Geography and Electronic Commerce: Measuring Convenience, Selection, and Price Anindya Ghose (with Chris Forman and Avi Goldfarb) 2007 NET INSTITUTE CONFERENCE
Introduction • What are some reasons consumers buy online? • Price • Convenience • Selection/Variety • Each of these components is largely driven by the local availability of offline options
What we do • Use the existing theory literature to motivate a set of hypotheses on online/offline channel substitution Use to show how changes in retail supply influence online/offline channel substitution • Examine data from Amazon “purchase circles” • Merge with data on store openings across locations
Why might this be interesting? • How do consumers substitute between online and offline channels? • Prior research has focused primarily on effects of price • Little systematic research on other factors affecting consumer purchase behavior in online world. • If you are a online retailer, how does offline entry influence online consumer behavior in you local market? • If you are an offline retailer, what kind of competition do you face online?
Hypotheses are motivated by theory literature on spatial competition • Consumers are distributed around unit circle • Incur transportation cost t for visiting offline retailer • Consumers incur disutility cost µ when purchasing online • Choice of channel is determined by location, t, µ, pd and pr Source: Balasubramanian (1998) • Three types of products • Popular products: stocked by all offline retailers • Less popular products: stocked by some offline retailers • Unpopular products: stocked only by online retailer
Hypothesis 1: Online purchasing for convenience • H1: As distance to offline stores decreases, the proportion of popular products being bought online also decreases. • Rationale: For popular products, increases in the distance to the closest offline retailer will increase the likelihood that a consumer will purchase from the direct (Internet) retailer, other things equal.
Hypothesis 2: Online purchasing for selection • H2: As availability of less popular products in offline markets increases, the proportion of less popular products bought online decreases. • Rationale: As product selection at offline retailers increases, consumers will be able to increasingly purchase less popular products online.
Hypothesis 3a & 3b: Online purchasing for price • H3a: As the discounted online price of a commodity falls relative to its list price, consumers are more likely to buy it. This increases the relative online sales of that commodity compared to other online items. • Rationale: Demand curves are downward sloping. • H3b: As distance to offline stores falls, discounted online items will be more likely to be bought offline. • Rationale: Items that are typically discounted online are also discounted offline. Thus, the impact of online discounts will be tempered by the existence of local retail stores with similar discounts.
Amazon • Entry Data Data sources • Entry data from Wal-Mart, Target, Barnes and Noble, and Borders • Collected from news wires and from company reports • Amazon Purchase Circle Data • Collected over 10 months from April 2005 to January 2006 • 1501 unique locations in the US Our dependent variable is a binary variable that is equal to one when the book appears among the top 10 books in that location.
Data • Product characteristics • Amazon price and list price • Overall Amazon rank (splined) • Results robust to USA Today rankings • Average rating and # of reviews on Amazon • Days since product launch • Store Openings • Discount Store Entry: Wal-Mart and Target (16.4%) • Large Bookstore Entry: Barnes & Noble and Borders (4.6%) • Location and date • Identify places within 5.4 miles of opening • Robust to use of 20 mile radius
Econometric strategy • Identification: • The product-location fixed effect controls for location-specific preferences for each product • Difference-in-difference: • Relative effect of entry in one location compared to another location • Key idea is that entry will impact the relative popularity of the books purchased online.
Results are robust to… • Using distance of 20 miles (rather than 5.4 miles) to define “local” • Defining the total set of books available in a month as 1000 not 300 • Using different breaks in the spline of the overall Amazon sales rank • Using a linear (not splined) definition of the overall Amazon sales rank • Interesting distinctions in subsamples: small/large markets, sales tax/no sales tax
What we find: geography matters Goal Findings • Consumers buy different products in different locations • Consumers in different locations use the online channel alternatively to obtain better convenience, selection, and price • Demonstrate that geography matters for electronic commerce • Offline transportation cost, online fit, and prices interact to determine channel choice. • Identified potential extension arising from retailer decisions to stock less popular products. • Provide empirical support for assumptions used in direct marketer circular city model • Demonstrate how convenience matters for electronic commerce • Convenience is the single biggest factor influence online/offline channel substitution in our data
Limitations • Only observe the top 10 products in each location. • This limits our ability to make inferences about very rare products. • This is not a “Long Tail” paper. • We lack information on consumer choice of channel. • Our conclusions are about relative popularity, not about quantities sold. • We use “within” market variation over time to make conclusions about “between” market variation
Conclusions • Geography matters for electronic commerce • The virtual world improves consumer welfare by easing some constraints of the physical world • BUT, the importance of these constraints varies across locations • We examine convenience, selection, and price. Prior literature has focused on price alone. • Strong evidence of convenience and price effects. • Selection effects only in university towns and large cities and taxable states • We do NOT identify the “long tail”