`Case 2:20-cv-00424-RAJ Document 100-6 Filed 04/06/23 Page 1 of 24
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`EXHIBIT F
`EXHIBIT F
`
`
`
`Case 2:20-cv-00424-RAJ Document 100-6 Filed 04/06/23 Page 2 of 24
`
`Price Parity Clauses and Product Prices:
`
`Evidence from Amazon and eBay
`
`
`
`Yu Song
`
`University of Michigan
`
`yyusong@umich.edu
`
`Abstract
`
`Price parity clauses (PPCs) have been adopted by many digital platforms, raising
`
`antitrust concerns. We study how Amazon’s removal of PPCs in March 2019
`
`affects product prices on Amazon and eBay. We find that, after the removal,
`
`products in categories with greater differences in commission rates between
`
`Amazon and eBay experience greater price reduction on both. The price reduction
`
`on Amazon is greater for products sold directly by Amazon than for those sold by
`
`third parties. These results suggest that Amazon’s removal of PPCs increases
`
`platform competition and reduces product prices.
`
`Keywords: price parity clauses, Amazon, eBay, platform competition
`
`1
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Case 2:20-cv-00424-RAJ Document 100-6 Filed 04/06/23 Page 3 of 24
`
`1. Introduction
`
`With the growing popularity of digital platforms, significant regulatory concerns have emerged,
`
`ranging from competitive price parity clauses (PPCs) (Baker and Scott Morton 2018) to consumer
`
`privacy (Koutroumpis et al. 2020) to the classification of platform participants as employees versus
`
`contractors (Hagiu and Wright 2019) to the enforcement of data portability (Parker et al. 2020).
`
`This paper examines the use of PPCs by dominant platforms. These PPCs—also termed most
`
`favored nations (MFN) provisions—prevent third-party sellers or service providers from charging
`
`lower prices on alternative channels than on the focal platform.1 For example, when Amazon
`
`adopted PPCs, the total price of an item listed by a seller on Amazon was required to be at or below
`
`the price at which the seller was offering that item on any other online sales channel. Such a policy
`
`draws attention from regulators because it could reduce competition, a possible violation of
`
`Section 2 of the Sherman Act.2
`
`In the past few years, PPCs have been the subject of antitrust enforcement in multiple European
`
`countries. For example, the European Commission investigated the practice of Amazon in Europe,
`
`and ended its PPCs throughout the European Union in 2013.3 In December 2018, Senator Richard
`
`Blumenthal recommended that the US Department of Justice and the Federal Trade Commission
`
`investigate whether the leading platforms violated antitrust rules by adopting PPCs. In March 2019,
`
`presidential candidate Elizabeth Warren proposed to break up tech giants Amazon, Facebook and
`
`Google, and suggested that Congress should pass a law banning large companies from operating
`
`
`1 There are two types of PPC: wide and narrow. Wide PPC prevents the provider from setting a lower price on all
`other platforms, including the provider’s own website; narrow PPC prevents the provider from setting a lower price
`on its own website, while leaving prices on other platforms unrestricted. This paper focuses on wide PPC.
`2 See 15 U.S.C. §§ 1-2 (2012). Federal Trade Commission (FTC) enforcement actions would be brought under Section
`5 of the FTC Act, applying Sherman Act principles. See 15 U.S.C. § 45 (2012).
`3 Jacob Kastrenakes (2013), “Amazon lifts EU restrictions barring merchants from setting lower prices elsewhere”,
`The Verge, August 31, https://www.theverge.com/2013/8/30/4676520/amazon-removes-eu-rule-requiring-lowest-
`price-parity-online, accessed September 2020.
`
`
`
`2
`
`
`
`Case 2:20-cv-00424-RAJ Document 100-6 Filed 04/06/23 Page 4 of 24
`
`and owning participants on the same online platform.4 Against this backdrop, Amazon decided to
`
`stop preventing third-party sellers from setting lower prices on other websites.5
`
`In this paper, we investigate the impact of Amazon’s removal of PPCs on product prices on
`
`Amazon and eBay. Amazon is a retailer itself and also operates a marketplace in which third-party
`
`sellers can sell their products directly to Amazon customers. Amazon charges third-party sellers a
`
`commission for each transaction. The commission rate—a percentage of the product price—varies
`
`across product categories.6, 7 Many sellers sell their products on both Amazon and eBay, one of
`
`Amazon’s biggest competitors. eBay charges lower commission rates in many product categories
`
`than Amazon8 and has not imposed PPCs.
