U
NIVERSITEIT VANA
MSTERDAM Master ThesisMARKET STRUCTURE AND PRICING COMPETITION AMONG HOTELS
August 2014
Name: Rob van Hemert
Student number: 10317244
Supervisor: Prof. dr. J. Strikwerda
Reviewed on September 1, 2014. Approved, there are some minor points and typos indicated by me in this file + one question at the conclusions
2
Preface
With pleasure and great pride I present my master thesis. It represents the end of a joyful ride and a decade of education.
First of all, I would like to thank Marco and Sam for giving me the opportunity to obtain my master’s degree and for their support throughout.
Second, I would like to thank my parents for their unconditional love and support, in good times and in bad times. I hope my accomplishment makes you proud.
I consider myself to be very lucky with Prof. dr. J. Strikwerda as a supervisor and have thoroughly enjoyed the inspirational meetings. You have been such a help, thank you. But most of all I am extremely happy to finally write this part of my thesis for my beloved Stefanie. Thank you so much for putting up with me through what have been challenging times. I could not have done this without your love and support, thank you.
3 Contents Preface ... 2 1 Abstract ... 5 2 Introduction ... 5 2.1 Overview ... 5 2.2 Theoretical Contribution ... 8 3 Literature Review ... 9 3.1 Industrial Organization ... 9 3.2 Market structure ... 9
3.3 Market structure in the travel agent industry ... 11
3.4 Market structure in the hotel industry ... 13
3.5 Performance ... 14
3.6 Game theory ... 15
3.7 Pricing competition and pricing strategies in hotel industry ... 16
4 Propositions and hypotheses ... 17
5 Method and data ... 18
5.1 Research philosophy and approach ... 18
5.2 Data Collection and cleansing ... 19
6 Results ... 27
6.1 Herfindahl-Hirschman-Index for travel agents ... 28
6.2 Four firm concentration ratio for travel agents ... 28
6.3 Herfindahl-Hirschman-Index for Dutch hotel market ... 30
6.4 Four firm concentration ratio for Dutch hotel market ... 30
6.5 Technical analysis pricing based competition ... 32
7 Conclusion and discussion ... 40
8 Limitations ... 41
4
10 Appendices ... 45
Table 1: Means, Standard deviation, Skew Turnover, Profit and Market Share ... 19
Table 2: Travel agent industry total turnover and profit per year ... 20
Table 3: Descriptive Statistics for the Travel agent industry (2004-2013) ... 20
Table 4: Descriptive Statistics for the Dutch Hotel Market (2004-2013) ... 21
Table 5: Sample hotels price and other characteristics ... 22
Table 6: Descriptive statistics hotel rates per day type ... 24
Table 7: Descriptive statistics per observation date for the complete check-in sample ... 27
Table 8: Correlation, means and standard deviations ... 32
Table 9 Results for Daysdiff and Avail as predictors for Rate ... 33
Table 10: Regression output. Dependent variable is natural logarithm of the rate. ... 37
Table 11: Price reaction to the clusters of hotels. See Figure 5 Panel A-C... 39
Table 12: Availability reaction to the clusters of hotels. See Figure 5 Panel A-C. ... 45
Table 13: Weekend reaction to the clusters of hotels. See Figure 5 Panel A-C. ... 45
Table 14: Time reaction to the clusters of hotels. See Figure 5 Panel A-C. ... 46
Figure 1: Overview value chain (adapted from Granados et al. (2008) ... 13
Figure 2: Inter-temporal development of average rate observed over the whole sample of check-in dates for 3 and 4 star hotels ... 24
Figure 3: Inter-temporal development of average rate observed for week day check-in dates for 3 and 4 star hotels. ... 25
Figure 4: Inter-temporal development of average rate observed for weekend check-in dates for 3 and 4 star hotels. ... 25
Figure 5: Average trends in the percentage rate change for the total arrival sample (panel A), the weekday arrival sample (panel B) and the weekend arrival sample (Panel C). ... 35
Graph 1: Herfindahl Hirschman Index Travel Agent Industry '04-'13... 28
Graph 2: Concentration Ratios Travel Agents '04-'13 ... 29
Graph 3: Herfindahl Hirschman Index for the Dutch Hotels '04-'13 ... 30
5
1 Abstract
This thesis provides insights in the market structure of the international travel agent industry and the Dutch hotel market by calculating the Herfindahl-Hirschman Index and concentration ratios for the period of 2004-2013. I find that the level of concentration in both industries is low and make suggestions to facilitate analyzing concentration levels and thus levels of competition more accurately in the future. In the next step of the thesis an analysis of inter-firm price based competition is performed using prices published on Booking.com for 11 hotels in Amsterdam that belong to a competitive set, which were collected for 13 consecutive weeks. I show to what extend pricing decisions are influenced by availability of hotels for a certain booking date, star rating, size, time left before check-in and pricing decisions made by competing hotels.
