Project topics and materials
Get Complete Project Material File(s) Now! »
Pricing strategies in online market places and Price Parity Agreements: evidence from the hotel industry
Abstract
This paper analyzes how the use of Price Parity Agreements (PPA) has im-pacted retail pricing strategies of hotels in online market places. We use a unique dataset of prices set by a panel of 863 hotels in Paris that were displayed on more than 25 Online Travel Agencies (OTA) over 3 years. Using a before-after design and controlling for external shocks, we demonstrate that the end of Price Parity Agreements imposed by public authorities to OTAs causes a de-crease of about 3.1% to 4.5% in the average level of prices set by hotels. This decrease may be explained by an increase of price discrimination. We show that the level of price discrimination across OTAs increases by 2.3% to 1.4% after the drop of Price Parity Agreements and that the the degree of inter-temporal price discrimination also increases by 3.6% to 2.1%. This paper discusses how theses results are related to theoretical foundations.
Keywords: Most Favoured Nation Clause, Price Parity Agreement, Online Book-ing Platforms, Before-after design, Price Discrimination.
Introduction
Price Parity Agreements, also called Most Favoured Nation clauses, have re-cently drawn the attention from competition agencies throughout the world. In April 2012, the US department of Justice filed an antitrust lawsuit against HaperCollins, Hachette, Simon & Schuster, Macmillan, Penguin and Apple be-cause they conspired to the end of e-book retailers’ freedom competition on retail prices. Publishers were blamed for taking the control on pricing deci-sions, substantially increasing the prices that consumers paid for e-books. In March 2016, the Supreme Court rejected Apple’s final appeal. Apple was thus sentenced to pay $450 millions to victims. At the same time, the European Com-mission opened a parallel investigation in December 2011. Publishers and Ap-ple agreed to stop all existing so called agency contracts that provided restric-tions on retail prices. In December 2012, the Commission concluded that the commitments were able to restore and preserve competition in the retail prices of digital books.
In spring 2015, the French, Swedish and Italian competition authorities ac-cepted the commitments offered by Booking.com (April 2015[3]) and Expedia Inc. (June 2015), which for a period of five years would remove any Price Par-ity Agreements restricting price differentials between Online Travel Agencies (OTAs). On August the 6th of 2015, the French Parliament passed a law that banned all agreements placing restrictions on hotel pricing.
From April 2014 (16 months before the end of PPAs) to July 2017 (24 months after the end of PPA), we collected the daily listed prices of a panel of 863 ho-tels on Kayak.com, Booking.com and Google Hotels. In addition, we collected individual information of each hotel (quality, number of rooms, services…) on TripAdvisor.com. As the removal of PPA affected simultaneously all hotels in France, and several other countries in Europe, we analyze the effects of this re-moval using a before and after design. The major limitation of this method is the risk of omitted variable bias (see Pearl (2009)[28]). To avoid omitting ex-planatory factors we use several control variables. We first control for external shocks in demand using public data on the number of effective nights booked in Paris. We also control for the entry of Airbnb as a direct competitor of the hotel industry using the number of search request for ”Airbnb Paris” on Google.
This article provides an empirical analysis of the effects of the removal of PPAs on 3 aspects of hotels’ pricing strategy: (i) the average level of price, (ii) the price discrimination across platforms and (iii) the inter-temporal price dis-crimination. We show that the end of PPAs imposed by public authorities to online travel agencies (OTAs) in France leads to a decrease of about 3.1% to 4.5% in the average level of hotels’ prices, an increase of about 2.3% to 1.43% of the dispersion of prices across platforms and an increase of about 3.6% to 2.1% in the degree of inter-temporal price discrimination. Thanks to our unique dataset, this paper provides the first empirical study of the impact of the drop of PPAs on the average level of hotels’ prices in Paris and contributes to the grow-ing litterature analysing the effects of PPAs on prices2. Moreover, to the best knowledge of the author, this article is the first to empirically study the impact of PPAs on price discrimination strategies.
The paper proceeds as follows. Section 2.2 reviews the existing theoretical and empirical literature on PPAs and input price discrimination. A definition of Price Parity Agreement is given in section 2.3. Section 2.4 describes the eval-uation methods. Section 2.5 presents the dataset we use for our analysis. The effects of PPAs on hotels’ pricing strategy are described in section 2.6 and 2.7. Robustness and extensions are discussed on section 2.8 and we keep the section 2.9 for the conclusion.