`
`We collect data for products on Amazon and eBay from Keepa.com, a market research firm
`
`that tracks historical product prices on these websites, to examine how the removal of PPCs affects
`
`product prices on Amazon and eBay. The removal of PPCs presents an exogenous shock to third-
`
`party sellers on Amazon and eBay. In addition, while Amazon as a retailer may have anticipated
`
`the removal of PPCs, it has no incentive to change prices for products it sells before the official
`
`removal.
`
`Our identification strategy exploits varying treatment intensity as a result of the differences in
`
`Amazon and eBay commission rates across product categories. We also ensure that the
`
`
`4 Elizabeth Warren
`(2019), “Here’s how we can break up Big Tech,” Medium, March 9,
`https://medium.com/@teamwarren/heres-how-we-can-break-up-big-tech-9ad9e0da324c, accessed September 2020.
`5 David McCabe (2019), “Amazon to end price practice critics said could violate antitrust law,” Axios, March 11,
`https://www.axios.com/amazon-price-practice-antitrust-elizabeth-warren-d802ba71-d376-4316-b9dc-
`cca4540959ac.html, accessed September 2020.
`6
` See
`different
`for
`rates
`Amazon’s
`commission
`https://sellercentral.amazon.com/gp/help/external/200336920, accessed September 2020.
`7 For example, one company might sell an office chair at the final price of $100 on Amazon. An office chair is in the
`category of office products, which has a commission rate of 15%. Thus, this company needs to pay $15 to Amazon as
`commission, whereas Amazon itself selling the same chair does not need to pay commission.
`8 See eBay’s commission rates for different categories at https://www.ebay.com/help/selling/fees-credits-
`invoices/selling-fees?id=4364, accessed September 2020.
`
`categories
`
`at
`
`
`
`3
`
`
`
`Case 2:20-cv-00424-RAJ Document 100-6 Filed 04/06/23 Page 5 of 24
`
`commission rates on both sites stay the same for these product categories during the study period.
`
`When products have higher commission rates on Amazon than on eBay, then after Amazon
`
`removes its PPCs, third-party sellers would be more likely to reduce their prices on eBay because
`
`their marginal costs of selling on eBay are lower. We take advantage of this varying treatment
`
`intensity to apply a difference-in-differences (DID) approach to examine how Amazon’s removal
`
`of PPCs affects product prices. This strategy allows us to control for common shocks to product
`
`prices during our study period.
`
`Our analysis reveals that Amazon’s removal of PPCs results in a greater price reduction on
`
`eBay for products that are subject to greater differences in commission rate between Amazon and
`
`eBay. The price of products on eBay decreases by 0.72 percent when the commission rate on
`
`Amazon is 1 percent higher than that on eBay. We also find that Amazon prices decrease by 0.63
`
`percent when the commission rate on Amazon is 1 percent higher than that on eBay; this result is
`
`consistent with the two platforms being close competitors. The price reduction on Amazon is more
`
`pronounced when the products are sold directly by Amazon than by its third-party sellers,
`
`suggesting that Amazon is more responsive to price changes in the external environment.
`
`This study contributes to the literature on PPCs. On the one hand, scholars point out that PPCs
`
`can benefit platforms, retailers, and customers. For platforms, theoretical studies have shown that
`
`PPCs can help resolve hold-up problems (e.g., Gans 2012, Ezrachi 2015) and prevent
`
`showrooming, a type of free-riding that undermines the platform’s ability to operate (Baker and
`
`Scott Morton 2018, Wang and Wright 2020). For retailers, PPCs may streamline the way inventory
`
`is managed and price is determined across channels, since the retailer only needs to set one price
`
`for all intermediaries. For customers, PPCs reduce possible perceptions of unfairness. Many
`
`companies use dynamic pricing—setting the price based on individual-level information such as a
`
`
`
`4
`
`
`
`Case 2:20-cv-00424-RAJ Document 100-6 Filed 04/06/23 Page 6 of 24
`
`customer’s browsing history and the time of booking—thereby raising concerns about whether the
`
`price being offered is fair. PPCs help maintain price transparency (Nicolau and Sharma 2019).