2 Introduction
2.1 Overview
When Google Chairman Eric E. Schmidt in 2011 said that “The Internet is the ultimate level playing field” he could not have been any further from the truth. Internet economics are of the kinds that are best described by power law mathematics, meaning that a small
percentage of the players claim a very large share of the market and its returns. The Internet has paved the way for many new businesses and business models and in 2008 12% of EU’s turnover came from e-commerce. Many online businesses act as intermediaries. In short, intermediaries reduce distribution and selling costs for the seller, while simplifying the choice process for the buyer (Palmer & McCole, 1999). In the early 90s many studies spoke of a phenomenon called disintermediation. The Internet would allow companies to communicate with and sell their products to consumers directly, rendering intermediaries largely obsolete. It was expected that cost of sales and thus the costs of products and services paid by
6 “As an information-based item, travel bookings are intangible, heterogeneous, and fixed
geographically, making the sale of travel highly dependent on getting the right information to potential customers at the appropriate stage in their purchase decision” (O’Connor &
Piccoli, 2003, p. 108). The disintermediation hypothesis turned out to be partly false. While traditional intermediaries, such as the world’s oldest travel intermediary Thomas Cook who announced in 2013 that it would cut 2.500 jobs, after already cutting another 1.100 jobs in 2012, struggle to keep up, their online rivals such as Booking.com and Expedia are
flourishing. Internet intermediaries have become increasingly important, influential and successful in electronic marketplaces over the past decades. In a time where there is more information available than there has been ever before, search engines such as Google and intermediaries such as Amazon, eBay and Priceline.com have found profitable ways to help consumers to find what they need. As a matter of fact, these intermediaries have become so powerful that they can hypothetically make or break the market participants that depend on their services. The intermediaries thus possess profit power, a concept described by De Kuijper in 2009, which refers to a firm’s ability to hold on to the value of its own activities as well as to extract value from the activities of others with whom they interact in their
commercial dealings, to increase the value available to the entire group, and to optimize the risks for themselves and allocate to others the risk that they do not want (De Kuijper, 2009, p. 7). The main premise here is that the Internet has created a transparent environment, in which firms that control or possess so called power nodes will be able to claim to largest share of profit, competitive success and power in relationships with competitors and partners. In short, what De Kuijper puts forward is that profit power is not equal to market power. Market power enables firms to raise the market price of a good or service over the marginal cost. Perfect competition states that individual firms should have no control over prices, when the products (or services) offered are homogeneous. De Kuijper argues that profit power enables the one in
7 possession of it to extract profits from its environment, whilst providing benefits at the same time.
The relationship between online travel intermediaries and hotels can best be described as a love / hate relationship. The general consensus is that the intermediaries provide many
benefits to hotels, mostly in terms of distribution power and the so called billboard effect. This means that intermediaries enable hotels to address an audience that they would not be able to address on their own. However, the intermediaries’ services come at a cost. Most intermediaries charge commissions over the bookings they provide for the hotels. These commissions range from 10 to 25 percent and have risen over the years. Hoteliers feel powerless against the industry giants and have become very reliant on their services. The intermediaries generated approximately 32.2% of the revenues in the Dutch hotel market in 2013 according to KPMG’s hospitality benchmark (Sweers & de Graaf, 2014). In 2012 this was 28.9% and so it appears that hotels are becoming more and more reliant on the
intermediaries.
Recently there have been some reports of uncompetitive conduct in the hotel industry. Some of the most notable incidents include the exclusions of the Choice Hotels International group by Expedia Inc. after Choice Hotels did not agree to new contract conditions imposed by Expedia (Lee et al., 2013) and the lawsuits against Booking.com for alleged price fixing and consumer deception. Although these issues themselves provide interesting grounds for research in many areas for scholars of all kinds, this study’s aim is to provide insight in the actual price competition between hotels belonging to a competitive set in Amsterdam. The main research question of this thesis is;
How do hotels belonging to a competitive set in Amsterdam compete on price given a certain market structure and what are the effects of competitors’ pricing decisions?
8 In order to answer this question is shall answer the following sub questions;
What are the levels of concentration in the international travel agent and Dutch hotel industry and what kind of market structure applies in the separate industries?
What are the barriers to entry in the respective industries? 2.2 Theoretical Contribution
The purpose of this thesis is to provide insights in the level of concentration in the two industries that together determine hotel prices; the Dutch hotel industry and the travel agent industry. Both industries have very different characteristics with the hotel industry being a low velocity market and the travel agent industry being a high velocity market. Given the recent signals of uncompetitive conduct of powerful travel agents towards hotels, the central question in this thesis is how this affects hotels’ pricing behavior online. The level of focus on individual firm behavior in this study presents unique insights and has not been presented in the hotel industry before as far as the author is aware.
9
3 Literature Review
This chapter presents the theoretical framework for this master thesis. 3.1 Industrial Organization
In this thesis I shall take the perspective on industrial organization (IO) as presented by Bain in his seminal 1956 article. Bain (1956) introduced the structure, conduct, performance hypothesis (SCP). Structure is determined by industry concentration and barriers to entry that affect the level of competition. Conduct refers to the way industry incumbents behave. Examples are controlling output, colluding, predatory pricing, advertising and research and development. The final construct of the SCP paradigm is performance. Within Bain-type IO performance is measured in terms of profitability, efficiency and innovation. Bain-type IO assumes a causal relationship between an industry’s structure, conduct and performance.
According to Barthwal (2007, p. 71) there are standard forms of market structure. These forms are perfect competition, monopoly, monopolistic competition, duopoly, bilateral monopoly, oligopoly and contestable markets. In order to define market structure, Bain suggests to define industries by four main characteristics; the degree of sellers concentration, the degree of buyers concentration, the degree of product differentiation and the condition of entry to the market.
3.2 Market structure
There are multiple measurements for market structure and competition. Below I shall provide an overview of the ones most commonly used by academics and discuss their suitability for this research. The Herfindahl-Hirschman Index (HHI) is a measure that compares firm size to industry size and provides an indication of the competition in an industry. It is calculated by squaring the market shares of all firms in a market and then summing the squares (Rhoades, 1993). This measure is used by the U.S. antitrust division. The maximum score for the HHI is
criticism in the literature points out that the direction of causality is not simply S—> C—> P, but that many firms manipulate the structure, so there is a conduct dimension as well on the structure level.
10 10,000, which is obtained if there is a single firm in the industry, which thus has a market share of 100 (100 x 100 = 10,000).
