Please enable JavaScript.
Coggle requires JavaScript to display documents.
Internet Exchanges (Forman, Ghose and Goldfarb 2009 (Methodology (Books: A…
Internet Exchanges
Forman, Ghose and Goldfarb 2009
Research Question
Does the Solow, Balasubramanian model be tested?
Tradeoffs between disutility of distance (in brick and mortar) and disutility of inspection, shipping and returning (in online)
H1: As distances to offline stores decrease, the likelihood of purchasing a commodity product online decreases
H2: As distance to offline store decreases, online purchases are more likely to be affected by specialized book stores than by discount stores. (Larger selection reduces transport cost)
Methodology
Books: A commodity product, allows a focus on location-related factors.
Regression of localtop10 on #localstores has OVB issues. Solution: To use store entry. (I think the issues that cause OVB in the 1st case will also cause OVB in the second case).
DID: Outcome = (Store Entry) + (NationalTop10) + (Store Entry)* (NationalTop10)
Assumptions: parallel Trends, No anticipatory effects.
Clustered Standard Errors: Location-Month similarities
Findings
People substitute away from online purchasing to offline purchasing when a store opens locally, particularly with bestsellers
However, the breadth of the product line at a local retail store does not affect purchases
Hypothesis Building
Circle, with online in the middle - fixed disutility costs. Customers uniformly distributed on the circle, distance from nearest retailer influences disutility of offline. Customers maximize utility of purchase
Salop: Fixed dis-utility costs from etailer(lack of instant gratification, shipping cost, inability to assess quality), and transportation costs from local store
All else equal, reduction in distance from a local store increases utility of purchasing from offline.
Uncertainty of stocking a niche product offline adds to the transportation cost. Here reduction in distance to offline store has less of an effect on consumer buying online.
Data
Amazon top 10 in each locality - scraped. National rank, relative price. Discount store entry and bookstore entry.
Brynjolfsson et al 2009
Research Question
Does the level of competition between internet retailers and traditional retailers differ across popular and niche products?
Internet retailers reduce search costs for niche products. Brick and mortar companies reduce search costs for popular products.
So, the internet retailer should face a lower level of competition from traditional stores when selling niche product versus popular
Findings
Insignificant effect of #stores on niche product sales online
Suggest internet focus on niche products and brick and mortar to focus on niche products to improve profits for both
Methodology
Data: Clothing Retailer sales records for catalog and internet and yellow pages listings for brick and mortar clothing retailers in the zip code.
Logistic regression: internet sales on #local stores - significant at 5%
Now, divide products by popular and niche by 80:20 rule. Now, regression is insignificant for niche products but significant for popular!
Endogeneity concerns: Simultaneity, or omitted variable socioeconomic factor affecting both #stores and internet sales. Solution: IV: Stores in 1994.
IV Conditions: 1) Correlated with regressor 2) As good as randomly assigned - no unobservables that affect both instrument and y (Exclusion Restriction) 3) Exclusion Restriction - z affects y only through x.
Logistic Regression
Odds1/Odds2 = e^\beta
Since y = 0 or 1, when we run a standard regression, we will get all kinds of decimal numbers. We need to convert this into a "probability that Y will be 1 for each given level of X.
Hypothesis building
Online characteristics - online disutility costs, Search costs, centralized warehousing, inexpensive warehousing, drop shipping, unlimited geographic dispersion, electronic delivery of products, offline transportation costs.
Physical store stocks based on sku demand - popular only - newsvendor model
e-tailers have dramatically lower inventory costs, to a large geography, they can stock niche products too profitably
Search costs at a physical store for an e-tailers niche products are high - the physical retailer may not carry it, or it may be difficult to find it in the store
Ghose, Smith and Telang 2006
Research Question
Are the sales of old books on amazon cannibalizing the sales of new books?
Impact on publisher, amazon and consumer welfare
Methodology - Analytical
Amazon provides a side-by-side comparison on commodities - making it easier to pinpoint the causal relation
Assuming a downward sloping demand curve
Find the demand and profits when only new book market
Now, specify this with used books available, to find substitution effect, price increase effect and used book demand effect
Buy used: Theta.q-Pu; Buy new and sell: Theta -P +alpha(1-k)Pu
Substitution Effect = Cross-Price elasticity of new books
(Number of percentage points change in used prices)
Demand of new books prior to entry of used books
Findings
16% of used books sold cannibalize new books. So while publishers lose some, the gains to amazon and consumers from the used book market more than make up for it.
same-price elasticity of new books ~ -1; for used books: ~-4.
Methodology - Empirical
Data
273 books from 5 categories of books, xml feed from amazon.com
book, price, condition of lowest priced book, amazon price, days since release, seller rating of lowest priced book, #books
Spec
1) log(rank) = b0 + b1.log(amazonPrice) + b2.log(usedPrice) + b3.Controls +e. Used price is the lowest available price - so its cross price elasticity of lowest priced book, not all used books.
Controls = log(time since book release), condition of lowest priced used book and their seller rating, log of number of used books offered.
2) (Whether a used book sold) = b0 + b1
(used price)+ b2
(very good)+b3*(good condition), etc, seller average rating, #seller ratings,
b1, b2 is the own-price and cross-price elasticity for new books
Since dependent variable is 0/1, the b1 needs to be transformed to find same-price elasticity.
Now, chug through a bunch of mundane math to find gains to amazon, publishers and users.
Fleder and Hosanagar 2009
Research Question
What is the impact of recommender systems on sales diversity?
Methodology
The customer chooses one product per time step. Two products available, white and black with consumer probability (p, 1-p)
Firm recommends a product that is accepted with a probability r. This updates urn2.
Findings
Individual purchase diversity can increase while overall sales diversity can reduce.
Standard recommender systems like collaborative filtering may cause a rich-get-richer effect. This means recommender systems that focus on increasing diversity may be better.