Related literature
This paper is directly related to the literature that studies the competitive ef-fects of PPAs on online platforms. Many recent papers have provided theoret-ical frameworks that help understanding the effect of PPAs. Boik and Corts (2016) [12] demonstrate that PPAs can lead to higher fees, prices and retailers profit. It can also deter the entry of low end firms. Johnson (2017)[20] extends the model of Boik and Corts (2016) and show that imposing price agreement reduces the competition between platform by cutting their incentive to reduce fees. On the other hand, Johansen and Verge´ (2017)[19] proved that when one allows each supplier to sell either through a platform or directly to consumers, whether Price Parity Agreements lead to higher or lower commissions depends on the degree of competition between the suppliers. In particular, they find that PPAs may simultaneously lead to higher profits for platforms and suppliers, and increase consumer surplus. Finally, Larrieu (2019)[22] demonstrates that the balance between the bargaining power of hotels and platforms appears as a key parameter in assessing the competitive effects of PPAs. Most of the time PPAs are detrimental to consumers, however they may also be welfare improv-ing when hotels own the bargaining power and competition between them is high.
While the theoretical literature is dense, the empirical literature on PPAs applied to online platform is still nascent. Mantovani et al. (2019)[23] analyze the dynamics of hotel prices listed on Booking.com in the period 2014-16 in the main Mediterranean islands of Italy, France and Spain. They show that prices decreased in 2015, the year in which the major antitrust decisions took place, whereas they bounced back in 2016. Hunold, Kesler, Laitenberger and Schlutter (2017)[17] focus on the direct selling channel of hotels and show that PPA influences the pricing and availability of hotel rooms across online sales channels. In particular, hotels tend to promote the direct online channel more actively at a lower price more often after the ban of PPAs.
This paper is also linked to the literature on input price discrimination. In-put price discrimination involves charging different prices to different resellers for the same good. Vulkan (2003)[31] exhibits that e-commerce is specifically prone to price discrimination because of all the data (browsing history, geoloca-tion,…) that allow sellers to analyze price sensitive groups of consumers. Mat-tioli (2012)[24] shows for instance that the online travel agency, Orbitz, targeted Mac users with costlier prices in comparison to Windows users. By definition, PPAs ban price discrimination between resellers as the resale price decided by hotels should be the same on all platforms. Arya and Mittendorf (2010)[6] ana-lyze the effects of a ban on input price discrimination across resellers in a setting where resellers are asymmetric and one operates in multiple markets. They find that price discrimination leads to price cuts in markets with lower demand and that, when these low demand markets are also less competitive, price discrim-ination can provide welfare gains by increasing the output on these markets. Miklos-Thal and Shaffer (2018)[25] focus on price discrimination across mar-kets rather than across buyers, and expose that discrimination may have a pos-itive allocation effect: welfare can rise even if total output decreases. Finally, Allain, Chambolle and Turolla (2019)[4] study a reform authorizing wholesale price discrimination that took place in France in 2008. They show that, on aver-age, suppressing the ban on input price discrimination indeed lowered national brand food prices by 3.36% on average compared to the private labels.
Price Parity Agreements
Price Parity Agreements, also reffered as Most Favoured Nation clauses, may refer to different situations depending on the agent who imposes the clauses. The traditional form takes place in a wholesale contract. For example, when a manufacturer asks its retailers not to better promote the brand of a competitor or when a retailer asks its suppliers not to sell its good at a lower price on another sales channels.
On online booking platforms, PPAs refer to agreements between the plat-form and the hotel that regulates the price and/or supply conditions for the third party, the final consumers, who are not part of the agreements. PPAs are a specific type of agreement that are imposed by platforms to hotels. They are not vertical agreements in the strict sense, because platforms acts as intermediaries, but they have an inherent vertical element. Hotels reservation platforms such as Booking.com, Expedia and HRS incorporated PPAs into their general terms and conditions. Under these conditions, if an hotel wants to be displayed on Booking.com, it has to offer its rooms on this platform at the best prices avail-able on any channel, at the best conditions. If we consider a hypothetical hotel that already offers 10 Queen bedrooms at a rate of 100 euros per rooms, Book-ing.com or any other platform with PPAs will require to have at least these same conditions of prices and capacities for the same type of rooms.