Alternatively, the Lerner Index, or price cost margin (PCM) expresses the difference between price and marginal costs and it depends on elasticity of demand. However, PCM tends to misrepresent the development of competition over time in markets with few firms and high concentration (Boone, 2007). It is for this reason that this way of measuring the level of competition might not be suitable for this research, since I expect there to be a high level of concentration in the travel agent industry.
A third measure of competition is Profit elasticity (PE) which was first introduced by Boone (2007) and it represents the percentage fall in profits due to a 1% increase in (marginal) costs. Boone argues that an increase of this elasticity indicates an increase in competition because firms will be punished harder for losing efficiency in competitive markets.
The final measure I will discuss is one which is often used by academics, the Concentration Index. Although there are different applications of the index in different studies, the four firm concentration index (CR4) and the eight firm concentration index (CR8)
are both used often to determine the level of competition and market power in an industry. The CR4 is calculated by summing the market share of the top four companies active in an
industry. Higher scores indicate less competition in a market, In general, if the CR4 is below
40% a market is considered to be highly competitive and can be labeled as perfect or monopolistic competition. Scores between 40-80% indicate oligopoly and scores of >80% indicate extremely concentrated oligopoly or monopoly. The CR8 is calculated the exact same
11 and market share were used as measures for market structure by Davies and Downward
(1996) and Davies (1999).
3.3 Market structure in the travel agent industry
The travel agent industry comprises thousands of sellers and it appears that there is some level of concentration when looking at some of the online travel agent industry giants, such as Expedia, TripAdvisor and Priceline.com. Clemons et al. (1998) concluded after studying the online travel agent industry that “this market is not characterized by perfect competition” and that “online travel agencies engage in both horizontal product differentiation and price
discrimination” (Clemons et al. 1998, p. 2). However, the online travel agencies that Clemons referred to in 1998 are very different from the ones we see nowadays. None of the
aforementioned travel agents decide what prices they offer to consumers, but they simply provide an easy way for consumers to compare thousands of hotels and airline tickets. The travel agents do encourage hoteliers to give them extra discounts on room prices for multiple night stays and this way they try to make their platform more attractive than other platforms. Ironically, all of the online travel agents have rate parity clauses in their supplier contracts. Rate parity allows travel agents to match the lowest price available anywhere, if the consumer finds a cheaper alternative. For example, if a hotel would offer a room for €100 on its own website, then the travel agent should be allowed to sell it for the same price. Obviously this prevents hotels from attracting consumers by presenting a price edge over the travel agents. The only exceptions to this rule are loyalty and membership programs. When a consumer is a member of a hotel’s community, the hotel is allowed to present lower prices to that consumer. This price may not be available publicly. The hotel can actually give a lower rate to the consumer because it saves on distribution costs in the form of commission on the booked revenue if the room is booked directly at the hotel. Rate parity provides travel agents with a guard against lower prices on other platforms and is, in my opinion, one of the main reasons
T e x t
12 of their success. Rate parity can be labeled as a barrier to entry present in the industry, for it prevents potential competitors to offer something that consumers so desperately want; the lowest price available.
Another barrier to entry in the travel agent industry is advertising. Both online and offline travel agencies invest heavily in advertising. Online travel agencies spend billions of dollars every year on search engine advertising. Priceline.com spent $1.8 billion on online advertising in 2013 alone and is Google’s biggest client in terms of advertising revenues. Priceline.com spent an additional $127 million on offline advertising in 2013. This kind of advertising prevents new entrants from being able to compete on terms of visibility online, since new entrants would have to invest heavily before yielding any revenues from bookings. This very problem was faced by Booking.com in 2005, when they invested so much in advertising they nearly defaulted and had to be saved by Priceline.com. Priceline.com acquired Booking.com for €110M in 2005. So if new entrants are not likely to compete in terms of visibility and are unable to offer better deals due to the rate parity clauses they can only win by outperforming incumbents in terms of usability and customer service.
This brings us to the final barrier to entry I will discuss; Intellectual Property (IP). Below I present an excerpt of the Priceline.com 2013 annual report in which they acknowledge the importance of IP for its success.
“Over time and through acquisitions, we have assembled a portfolio of patents, trademarks, service marks, copyrights, domain names, and trade secrets covering our services. We regard the protection of our intellectual property as critical to our success. We protect our intellectual property rights by relying on national, federal, state and common law rights in the United States and internationally, as well as a variety of administrative
13 Based on these barriers to entry it can be concluded that entry in the travel agent industry is not easy and that there are signs of imperfect competition, or even oligopoly. This means that it is likely that the level of concentration in the industry is high.
Figure 1 presents an overview of the value chain in the hospitality industry as presented by Granados et al. (2008).
Figure 1: Overview value chain (adapted from Granados et al. (2008)
The value chain is divided in offline (left) and online (right). In the first analysis, that of the level of concentration in the travel agent industry, both offline and online agents are included. The final part of this thesis, in which pricing based competition is researched on the Online Travel Agency (OTA) Booking.com focuses on the online section and more
specifically the link between OTAs and hotels. 3.4 Market structure in the hotel industry
Prior studies on Industrial Organization in the hotel industry have reported different results. Matovic (2002) conducted a research about the competitive market structure of the U.S. lodging industry between 1996 and 1999 and concluded that the industry was becoming more competitive on the brand level, which is caused primarily by the introduction of new brands
14 and brand extensions (Matovic, 2002, p. 116). Martel (1974) characterized the U.S. hotel industry as exhibiting tendencies of monopolistic competition, but found no evidence of predatory pricing by the leading firms (Matovic, 2002, p. 23). Yup Chung (2000) studied super deluxe hotels in Seoul and qualified their market as an oligopoly, mainly due to the high barriers to entry. Yup Chung goes on by stating that the super deluxe market cannot be
labeled as monopolistic competition, because of the existence of these high barriers to entry, in the form of high fixed costs. Another interesting remark made by Yup Chung refers to the high level of interdependence in oligopolistic competition in the hotel industry, by stating that “the consequences to a hotel of employing a specific pricing strategy depend not just on that
hotel’s strategy, but also on what pricing strategies the competing hotels employ” (Yup
Chung, 2000, p. 138). In 1999, Davies studied the applicability of the Industrial Organization approach for the UK hotel sector. His main results were that there was evidence of oligopoly, rather than competition or contestability (Davies, 1999, p. 1). Gu-Shin et al. (2010) reapplied the SCP paradigm to the Taiwanese international tourist hotel industry and found that firm profitability was positively related to market share.