`
`On the other hand, the increased price of the products or services can reduce consumer welfare
`
`and raise antitrust concerns. Edelman and Wright (2015) examine consumers’ decisions to
`
`purchase either directly or through platforms and find that PPCs harm consumers by diverting
`
`them from direct channels. Boik and Corts (2016) explicitly consider the fee and price effects of
`
`platform PPCs in a theoretical model. They show that PPCs raise platform fees and retail prices
`
`and that they increase platform profits when aggregate demand is sufficiently inelastic. Johnson
`
`(2017) finds similar results but extends his model to consider differences between the agency and
`
`wholesale models. His finding shows that PPCs may harm consumers and reduce competition
`
`between platforms by reducing their incentives to lower prices. Baker and Chevalier (2012) and
`
`Scott Morton (2012) point out that antitrust enforcement against anticompetitive platform PPCs
`
`could have noticeable benefits for productivity and consumer welfare.
`
`Empirical studies on PPCs are limited and many focus on the impact of the removal of PPCs
`
`on online travel agencies (OTAs) in Europe, where the National Competition Authorities and the
`
`European Union’s enforcement institution brought cases against them. From 2015 to 2017, PPCs
`
`on Booking.com were eliminated in many European countries, including France, Italy, Sweden,
`
`and Germany. Hunold et al. (2018) show that hotels in Germany increased their sales channels and
`
`engagement with multiple OTAs when narrow PPCs were removed on Booking.com. They also
`
`find that price of the direct channel among hotel chains is more often strictly lower than the prices
`
`on all other visible online sales channels following the abolition of PPCs. De los Santos et al.
`
`(2020) study wholesale contracts and agency contracts under bargaining in the e-book sector and
`
`address that reinstitution of PPCs would raise the consumer prices by about three percent and non-
`
`
`
`5
`
`
`
`Case 2:20-cv-00424-RAJ Document 100-6 Filed 04/06/23 Page 7 of 24
`
`fiction books prices by up to nine percent through a counterfactual exercise. Ennis et al. (2020)
`
`study how switching from wide PPC to narrow PPC affects online price differential between OTAs
`
`and the direct channel. They find that the probability that the direct channel is cheaper than OTAs
`
`is increased significantly for mid-level and luxury hotels. Similarly, Mantovani et al. (2021)
`
`examine the removal of PPCs in France in 2015 for hotels listed on Booking.com and find a
`
`significant reduction in room prices in the short run, but a more limited effect in the medium run.
`
`This paper provides empirical evidence beyond the OTA setting. In addition, the special
`
`structure of Amazon (being both a retailer and a platform provider) allows us to observe the
`
`responses of both Amazon itself and third-party sellers to Amazon’s removal of PPCs. Our study
`
`also examines price changes on a competing platform. The prior empirical studies investigate price
`
`changes on OTAs after the removal or change of its PPCs in Europe, but do not look at the prices
`
`on competing platforms, where one would expect to see an immediate effect. Our study examines
`
`price reduction not only on Amazon (the indirect impact of eliminating PPCs) but also on eBay
`
`(the direct impact). Therefore, our findings can enrich our understanding of PPCs.
`
`This paper adds to the growing literature on how regulation affects firms and consumers (e.g.,
`
`Israeli 2018, Calzada and Gil 2019, Lu et al. 2019, Goldfarb et al. 2020). A subset of this work
`
`focuses on platform responses to regulation (e.g., Weyl 2010, Chen and Liu 2011). Scholars have
`
`examined many settings, such as platform firms changing their ad campaigns in response to
`
`privacy regulation (Goldfarb and Tucker 2011), changing e-book prices in response to vertical
`
`price restraints (De los Santos and Wildenbeest 2017), and changing prices and price dispersion in
`
`response to transparency regulation (Ater and Rigbi 2018).
`
`Broadly, this paper relates to competition in platform markets (e.g., Caillaud and Jullien 2003,
`
`Rochet and Tirole 2003, Armstrong 2006, Economides and Katsamakas 2006, Armstrong and
`
`
`
`6
`
`
`
`Case 2:20-cv-00424-RAJ Document 100-6 Filed 04/06/23 Page 8 of 24
`
`Wright 2007, Karle et al. 2020). We find that the prices on Amazon and eBay are interrelated and
`
`that policy changes on Amazon affect prices on eBay. Moreover, in response to eBay’s price
`
`change, Amazon itself—as well as its third-party sellers—adjust their own prices.
`
`The rest of the paper proceeds as follows. Section 2 provides the background. Section 3
`
`develops hypotheses. Section 4 describes the datasets and provides descriptive statistics. Section
`
`5 presents the main empirical results and robustness checks. Section 6 concludes.