A distinction has been made by competition authorities between narrow and broad agreements. A narrow agreement between a hotel and an OTA commits the hotel not to charge a lower price on its own distribution channels that the one charged by the OTA to consumers. Under these agreements, the hotel is not able to offer better prices or conditions to someone who makes a reservation offline. The price and conditions between OTA and the direct channels of the hotel have to be the same or better on the OTA. In contrast, a broad (or wide) agreement between a hotel and a OTA commits the hotel not to charge a lower price on competing OTAs. Under these agreements, the hotel has to put the exact same prices and conditions on every booking platform.
Empirical setting and evaluation methods
In December 2013, in Germany, the Bundeskartellamt prohibited HRS from con-tinuing to apply its prices agreements while at the same time it initiated pro-ceedings against Booking.com and Expedia for applying similar clauses. In both cases, price agreements were suspected of anti-competitive effects. First because they could induce a reduction of competition between the hotel book-ing website and platforms but also because parity clauses may lead to the fore-closure of smaller platforms or new entrants.
In France, within the scope of a procedure initiated before the Autorite´ de la concurrence by the main French hotel unions and the Accor group, Book-ing.com commits to change its commercial practices. Booking.com amended the Price Parity Agreements and removed any clause imposing parity obligations in terms of the availability of rooms or commercial conditions. These commit-ments were undertaken for 5 years and came into force on the first of July 2015. France, Sweden and Italy worked together, in close coordination with the Euro-pean Commission, to obtain similar commitments from Booking.com in these three countries. In addition, in August 2015 the Article 133 of the Loi Macron renders null and void all OTA Price Parity Agreements. From this date, plat-forms such as Booking.com, Expedia, HRS or Oui.Sncf no longer have the right to impose Price Parity Agreements to hotels.
Our objective is to evaluate the causal effect of the ban of PPAs. Program evaluation methods are widely applied in economics to assess the effects of pol-icy interventions and other treatments of interest. Examples include minimum wages and employment policies (Card & Krueger, 1994[14]), health care inter-ventions (Newhouse, 1996[27]), and more recently large-scale online A/B stud-ies in which IP addresses visiting a particular web page are randomly assigned to different designs or contents (see Bakshy et al., 2014[9]). Abadie and Cattaneo (2018)[2] describe the main methodological frameworks of the econometrics of program evaluation and show that randomized experiments and difference in difference are from the most robust tools to perform program evaluation. Un-fortunately, as hotels from our panel are all simultaneously affected by the drop of PPAs, we don’t have control groups and we cannot use one of these methods.
Thereby, we run a before-after analysis, comparing the change in pricing strategies of hotels before and after the ban of PPAs in France. This evaluation method is simple and powerful but its major limitation is the risk of Omitted Variable Bias (see Pearl (2009)[28]). The tourism industry is very sensitive to external events such as major sport events, political crisis, terrorism etc. During the period of our study (April 2014 to July 2017) many events occurred and may have important effects on hotel pricing strategies. In 2015, Paris suffered from major terror attacks. First in January the 7th (Charlie Hebdo terror attacks) and the second one in November the 13th (6 simultaneous attacks in Paris). On the other hand, in 2016, France was the host of the soccer UEFA European Championship between June the 10th, and July the 10th.
To avoid omitting explanatory factors we use several control variables in ad-dition to market fixed-effects (seasonality, hotel and platform characteristics…). We first control for external shocks in demand using public data on the num-ber of effective nights booked in Paris. We expect the average prices of hotels to be positively correlated with volume of booked nights. We also control for the entry of Airbnb as a direct competitor of the hotel industry. Farronato and Fradkin (2018)[16] demonstrate that Airbnb revolutionized the lodging market by making additional rooms available and especially during peak periods when hotel rooms often sell out and rates skyrocket. This research shows that in the 10 cities with the largest Airbnb market share, including Paris, the entry of Airbnb resulted in 1.3% fewer hotel nights booked and a 1.5% percent loss in hotel rev-enue. We control for the growth of Airbnb using the number of search requests for ”Airbnb Paris” on Google. Here we expect the average price of hotels to be negatively correlated with the increasing notoriety of Airbnb as a direct com-petitor of hotels in Paris.
Data
Booking information such as the price of the room, the date of Booking, the type of room or the profile of the customer are highly sensitive data and are considered as industrial secrets by the big players of this industry. We didn’t get access to these data so we decided to construct our own data-set.