In terms of supplier power the issues raised in the previous section indicate that travel agents hold significant power over hotels and largely prevent them from undercutting their prices. Compared to the travel agent industry barriers to entry in terms of capital are actually relatively low in the hotel industry and therefore I expect there to be monopolistic competition in the hotel industry.
3.5 Performance
To operationalize firm performance I will use return on sales (ROS). ROS is calculated by dividing a firm’s profit before taxes by a firm’s turnover. ROS was used as a measure of performance in similar studies, such as Davies (1999). Other measures for performance
15 include profit (Davies & Downward, 1996; Pan, 2005), sales and ratio (Davies & Downward, 1996).
3.6 Game theory
Game theory contributed to a better application of Industrial Organization in the real world. Its basic premises are that any game in the market consists of four elements; i) players, ii) rules, iii) outcomes and iv) pay-offs. Together these elements define the structure of the game. The fundamental assumptions in game theory are that all players act rationally and have common knowledge, referring to the game structure. Depending on the timing of moves made by players and the uncertainty of payoffs of competitors, different types of games can be distinguished. There are static games, in which each player moves only once, without information of its competitors and there are more dynamic games, both with and without complete information about competitors’ moves and payoffs. Hotels and airlines are known to apply revenue management and so in the hotel industry a dynamic game is played with a high frequency of pricing decisions. Hypothetically, players of the game can have full information since competitors’ rates are published online. In practice however, hotels tend to focus on a small subset of their market. For the hotel industry, one major constraint is the capacity constraints. This means that in the short run, hotels are not able to expand their capacity. This constraint is one of the variations in the Bertrand competition. Hansen and Kanafani (1989) described a model of hub-competition for the airline industry as a non-cooperative game, in which competing firms solely try to maximize their own profitability. Although the airline industry and the hotel industry have some clear differences, the same non-cooperative game theory applies. Hotels, very much like airlines, compete for market share and traveler
revenues on the basis of fare, service quality and capacity (Hansen & Kanafani, 1989, p. 29). in the hotel industry
16 3.7 Pricing competition and pricing strategies in hotel industry
According to traditional economic theory firms competing in competitive markets have no control over prices and are price-takers. This implies that firms in a competitive market cannot reduce their prices, because that would create a position that cannot be sustained over time, since the firms would be selling their services under its marginal cost (Gerardi & Shapiro, 2009). Collins and Parsa (2006) identify three common approaches to pricing in the hospitality industry; cost-based pricing, customer-driven pricing and competition-driven pricing. Cost-based pricing represents a financially driven approach, where products are priced to earn a profit above and beyond all costs associated with producing the product. With customer-based pricing, prices are based on customers’ willingness to pay, while competition-based pricing refers to a market-driven approach, in which prices are set to attain a targeted market-share level (Collins & Parsa, 2006, p. 93). Abrate et al. (2012) find that the inter-temporal pricing structure for hotels depends on star rating, type of customer and the number of available rooms. Pricing discrimination is often applied in multisided markets, such as the airline industry, where people who travel last minute generally have a higher willingness to pay than the people who book early (Evans, 2006). The hotel industry shows the same characteristics and because of the differences in price elasticity of demand, hotels are able to charge different prices for homogenous goods, e.g. rooms, by applying pricing strategies, often referred to as yield management. Price discrimination is categorized into three types; first-, second- and third degree price discrimination. Price discrimination can only exist if customers have different elasticities of demand and firms have a way to segment and target customers according to these differences (Phillips, 2005).
in an efficient, competitive market market prices = marginal costs
17
4 Propositions and hypotheses
The literature review leads to the following propositions and hypotheses;
Proposition 1: Based on the recent reports of uncompetitive conduct and the high barriers to entry in the travel agent industry I expect the level of concentration, measured by CR4, CR8
and HHI, in that industry to be high. This would indicate that the travel agent industry is an oligopoly.
Proposition 2: Based on the high levels of supplier power I expect there to be low levels (CR4
<.40) of concentration in the Dutch hotel industry and that its market structure can be classified as monopolistic competition.
For my empirical analysis of the travel agent and Dutch hotel industry I have the following expectations. I expect the SCP paradigm (Bain, 1957) to apply to both the travel agent industry and the Dutch hotel industry. Following this expectation, I expect market share and concentration to be positively related to profitability, measured by ROS and/or profits.
For the analysis of hotel pricing I expect the general principles of revenue
management to hold. This means that I expect a positive effect of demand on room prices. Secondly, I expect star rating to have a positive effect on room prices. Thirdly, I expect a negative effect of the number of days left until a certain check in date on the room prices, meaning that the more time there is left, the lower the room prices will be.
Lastly, I expect pricing decision to be influenced by competitors’ previous pricing decisions and thus I expect to find proof of competition-based pricing.
18
5 Method and data
5.1 Research philosophy and approach
The author will adopt a positivist approach to research, meaning that research will be undertaken in a value-free way (Saunders, 2009). The author will draw on previously established theories in order to form hypotheses. In order to be able to take the positivist approach to research Gill and Johnson (2010) argue that it is likely to use a highly structured methodology in order to facilitate replication. Saunders (2009) adds that the emphasis will be on quantifiable observations that lend themselves to statistical analysis. The research
approach is best described as deduction, meaning that a theory will be developed and subjected to a rigorous test (Saunders, 2009).