`
`
`
`2. Background
`
`Amazon, founded as an online retailer of books in 1994, quickly diversified into electronics, CDs,
`
`video games, home and kitchen, baby products, and many other product categories. Amazon
`
`Marketplace was launched in November 2000, allowing third-party sellers to sell directly to
`
`customers. Thus, Amazon is both a retailer and a platform provider: products on Amazon can be
`
`sold by Amazon and by its third-party sellers.
`
`To help third-party sellers handle back-end operations, Amazon initiated the Fulfilled by
`
`Amazon service in 2006, whereby third parties pay for services but gain flexibility in their selling
`
`practices. The service provides storage, packaging, and shipping assistance, taking the burden off
`
`the sellers, and allows sellers to ship their merchandise to an Amazon Fulfillment Center, where it
`
`is warehoused until sold. When an order is placed, Amazon employees physically prepare, package,
`
`and ship the products.
`
`eBay is another leading online marketplace and one of Amazon’s major rivals. Many sellers
`
`choose to multi-home: they sell on both platforms. Unlike Amazon, eBay does not offer products
`
`itself; it is a pure marketplace connecting buyers and third-party sellers. At the time of this study,
`
`eBay does not offer fulfillment services.
`
`
`
`7
`
`
`
`Case 2:20-cv-00424-RAJ Document 100-6 Filed 04/06/23 Page 9 of 24
`
`
`
`3. Hypothesis Development
`
`Since most products are charged lower commission rates on eBay than on Amazon, their selling
`
`costs are lower on eBay. When Amazon imposed PPCs, third-party sellers selling on both Amazon
`
`and eBay could not sell their products at lower prices on eBay even though they faced lower selling
`
`costs.9 With the removal of PPCs on Amazon, third-party sellers no longer face price floors when
`
`setting their prices on eBay and are therefore likely to reduce the price on eBay. This price
`
`reduction is likely to be larger the larger the difference between Amazon and eBay commission
`
`rates.10 Hence, we hypothesize:
`
`Hypothesis 1a: The greater the difference in a product’s Amazon and eBay commission rates, the
`
`greater its price reduction on eBay is likely to be after Amazon’s removal of PPCs.
`
`Most products in our study have commission rates of 10 percent on eBay (see Table 1). The
`
`differences between Amazon’s and eBay’s commission rates are thus mainly determined by
`
`Amazon’s rates. Therefore, Amazon’s commission rates may also matter when analyzing the price
`
`change after it removes its PPCs. Products with higher commission rates on Amazon have greater
`
`selling cost reduction and thus the removal of PPCs should result in significantly lower prices on
`
`eBay. Therefore, we hypothesize:
`
`Hypothesis 1b: The higher a product’s commission rate on Amazon, the greater the price reduction
`
`
`9 To implement this policy, Amazon’s software scans price on other websites looking for price parity violations, and
`Amazon will take automated action on sellers found to be in violation (e.g., making the product harder to find).
`Moreover, retailers choose to raise price on other websites rather than losing sales on Amazon, because they earn a
`large portion of their sales from Amazon. Source: https://www.bloomberg.com/news/articles/2019-08-05/amazon-is-
`squeezing-sellers-that-offer-better-prices-on-almart, accessed September 2020.
`
`10 We define the difference in commission rates between Amazon and eBay as 𝐶𝑜𝑚𝑚𝑖𝑠𝑠𝑖𝑜𝑛 𝑅𝑎𝑡𝑒,-../0/12/=
`𝐶𝑜𝑚𝑚𝑖𝑠𝑠𝑖𝑜𝑛 𝑅𝑎𝑡𝑒456781−𝐶𝑜𝑚𝑚𝑖𝑠𝑠𝑖𝑜𝑛 𝑅𝑎𝑡𝑒/:6;.
`
`
`
`8
`
`
`
`Case 2:20-cv-00424-RAJ Document 100-6 Filed 04/06/23 Page 10 of 24
`
`on eBay is likely to be after Amazon’s removal of PPCs.
`
`Amazon is known to use algorithms to price products it sells directly (e.g., Chen and Wilson
`
`2017). Its algorithms collect price information for the same product on many other websites. When
`
`sellers reduce their product prices on eBay or other sales channels, Amazon’s algorithms are likely
`
`to detect the price changes and Amazon, as a retailer, is likely to reduce its own prices for those
`
`products. Following Hypotheses 1a and 1b, such a price reduction of a product sold by Amazon
`
`directly is likely to be greater for a product with a greater commission rate difference between
`
`Amazon and eBay or with a higher commission rate on Amazon.
`
`Third-party sellers on Amazon also have incentives to reduce their prices to stay competitive.