The backbone of our data are prices of hotel rooms in Paris collected on a daily basis on the meta-search engine Kayak.com between April 2014 to July 2017 for more than 863 hotels on more than 25 platforms. We also collected data on hotels characteristics on TripAdvisor for the same time period. We moni-tor for specific shocks in hotels’ demand using data on the number of effective nights booked in Paris. These data are provided by the French National In-stitute of Statistics and Economic Studies (INSEE). Finally, we control for the entry of Airbnb as a direct competitor of the hotel industry using the number of search request for ”Airbnb Paris” on Google. The following sections describe the collected data.
Kayak data
Kayak is a fare aggregator and travel metasearch engine. From its many ser-vices, it provides users a way to compare the price of hotels sold on different platforms. Kayak Software Corporation was acquired by Booking Holdings on May 21, 2013.
We collected data on Kayak.com for 863 hotels in Paris between April, 10th of 2014 to July, 1st of 2017. Data were collected every day from 6am to 8am using a web-scrapper robot. We collected 6,097,632 prices offered by hotels in Paris on more than 25 platforms for a twin bedroom for 2 people for 1 night available in 4, 14 and 30 days. For 2 months (June 2014 and September 2015) we also collected data directly from Booking.com, Expedia and Voyage-SNCF to cross-check the prices of room observed on Kayak.com. Results show that less than 4% of prices were significantly different between Kayak.com and Booking.com, Expedia or Voyage-Sncf. Data collected on Kayak are therefore representative of the available online prices for hotel rooms.
Among these 863 hotels, we restricted our panel to the hotels that were dis-played on Kayak at least 60% of the time between April 2014 and July 2017, and were present at least once during the 5 first days and the 5 last days of our sample. These restrictions aim at creating a panel that exclude hotels that are closed, changed their name or that decided to no longer be present on Internet during this period. In addition, we decide to exclude from our analyzes hotels of very low quality (zero or 1 star) as it exists very few hotels bellow 1 star in Paris. Finally, these restrictions lead to create a panel of 535 hotels to whom we collected their prices on a daily basis.
Table of contents :
1 Most Favoured Nation Clauses on the Online Booking Market
1.1 Introduction
1.2 The online booking industry
1.3 The model
The Setup
Bargaining assumptions and solution concept
1.4 Benchmark : No Restriction
Stage-2
Stage-1
1.5 Most Favoured Nation Clauses
Stage 2
Stage-1
Equilibrium outcomes
1.6 Robustness and extensions
Observability of the output of the negotiations
Endogenous MFN adoption
1.7 Conclusion
1.8 Appendix
Proof of Lemma 1
Proof of Lemma 2
Proof of Proposition 4
Proof of Corollary 1
Proof of Proposition 6
Proof of Lemma 5 and Proposition 7
2 Pricing strategies in online market places and Price Parity Agreements: evidence from the hotel industry
2.1 Introduction
2.2 Related literature
2.3 Price Parity Agreements
2.4 Empirical setting and evaluation methods
2.5 Data
2.5.1 Kayak data
2.5.2 TripAdvisor data
2.5.3 INSEE data
2.5.4 Airbnb
2.6 Effects of PPAs on the average level of price
2.6.1 Specifications
2.6.2 Results
2.7 Effects of PPAs on hotels’ price discrimination strategy
2.7.1 Across Platforms Price Discrimination
2.7.2 Inter-Temporal Price Discrimination
2.8 Robustness and extensions
2.8.1 Impact of Geographical Competition
2.9 Conclusion
2.10 Appendix
2.10.1 Price differences between neighbourhoods
2.10.2 Inter-Temporal Price Discrimination depending on hotels haracteristics
3 Evaluation des amendes dans les cas de cartel en France
3.1 Introduction
3.2 M´ethodologie
3.2.1 Le montant des amendes en pratique
3.2.2 Niveau optimal de l’amende
3.2.3 D´etermination de l’amende de r´ef´erence
3.3 Les donn´ees
3.3.1 Construction de la base de donn´ees
3.3.2 Statistiques descriptives
3.3.3 Analyse du montant initial et final des amendes
3.4 Resultats
3.4.1 Hypoth`eses sur les valeurs de param`etres
3.4.2 Propri´et´es dissuasives des sanctions
3.4.3 Propri´et´es compensatoires des sanctions
3.5 Conclusion
3.6 References
3.7 Annexes
3.7.1 Liste des cartels contenus dans notre base de donn´ees
3.7.2 Part des amendes au dessus de DF entre 2012 et 2018
3.7.3 Part des amendes au dessus de CF entre 2006 et 2018