The purpose of this research on the one hand is to provide an overview of the market structure of the two aforementioned industries, and on the other hand to describe actual firm behavior given this market structure. The research is therefore best described as descripto-explanatory research. The research shall be conducted by means of both a longitudinal and a cross-sectional study in the hotel and travel agent industry. In the first part of the research we shall measure the level of competition in both the hotel industry itself and in the hotel
industry’s distribution using the measures described in section 3.2. After we have determined the level of concentration in both industries a multiple case study on a small sample of the Dutch hotel market will be performed with the aim to extend on prior research done by Abrate et al. (2012) and to provide insights on how hotels compete on price given a certain market structure.
19 5.2 Data Collection and cleansing
The data needed to analyze the level of concentration in the travel agent industry was collected using the Orbis database which is provided by Bureau van Dijk. NACE Rev.2 industry code 7911, which stands for travel agency activities, was used to identify the companies of interest in the period between 2004 and 2013. An initial search returned 304,187 results, which were then further reduced by filtering for companies with more than 50 employees. 3,384 results were left. Upon further inspection it appeared that not all the firms were still active, or had financial data available for the whole period. 3,247 firms had data for at least one of the years observed. Table 1 presents the descriptive statistics for this sample per year. The total turnover and profit of the sample per year is depicted in
The sample is dominated by traditional offline travel agencies and tour operators. With traditional I refer to a certain type of business model, in which the firm has a contract with wholesale prices that are discussed and negotiated before the year starts. The other type of business model is referred to as agency model. In this case, the intermediary simply gets a percentage of sales through their platform, as is the case for most online travel agents. Table 2.
Table 1: Means, Standard deviation, Skew Turnover, Profit and Market Share Turnover (€1.000) Profit (€1.000) Market Share
N M STD skew M STD skew M STD skew
‘04 827 134.497,86 1.435.207,18 25,05 1.442,63 32.246,77 10,25 0,0012 0,0130 24,87 ‘05 902 159.011,59 1.558.927,46 23,67 2.525,42 28.869,06 19,35 0,0011 0,0110 23,47 ‘06 997 131.660,34 1.548.824,54 28,35 1.914,93 35.464,75 18,34 0,0010 0,0119 28,17 ‘07 1077 149.526,04 1.729.493,52 26,01 2.624,03 44.521,79 21,32 0,0009 0,0108 25,79 ‘08 1209 110.762,69 743.861,09 15,93 12,94 64.430,48 -15,16 0,0008 0,0056 15,75 ‘09 1345 85.136,60 582.756,42 18,67 1.148,43 13.462,14 5,16 0,0008 0,0051 18,48 ‘10 1619 79.639,20 563.120,57 19,74 2.905,92 43.524,08 25,22 0,0007 0,0046 18,72 ‘11 1668 83.614,79 607.084,57 19,96 2.567,39 51.539,09 16,11 0,0006 0,0044 19,61 ‘12 1597 89.617,77 651.812,41 19,59 3.165,35 50.809,14 13,36 0,0006 0,0046 19,36 ‘13 677 133.463,82 965.965,31 13,59 11.060,74 71.396,64 9,74 0,0015 0,0107 13,59
20 The sample is dominated by traditional offline travel agencies and tour operators. With traditional I refer to a certain type of business model, in which the firm has a contract with wholesale prices that are discussed and negotiated before the year starts. The other type of business model is referred to as agency model. In this case, the intermediary simply gets a percentage of sales through their platform, as is the case for most online travel agents.
Table 2: Travel agent industry total turnover and profit per year
Turnover (€1.000) Profit (€1.000) n Total Total 2004 827 111.229.729 1.172.861 2005 902 143.428.457 2.293.079 2006 997 131.265.362 1.914.931 2007 1077 161.039.545 2.763.099 2008 1209 133.912.092 14.718 2009 1345 114.508.724 1.383.862 2010 1619 128.935.861 3.928.804 2011 1668 139.469.463 3.258.016 2012 1597 143.119.583 3.643.322 2013 677 90.355.005 2.720.943
Table 3 presents the means and standard deviations for ROS, market share and
concentration for the travel agent industry sample. The figures indicate that average ROS, market share and concentration levels over the whole period are very low. Table 3 also shows high standard deviations and skews for all variables indicating that the sample is normally distributed and there are large differences between the travel agents in terms of profitability, market share and individual concentration scores.
Table 3: Descriptive Statistics for the Travel agent industry (2004-2013)
Statistic Return on Sales Market Share Concentration
Mean ,0178 ,0068 ,0014
Standard Deviation ,26859 ,03518 ,01279
Skew -3,194 9,030 11,368
In order to analyze the Dutch hotel industry concentration a similar procedure for data collection was followed. The industry identifier that was used is 551, which stands for Hotels
21 and similar accommodations. As output variables the companies’ turnover and profit and loss for a period starting 2004 and ending 2013 were defined. The results were then filtered on geographical location using the Netherlands, which led to 193 results. Out of the results from the query only 109 had the year 2013 as most recently available. Out of the 109 results, 47 did not have turnover data available. 62 observations remained and together accounted for over €2.1B in turnover in 2013, with an average of €34.2M. The average profit in 2013 for the 62 observations equaled €1.9M, while the total profit equaled €118.9M. The observed companies employ 17,317 workers, with an average of 299 employees per observation. Table 4 presents the ROS, market share and concentration level for the whole period. Average levels are very low, as was the case for the travel agent sample and again we see high standard deviations and skews.