`
`If they are single-homing on Amazon or if their products compete with Amazon’s own offerings,
`
`they are willing to lower prices to avoid losing customers to eBay sellers or to Amazon itself.
`
`Overall, the prices of products sold both by Amazon itself and by its third-party sellers are likely
`
`to decrease on Amazon after the policy change. Hence, we hypothesize:
`
`Hypothesis 2a: The greater the difference in a product’s Amazon and eBay commission rates, the
`
`greater its price reduction on Amazon is likely to be after Amazon’s removal of PPCs. Also, the
`
`higher a product’s Amazon commission rate, the greater its price reduction on Amazon is likely to
`
`be after Amazon’s removal of PPCs.
`
`On the other hand, a third-party seller’s incentive to lower prices might not be as strong as that
`
`of Amazon as a retailer. Because Amazon does not change its commission rates after the removal
`
`of its PPCs, the cost structure for selling on Amazon remains unchanged for third-party sellers. In
`
`other words, the commission rate for third-party sellers remain the same while Amazon itself
`
`continues to not pay commissions. Thus, the room for price decrease is smaller for third-party
`
`
`
`9
`
`
`
`Case 2:20-cv-00424-RAJ Document 100-6 Filed 04/06/23 Page 11 of 24
`
`sellers than for Amazon itself. Therefore, we hypothesize:
`
`Hypothesis 2b: Amazon, as a retailer, will reduce its product prices more than its third-party
`
`sellers do after its removal of PPCs.
`
`
`
`4. Data
`
`We collect data from Keepa, which keeps track of the historical prices of products on Amazon and
`
`eBay. We exclude certain categories, such as e-books, videos, and music, as products in these
`
`categories are often exclusively offered by Amazon and thus the removal of PPCs has little impact
`
`on their prices. Ultimately, we randomly collect data in 17 product categories and ensure that
`
`commission rates for these product categories stayed the same during our study period on both
`
`Amazon and eBay.
`
`For every product on Amazon, we collect (a) the Amazon Standard Identification Number
`
`(ASIN), unique to each product; (b) the category, which determines the commission rate; (c) the
`
`price history for Amazon as the seller if Amazon sells the product; (d) the price history for third-
`
`party sellers—specifically, the lowest price among all third-party sellers;11 and (e) the price history
`
`on eBay for those products that are offered or matched on eBay in new condition.
`
`We focus on the prices of products three months before and three months after the month of
`
`Amazon’s PPC removal (March 2019). We calculate the monthly average price—on Amazon and
`
`eBay—of every product in our sample. For products sold both directly by Amazon as a retailer
`
`and by its third-party sellers, we collect pricing data from both channels. To analyze the price
`
`
`11 If the product is sold by multiple third-party sellers, we take the lowest price among them, as we assume that
`consumers are likely to pay the lowest price for a given product. Keepa provides shipping costs for items fulfilled by
`third-party sellers only and for items sold on eBay. We include shipping costs when calculating the final prices of
`these products.
`
`
`
`10
`
`
`
`Case 2:20-cv-00424-RAJ Document 100-6 Filed 04/06/23 Page 12 of 24
`
`changes as a result of Amazon’s removal of PPCs, we keep the product only if its price has been
`
`first tracked by Keepa before March 2019. As third-party sellers can use different IDs when selling
`
`on the two sites, we cannot match the sellers across the sites.
`
`Table 1 provides a summary of products on the two sites in our dataset. Because not all Amazon
`
`products are offered on eBay or can be matched to an identical product in new condition on eBay,
`
`we have data for fewer products on eBay. The difference in the distribution of the products across
`
`categories is likely to capture both the multi-homing tendency of the sellers and the ease of
`
`matching. We also report the commission rates of the product categories on the two sites.12 For
`
`most categories, eBay charges a lower commission rate. We observe more variation in the
`
`commission rates on Amazon than on eBay, which we leverage to identify the PPC effects.
`
`
`
`5. Empirical Analysis
`
`5.1 Analysis of Prices on eBay
`
`We use the following DID model to analyze the price change on eBay:
`
` =𝛽C+𝛽E𝐶𝑜𝑚_𝐷𝑖𝑓𝑓-×𝑃𝑜𝑠𝑡A+𝜆A+𝜃-+𝜀-A, (1)
`𝑙𝑜𝑔 𝑝𝑟𝑖𝑐𝑒-A
`where 𝑖 identifies a unique product and 𝑡 is the month. The dependent variable is the logarithm of
`the monthly average price of product 𝑖 in month 𝑡 on eBay. The variable 𝐶𝑜𝑚_𝐷𝑖𝑓𝑓- is the
`commission rate difference between Amazon and eBay for product 𝑖.13, 14 𝑃𝑜𝑠𝑡A is a dummy
`
`variable equal to 1 if the observation is collected after March 2019 and 0 otherwise.
