory has been used to study auction channel outcomes. It then further discusses how seller characteristics related to the demand in non-auction channels are expected to influence the auction channel outcomes of interest.
2.1. Search Theory and Online Auctions This research utilizes search theory to understand seller behavior in online auctions as part of a multichannel strategy. Search theory specifies that the process of search is influenced by both the potential benefits and the costs of continuing the search (Diamond 1985). When applied to the study of online auctions, search theory can explain a seller’s rationale in setting her reserve price (Ashenfelter 1989; Genesove 1995). Setting a high reserve price decreases the chance that the reserve price will be met and the auction will end in a sale, but increases the minimum sale price when a sale occurs. Conversely, lowering the reserve price increases the likelihood of sale, but decreases the minimum sale price. In other words, the seller faces a tradeoff between the sale price and the time required to find a buyer, making search a relevant foundation for understanding the process through which sellers set their reserve price.
Ashenfelter (1989) first suggested that search theory may explain how sellers set and then adjust their reserve price over time. Genesove (1995) empirically tested a model of search in the context of wholesale auto auctions, examining how the mean and variance of market prices for a vehicle influence the search process of sellers in the auctions. Building on this work, Ashenfelter et al. (2003) modeled the sale rate of art as a search process, examining how sellers’ reserve prices depend upon market characteristics (i.e., the variance in prices). Genesove and Mayer (2001) found that sellers who originally paid higher prices for condominiums also set higher asking prices when subsequently selling, showing that the reserve price and the time on market are influenced by seller-specific characteristics. Kuruzovich et al. (2010) examined how seller search across sequential online auctions influences the price the seller obtains for an item.
Overall, this prior research suggests that search theory can be useful to understanding how sellers set reserve prices when using online auctions alongside other channels. Just as consumers may search for the right product and price across the variety of channels available to them (Ratchford et al. 2003; Ratchford and Srinivasan 1993), sellers of unique items can be conceptualized as searching for a high-valuation buyer across multiple channels, trying to maximize the difference between the price they would obtain and the cost of search (Genesove 1995; Kuruzovich et al. 2010).