Table 4: Descriptive Statistics for the Dutch Hotel Market (2004-2013)
Statistic Return on Sales Market Share Concentration
Mean ,0029 ,0161 ,0011
Standard Deviation ,35474 ,02763 ,00488
Skew -5,886 4,514 7,384
For the final analysis, regarding hotels’ pricing behavior given a certain market structure, the lowest rates published on Booking.com for a double room with a length of stay of 1 night for 11 hotels were collected every week on Monday at 3AM for a booking period of 90 days ahead using a rate shopper called OTAinsight. The 11 hotels together form the competitive set as defined by one of the hotels included in the set. The star ratings vary between 3 and 4 stars and the size range of the hotels included lies between 36 and 520. The review scores as published on Booking.com of the hotels included lie between 7.4 and 9 and all the hotels are located in Amsterdam within a maximum range of 5.5 kilometers of each other. In total 3,520 rates were gathered for 28 arrival dates, during 13 weeks. This number includes ‘Sold Out’ as a rate for a given date. Out of the 3,520 observation, there were 2,862 observations with rates. This means that on 678 observations no rates were published, or the rate published was not
22 meant to be sold. This is a practice two of the observed hotels employed when they still had rooms in their allotment contract with Booking.com, but did not have the physical rooms left to sell online. An allotment contract gives a travel agent the right to sell a certain amount of rooms up and until n days prior to arrival. If the agent does not succeed in selling the room before the cutoff date, then the room will be returned to the hotel’s inventory. Travel agents use allotment contracts to protect themselves against hotels that are unwilling to sell their rooms through the agent’s channel, particularly when demand is high and the hotel could easily sell all of its rooms without the intermediary, saving commission costs. Observed rates of this kind were €989.00 and €720.00 and they were labeled as missing values leaving 2,858 observations. The period used for analysis lies between May 26th and June 22nd. Some other data was collected during the same period, with regards to how many hotels were available per observation date and arrival date combination. The descriptive statistics per hotel are shown in
Table 5.
Table 5: Sample hotels price and other characteristics
Hotel N Star Size M max min STD
Hotel 1 279 3 81 129,45 240,00 65,00 32,57 Hotel 2 253 4 175 192,01 361,90 131,21 38,52 Hotel 3 232 4 215 125,48 189,05 84,55 24,69 Hotel 4 265 4 260 120,49 248,98 89,00 29,13 Hotel 5 290 4 446 141,20 300,00 90,00 32,95 Hotel 6 235 4 116 152,28 249,00 88,74 37,47 Hotel 7 268 4 520 111,51 199,00 72,09 19,72 Hotel 8 283 4 138 135,22 299,00 79,20 43,74 Hotel 9 269 3 36 144,58 285,00 80,00 39,68 Hotel 10 306 3 68 138,10 398,82 75,60 49,33 Hotel 11 178 4 43 225,73 439,00 139,00 47,49
As described in the hypotheses it is expected that the observed hotels will follow the basic principles of revenue management and that prices tend to rise as an arrival date approaches. Also, based on the benchmark figures provided by KPMG a significant difference between days of the week is expected, since the demand for weekends (Fridays and Saturdays) appears to be higher than the demand for weekday. The descriptive statistics for the observed rates are shown in
23 Table 6.
24
Table 6: Descriptive statistics hotel rates per day type
Day of week N M max min STD
Sunday 413 133,17 319,00 75,00 41,85 Monday 416 137,27 389,00 70,00 54,94 Tuesday 468 140,56 359,00 65,00 45,47 Wednesday 481 135,10 299,00 70,00 39,59 Thursday 443 141,12 439,00 84,60 42,27 Friday 371 157,02 361,90 79,00 45,09 Saturday 266 181,64 398,82 94,05 42,78 Total 2858 144,14 439,00 65,00 46,86
Figure 2, Figure 3 and Figure 4 onder present the development of the inter-temporal mean rate observed for 3 and 4 star hotels in the sample. There appear to be differences between 3 and 4 star hotels in the sample regarding to how their pricing changes inter-temporally. For 3 star hotels the pattern for week days differs from weekend days, however the 4 star sample shows a less dramatic difference, indicating less variation between week and weekend days.
Figure 2: Inter-temporal development of average rate observed over the whole sample of check-in dates for 3 and 4 star hotels
Figure 2 shows two very different pricing developments for 3 and 4 star hotels through time. 3 star hotels seem to dramatically lower their prices starting 8-9 weeks prior to arrival, but then start increasing their rates from 5 weeks prior to arrival. Figure 3 shows that the same trends occur for the week day sample. For the weekend sample however, both 3 and 4 star hotels in the sample seem to have increasing prices as check in dates approach. Interestingly,
25 3 star hotels in the sample show a higher average price for weekend check in dates than 4 star hotels do, as is presented in Figure 4.
Figure 3: Inter-temporal development of average rate observed for week day check-in dates for 3 and 4 star hotels.
26 Table 7 shows the descriptive statistics for the price and price change per booking date.
For the whole sample of 2,858 check-in dates with prices, there were 1,190 rate changes, indicating that 41.6% of the rates observed were changed during the 13 weeks of observation. This goes to show that the hotels in the sample very actively apply revenue management techniques and thus change their prices very often.