`
`
`12 Zhu and Liu (2018) pointed out that Amazon appears to set commission rates strategically: The commission rates
`tend to be higher for product categories that are less competitive, in which margins for third-party sellers are higher.
`13 The commission rate for some categories is based on price, which might vary over time, so that a given product’s
`commission rate can be different in different months. To avoid endogenous changes in commission rates, we fix
`
`𝐶𝑜𝑚_𝐷𝑖𝑓𝑓- using 𝐶𝑜𝑚_𝐷𝑖𝑓𝑓- in March of the product 𝑖, regardless of variation in commission rates for some products
`
`caused by price change over the months of our analysis.
`14 When the commission rate difference is negative, we set it to zero because when a product’s commission rate is
`higher on Amazon than on eBay, prices are not likely to affected by Amazon’s removal of PPCs. The results are
`qualitatively the same if we use negative commission rate differences.
`
`
`
`11
`
`
`
`Category
`Unlocked Cell Phones
`Industrial & Scientific
`Appliances
`Electronics
`Baby Products
`Books
`CDs & Vinyl
`Clothing
`Health Care
`Home & Kitchen
`Office Products
`Patio, Lawn & Garden
`Personal Care
`Pet Supplies
`Software
`Sports
`Tools & Home Improvement
`Toys & Games
`Video Games
`Total
`
`
`Commission
`rate (%)
`8
`12
`8/1515
`8/1516
`15
`15
`15
`15
`15
`15
`15
`15
`15
`15
`15
`15
`15
`15
`15
`
`
`Frequency
`2,547
`1,345
`1,987
`75,965
`12,662
`719
`141
`18,436
`47,642
`59,624
`64,548
`6,789
`40,942
`46,467
`161
`59,723
`25,434
`3,451
`40,516
`509,099
`
`% of all
`products
`0.50
`0.26
`0.39
`14.92
`2.49
`0.14
`0.03
`3.62
`9.36
`11.71
`12.68
`1.33
`8.04
`9.13
`0.03
`11.73
`5.00
`0.68
`7.96
`100
`
`Commission
`rate (%)
`10
`10
`10
`10
`10
`12
`10
`10
`10
`10
`10
`10
`10
`10
`10
`10
`10
`10
`10
`
`
`eBay
`
`Frequency
`35
`555
`39
`2,565
`189
`40
`8
`35
`899
`2,438
`1,385
`22
`1,152
`1,358
`4
`1,360
`951
`85
`1,806
`14,926
`
`% of all
`products
`0.23
`3.72
`0.26
`17.18
`1.27
`0.27
`0.05
`0.23
`6.02
`16.33
`9.28
`0.15
`7.72
`9.10
`0.03
`9.11
`6.37
`0.57
`12.10
`100
`
`Case 2:20-cv-00424-RAJ Document 100-6 Filed 04/06/23 Page 13 of 24
`
`Table 1. Distribution of Products across Categories
`Amazon
`
`
`
`To control for possible seasonality that may affect all products, we include a month fixed effect
`
`𝜆A, which absorbs the main effect of 𝑃𝑜𝑠𝑡A. 𝜃- is a product-level fixed effect and absorbs the main
`effect of 𝐶𝑜𝑚_𝐷𝑖𝑓𝑓-. We cluster standard errors at the category level to account for autocorrelation
`
`in the data within categories and over time.
`
`Column (1) of Table 2 shows that a 1-percent increase in the commission rate difference leads
`
`to a 0.721-percent price reduction on eBay. This supports Hypothesis 1a.
`
`
`15 15% for the portion of the total sales price up to $300.00 and 8% for any portion of the total sales price greater than
`$300.00.
`16 15% for the portion of the total sales price up to $100.00 and 8% for any portion of the total sales price greater than
`$100.00.
`
`
`
`12
`
`
`
`Case 2:20-cv-00424-RAJ Document 100-6 Filed 04/06/23 Page 14 of 24
`
`We further examine how the products’ commission rates on Amazon affect price change on
`
`eBay. We use a similar model but replace the variable 𝐶𝑜𝑚_𝐷𝑖𝑓𝑓- with 𝐶𝑜𝑚-, which measures the
`
`product’s commission rate on Amazon.
