Internet Exchanges for Used Books

      [Internet Exchanges for Used Books: An Empirical Analysis of Product Cannibaliztion and Welfare Impact]
 
  #Intro, Keywords
-degree to which used products cannibalize new-product sales for books 
-books are one of the most prominent used-product categories sold online
-data set collected from Amazon's new and used book marketplaces 
-Jeff Bezos: found that used books business does not take business away from sale of new books, opposite happed
: offering customers lower-priced option causes them to visit site more frequently, leading to higher sales of new books
: when customer sells used books, gives them a budget to buy more new books
-analyze impact of secondary markets on welfare to publishers, retailers, and consumers
-when consumer sells used book,
1) retailer gets commission k per used good sold
2) rest (1-k) is the gain to the consumer 

  #Data
-Amazon publicly available information on new and used book prices, sales ranks
-1st sample: 180 day period: Sep 2002-Mar 2003 = 273 individual book titles 
Category: NYT bestsellers, former NYT bestseller, new and upcoming books 
-2nd sample: 85 day period: Aprit 2004-July 2004 = 41,994 obserservations of 393 titles
-data includes: price, condition, seller rating 
-only in 2nd sample, unique identifier exists - can know if sold or not 
Sales of book
-before hard to measure, quantity sold unobservable
-stable relationship between ordinal sales rank and cardinal number of sales 
-calibration of beta=-0.8555 (purchase several copies and track sales rank before/after purchase)
-Sales rank calculated based only on new-book sales

  #Model
  1) 1st regression
-y: log of rank 
-x: Amazon price, lowest-priced used book in market 
-control: log of time since book released, condition of lowest-priced used book, seller rating for lowest-priced used book,
log of number of used books offered for sale for particular book 
-no endogeneity: quantity of selling books predetermined, quantity/price not jointly determined

  2) 2nd regression: multinomial logit model 
-observe each of the marketplace offers shown to customers with offer each customer chose to purchase
-OLS with book level fixed effect

    #Result
 -Amazon price: positive/significant effect on rank
 -Use book price: negative/significant effect on rank = no cannibalization (opposite)
 -Condition rating/seller rating controls: insignificant in driving sales of new books
 -Condition rating/seller rating controls: positively significant in choice of used-book 
 *Used books are poor substitutes for new books 
 -cross-price elasticity of new-book demand w/ respect to used-books prices = 0.0888
 -16% of used-book sales at Amazon cannibalize new-book purchases 
 -increase in book readership from used-book marektplace - increases CS $65.67 million annually
 -$45.05 million loss in public welfare, #65.76 million increase in Amazon's profits
 -increase in total welfare to society of $87.92 million annually from used-book markets