27
Table 7: Descriptive statistics per observation date for the complete check-in sample
Observation Date Variable N Mean Max. Min. STD
24.03.14 Rate 242 147,73 300,00 85,00 44,88 Rate 0 31.03.14 Rate 250 140,83 300,00 79,00 43,55 Rate 153 -7,11 88,20 -120,00 28,54 07.04.14 Rate 251 143,33 309,60 79,00 44,36 Rate 87 6,60 117,00 -93,60 35,75 14.04.14 Rate 250 144,61 300,00 84,55 42,51 Rate 53 9,02 60,00 -59,50 20,22 21.04.14 Rate 250 143,38 361,90 75,60 44,34 Rate 57 -5,26 98,70 -87,50 27,81 28.04.14 Rate 250 142,39 361,90 77,40 45,14 Rate 67 -3,83 60,00 -75,00 22,33 05.05.14 Rate 251 142,36 289,00 65,00 47,09 Rate 109 0,76 60,00 -77,40 22,16 12.05.14 Rate 250 141,50 389,00 70,00 46,69 Rate 120 0,05 130,00 -60,00 25,17 19.05.14 Rate 250 141,13 389,00 70,00 46,54 Rate 135 -1,15 55,00 -94,00 23,86 26.05.14 Rate 247 146,10 398,82 80,00 51,67 Rate 171 0,55 52,50 -80,00 22,01 02.06.14 Rate 193 149,19 439,00 79,00 55,97 Rate 117 7,78 100,00 -66,90 25,53 09.06.14 Rate 120 145,96 369,00 72,09 51,92 Rate 80 1,71 80,00 -70,00 23,81 16.06.14 Rate 54 159,96 309,00 79,00 49,54 Rate 41 -1,12 106,90 -95,00 33,80 Total Rate 2858 144,14 439,00 65,00 46,86 Rate 1190 0,37 130,00 -120,00 26,10 6 Results
In order to be able to define the level of concentration in the travel agent industry two
measures were used. First the Herfindahl Hirschman-Index (HHI) is calculated for the period between 2004 and 2013. Afterwards the N-firm concentration ratio is used. The same
measures are used to define the level of concentration in the Dutch hotel industry. In the final step, an analysis of hotel conduct on the online booking platform Booking.com is made, by analyzing the online published rates of 11 hotels over a period of 28 days starting May 26th 2014 and ending June 22nd 2014.
28 6.1 Herfindahl-Hirschman-Index for travel agents
The concentration index H= ( ) is calculated using the sum of squares of individual market share, where Si denotes a firm’s market share represented by the ratio of each firm’s
turnover to total industry turnover (Davies, 1999). The total scores of the HHI per annum are presented in Graph 1.
Graph 1: Herfindahl Hirschman Index Travel Agent Industry '04-'13
Graph 1 presents the Herfindahl-Hirschman Index of the travel agent industry for the period of 2004 to 2013. The highest score possible would be 10,000 and so the scores
between 3.15 and 13.97 are very low, indicating a competitive market. After initial shocks in the index between 2004 and 2007, the index has decreased from 12.50 in 2006 to 3.81 in 2008. This shift can be explained by the first economic recession. Since 2009 the index has remained relatively stable, showing an upward trend between 2012 and 2013, which I shall explain further on. Overall, the downward slope of the curve indicates that the market power of the largest firms in the industry has declined over time.
6.2 Four firm concentration ratio for travel agents
The CR4 ratio for the travel agent industry for the period 2004 to 2013 is depicted in graph 2.
The CR4 is calculated by summing the market shares of the top 4 firms in an industry. Higher
0 2 4 6 8 10 12 14 16 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
Herfindahl Hirschman Index Travel Agent Industry '04-'13
29 scores indicate less competition in a market, In general, if the CR4 is below 40% a market is
considered to be highly competitive. The CR4 and CR8 are presented in Graph 2.
Graph 2: Concentration Ratios Travel Agents '04-'13
The four firm concentration index and eight firm concentration index presented in Graph 2 show similar curves as graph 1, with a decline in concentration in 2008, followed by a stable period between 2008 and 2012. For both graphs the observed number of firms seems to heavily influence the curves. Since the dataset contained only 677 firms for the year 2013, compared to the 1597 firms in 2012, it is likely that the upward trend in 2013 is observed for this reason and thus should not be interpreted as an accurate presentation of the actual
developments. Given the rule of thumb that an industry with a CR4 < 0.40 is considered to be
highly competitive and that the HHI is considered to be very low, it would be reasonable to label the travel agent industry as highly competitive, but at the same time the market structure can be labeled as an oligopoly due to the high barriers to entry that were described in chapter 3 and this is supported by the outcomes of the CR8, which show that the top 8 firms in the
industry account for approximately 41% of revenues in 2012. Also, the sample of travel agents that is used in this study consists of all firms listed under the industry identification code 7911 and does not make any distinction in whether a firm is operating online, offline or both. The concentration index and the HHI would have looked differently if this distinction
51% 47% 50% 55% 34% 32% 30% 30% 31% 48% 62% 62% 59% 62% 45% 42% 39% 40% 41% 65% 0% 20% 40% 60% 80% 100% 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
Concentration Ratios Travel Agents '04-'13
CR4 CR8
30 would have been made, perhaps resulting is higher (or even alarming) ratios. In order to facilitate accurate analysis of the level of concentration in industries the NAICS system
should provide more distinctive categories to account for the different playing fields that firms populate. When looking at the revenue development of Priceline.com over the past decade we might expect there to be a shift in concentration in the year to come. Annual revenue
increased from €1.8B in 2009 to €5.2B in 2013, with an average annual growth of 30%. 6.3 Herfindahl-Hirschman-Index for Dutch hotel market
The HHI for the Dutch hotel market per year is presented in Graph 3.
Graph 3: Herfindahl Hirschman Index for the Dutch Hotels '04-'13
Graph 3 shows that for the Dutch hotel market the HHI remained relatively stable and is very low throughout the observed period, indicating a competitive market.