`
`As shown in Column (2) of Table 2, a 1-percent higher commission rate on Amazon leads to a
`
`0.540-percent reduction in product prices on eBay due to Amazon’s removal of PPCs. This
`
`supports Hypothesis 1b.
`
`
`
`5.2 Analysis of Prices on Amazon
`
`We then evaluate price changes for sellers on Amazon. In Column (3) of Table 2, we find that a
`
`1-percent increase in commission rate difference leads to a 0.630-percent price reduction on
`
`Amazon.17 Column (4) of Table 2 shows that a product with a 1-percent higher commission rate
`
`on Amazon has a 0.450-percent price drop on Amazon after the removal of PPCs. These results
`
`are consistent with Hypothesis 2a.
`
`Because these products can be sold by Amazon (as a retailer) and/or by third-party sellers, we
`
`also analyze the impact of commission rate differences on the price changes separately. The results,
`
`as shown in Columns (1) and (3) of Table 3, indicate that a 1-percent higher commission rate
`
`difference between Amazon and eBay leads to 0.920-percent and 0.426-percent reductions in the
`
`average price of products sold by Amazon and by third-party sellers, respectively. Furthermore,
`
`Columns (2) and (4) of Table 3 present the empirical results based on Amazon’s commission rate:
`
`a 1-percent higher commission rate on Amazon results in 0.694-percent and 0.305-percent
`
`decreases in the average price of products sold by Amazon and by third-party sellers, respectively.
`
`
`
`
`17 The price here is defined as the lowest price among that of Amazon itself as a retailer and those of all third-party
`sellers offering the product.
`
`
`
`13
`
`
`
`Case 2:20-cv-00424-RAJ Document 100-6 Filed 04/06/23 Page 15 of 24
`
`Table 2. Impact of Amazon’s Removal of PPCs on Price on eBay and Amazon
`
`eBay
`
`Amazon
`
`(1)
`Commission
`rate difference
`-0.721***
`(0.121)
`
`(2)
`Amazon
`commission rate
`
`
`(3)
`Commission
`rate difference
`-0.630***
`(0.161)
`
`(4)
`Amazon
`commission rate
`
`
`
`
`-0.540***
`(0.087)
`
`
`
`Yes
`Yes
`90,633
`0.126
`
`Yes
`Yes
`90,633
`0.134
`
`Yes
`Yes
`3,032,247
`0.025
`
`-0.450***
`(0.114)
`
`Yes
`Yes
`3,032,247
`0.025
`
`
`Variable
`Commission rate
`
`difference × Post
`× Post
`
`
`Amazon commission rate
`
`
`Product fixed effect
`Month fixed effect
`Number of observations
`R-squared
`
`Notes: Dependent variable: Logarithm of (price +1) ×100.
`
`Standard errors in parentheses, clustered at category level.
`*** Significant at the 1% level. ** Significant at the 5% level. * Significant at the 10% level
`
`
`To further examine the significance of the difference in price reduction between Amazon itself
`
`and by a third-party seller. We adopt the following model:
`
`and its third-party sellers, we restrict to the products that are offered both by Amazon as a retailer
`
` =𝛽C+𝛼𝐶𝑜𝑚_𝐷𝑖𝑓𝑓-×𝑃𝑜𝑠𝑡A×𝐹𝑖𝑟𝑠𝑡𝑃𝑎𝑟𝑡𝑦-O +𝛾E𝐶𝑜𝑚_𝐷𝑖𝑓𝑓-×𝑃𝑜𝑠𝑡A+
`𝑙𝑜𝑔 𝑝𝑟𝑖𝑐𝑒-OA
` 𝛾T𝑃𝑜𝑠𝑡A×𝐹𝑖𝑟𝑠𝑡𝑃𝑎𝑟𝑡𝑦-O+𝛾U𝐶𝑜𝑚_𝐷𝑖𝑓𝑓-×𝐹𝑖𝑟𝑠𝑡𝑃𝑎𝑟𝑡𝑦-O+𝛿𝐹𝑖𝑟𝑠𝑡𝑃𝑎𝑟𝑡𝑦-O+𝜆A+𝜃-+𝜀-OA,(2)
`where 𝑗 identifies the seller, 𝐹𝑖𝑟𝑠𝑡𝑃𝑎𝑟𝑡𝑦-O is a dummy variable equal to 1 if the seller of product
`𝑖 in month 𝑡 is Amazon itself and 0 if it is a third-party seller, and 𝛼 is the coefficient of interest.
`𝐶𝑜𝑚_𝐷𝑖𝑓𝑓- can also be replaced by 𝐶𝑜𝑚- for similar analysis.
`
`Columns (5) and (6) of Table 3 reports the result. The significant coefficients and the
`
`differences in the magnitudes further support Hypothesis 2b.
`
`
`
`
`
`
`
`14
`
`
`
`Case 2:20-cv-00424-RAJ Document 100-6 Filed 04/06/23 Page 16 of 24
`
`Table 3. Impact of Amazon’s Removal of PPCs on Amazon’s Price
`Amazon as retailer
`Third-party sellers
`
`Amazon as retailer and
`third-party sellers
`difference
`
`(5)
`Commission
`rate
`difference
`-0.508**
`(0.231)
`
`(6)
`Amazon
`commission
`rate
`
`
`
`
`-0.391**
`(0.156)
`
`
`
`Yes
`Yes
`835,191
`0.002
`
`-0.350*
`(0.182)
`
`
`
`-0.299***
`(0.096)
`
`Yes
`Yes
`835,191
`0.003
`
`(4)
`Amazon
`commission
`rate
`
`
`
`
`
`
`
`-0.305***
`(0.095)
`
`Yes
`Yes
`2,851,177
`0.015
`
`
`Variable
`Commission rate
`
`party
`
`Amazon commission rate
`
`
`Commission rate
`
`
`Amazon commission
`
`
`Product fixed effect
`Month fixed effect
`Number of observations
`R-squared
`
`(1)
`Commission
`rate
`difference
`
`
`(2)
`Amazon
`commission
`rate
`
`
`(3)
`Commission
`rate
`difference
`
`
`
`
`-0.920***
`(0.260)
`
`
`
`Yes
`Yes
`558,192
`0.005
`
`
`
`
`
`-0.694***
`(0.165)
`
`Yes
`Yes
`558,192
`0.006
`
`
`
`-0.426***
`(0.135)
`
`
`
`Yes
`Yes
`2,851,177
`0.015
`
`difference × Post × First
`× Post × First party
`difference × Post
`rate × Post
`Notes: Dependent variable: Logarithm of (price +1) ×100.
`
`Standard errors in parentheses, clustered at category level.
`*** Significant at the 1% level. ** Significant at the 5% level. * Significant at the 10% level.
`
`
`5.3 Robustness Checks
`
`We first test the parallel trends assumption underlying the DID analysis. We conduct a leads-and-
`
`lags analysis that allows us to incorporate the continuous nature of the treatment before the policy
`
`
`
`
`
`change. Specifically, we adopt the model as follows:
`
`𝑙𝑜𝑔 𝑝𝑟𝑖𝑐𝑒-A
`
` =𝛽C+
`
`𝛽[𝐶𝑜𝑚_𝐷𝑖𝑓𝑓-×𝑃𝑜𝑠𝑡A\[
`
`+𝜆A+𝜃-+𝜀-A. (3)
`
`U[]^U
`
`As shown in Table 4, we do not find significant differences in trends before Amazon’s removal
`
`of PPCs. The coefficients of the interaction terms are in general significant and the magnitudes for
`
`Amazon are larger than those for eBay, as expected.
`
`
`
`15
`
`
`
`Case 2:20-cv-00424-RAJ Document 100-6 Filed 04/06/23 Page 17 of 24
`
`
`Table 4. Leads-and-Lags Analysis
`
`
`Commission rate difference × Post (t-2)
`Commission rate difference × Post (t-1)
`Commission rate difference × Post (t)
`Commission rate difference × Post (t+1)
`Commission rate difference × Post (t+2)
`Commission rate difference × Post (t+3)
`Notes: Dependent variable: Logarithm of (price +1) ×100.
`
`
`
`
`
`
`
`
`
`
`
`
`Product fixed effect
`Month fixed effect
`Number of observations
`R-squared
`
`(1)
`eBay
`0.031
`(0.130)
`0.019
`(0.115)
`-0.247*
`(0.150)
`-0.350***
`(0.094)
`-0.214**
`(0.100)
`-0.233*
`(0.125)
`Yes
`Yes
`90,633
`0.136
`
`(2)
`Amazon
`0.006
`(0.068)
`0.051
`(0.065)
`-0.224*
`(0.125)
`-0.293***
`(0.100)
`-0.201*
`(0.105)
`-0.200**
`(0.093)
`Yes
`Y