6.4 Four firm concentration ratio for Dutch hotel market
The CR4 ratio for the Dutch hotel market for the period 2004 to 2013 is depicted in Graph 4
on the next page. Graph 4 shows that the CR4 declined from 49% in 2004 to 33% in 2013,
indicating that the four largest firms in the market have lost market share. This is likely to be caused by the increased numbers of individual hotels which have come in to existence. Yet, despite the downward trend the Dutch hotel market can still be labeled an oligopoly or
0,00 2,00 4,00 6,00 8,00 10,00 12,00 14,00 16,00 18,00 20,00 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
Herfindahl Hirschman Index for the Dutch Hotels '04-'13
31 monopolistic competition, depending on how individual firms behave (e.g. whether they have control over the prices or not).
Graph 4: Four Firm Concentration Ratio for the Dutch Hotels '04-'13
Because of the small number of hotels that were present in the dataset the CR8 was not
presented for hotels and because of the small number of hotels present, the CR4 might
misrepresent the actual level of concentration. However, combined with the high level of supplier power in the industry it would be fair to say that there is monopolistic competition, rather than oligopoly.
A Pearson test for correlation was performed on the 10 year averages of market share, ROS, profits and concentration (HHI) for both the travel agent industry and the Dutch hotel industry. For the travel agent industry this yielded a strong positive relationship (r = .845, p
< 0.01) between market share and profit. The same relationship was found in the Dutch hotel
sample (r = .840, p < 0.01). HHI and average profits showed a weak positive relationship (r
= 0.77, p < 0.01) for the travel agent sample and a strong positive relationship (r = 0.836, p < 0.01) for the Dutch hotel sample. Lastly, no significant relationships between ROS and
concentration were found in either sample.
49% 45% 42% 42% 41% 37% 38% 34% 33% 33% 0% 20% 40% 60% 80% 100% 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
Four Firm Concentration Ratio for the Dutch Hotels '04-'13
32 6.5 Technical analysis pricing based competition
In order to determine how time is related to the rates offered in the market the correlation between the number of days left until the arrival date and the average rate offered by the hotels was tested. First a Pearson Correlation test was performed on the whole sample of arrival dates and no significant correlation was found. The dataset was then split to divide week days and weekend days. A weak negative relationship (r = -.079, p = 0.039 ) between number of days left before arrival and average rate offered was found for weekend days, meaning that prices tend to rise as the day of arrival approaches. For week days there was no such significant relationship. In the next step size, star rating and availability were tested for correlations. The results are depicted in
Table 8 below.
Table 8: Correlation, means and standard deviations
In order to test the expectations presented in section 4, a multiple regression analysis was performed with size and star ratings as control variables. In the second step the
M STD 1 2 3 4 5 6
Whole Sample Size 195,53 156,25 -
Star 3,70 0,46 ,557** - DayType 3,81 1,88 ,004 -,051** - Avail 9,75 1,72 ,009 ,077** -,341** - DaysDiff 41,74 23,83 -,017 ,002 ,051** ,142** - Rate 144,14 46,86 -,254** ,095** ,235** -,383** ,001 - Weekend Size 187,92 148,92 - Star 3,68 0,47 ,582** - DayType 3,35 2,93 -,021 -,096* - Avail 9,16 2,27 ,110** ,165** -,550** - DaysDiff 43,93 23,98 -,028 -,013 ,069 ,054 - Rate 152,16 48,38 -,276** -,009 ,490** -,527** -,079* - Weekday Size 197,90 158,42 - Star 3,71 0,45 ,550** - DayType 3,95 1,36 ,013 -,035 - Avail 9,94 1,46 -,045* ,029 -,207** - DaysDiff 41,05 23,74 -,013 ,009 ,058** ,208** - Rate 141,64 46,10 -,247** ,135** ,112** -,306** ,021 -
33 independent time variable daysdiff was added to the model to check for the inter-temporal effect that occurs when an arrival day approaches. In the final step independent variable avail (total number of hotels with availability for a certain check-in date) was added in order to account for increasing demand and thus to operationalize scarcity. The results for the whole sample of check-in dates, weekend dates and weekdays are shown in
Table 9 Results for Daysdiff and Avail as predictors for Rate
Variables
Whole Sample (1) Weekend (2) Weekdays (3)
B SE B SE B SE Step 1 Size -,13 ,01 -,45* -,13 ,01 -,41* ,01 ,01 -,46* Star 35,17 2,13 ,34* 23,73 4,61 ,23* 39,30 2,38 ,39* Step 2 Size -,13 ,01 -,45* -,13 ,01 -,41* -,13 ,01 -,46* Star 35,19 2,13 ,34* 23,78 4,59 ,23* 39,27 2,38 ,39* DaysDiff -,01 ,03 -,01 -,18 ,73 -,09** ,02 ,04 ,01 Step 3 Size -,13 ,01 -,47* -,13 ,01 -,40* -,14 ,01 -,49* Star 39,56 1,92 ,39* 32,10 3,85 ,31* 41,89 2,21 ,41* DaysDiff ,10 ,03 ,05* -,12 ,61 ,06 ,17 ,04 ,09* Avail -11,32 ,43 -,42* -11,30 ,65 -,53* -11,25 ,59 -,36*
(1) N = 2858. R2 = .146* for step 1. R2 = .145 for step 2. R2 = .314* for step 3. * p < .01
(2) N = 679. R2 = .108* for step 1. R2 = .115** for step 2. R2 = .388* for step 3. * p < .01 ** p < .05
(3) N = 2179. R2 = .165* for step 1. R2 = .164 for step 2. R2 = .286* for step 3. * p < .01 ** p < .05
.
Table 9 Results for Daysdiff and Avail as predictors for Rate
Variables
Whole Sample (1) Weekend (2) Weekdays (3)
B SE B SE B SE Step 1 Size -,13 ,01 -,45* -,13 ,01 -,41* ,01 ,01 -,46* Star 35,17 2,13 ,34* 23,73 4,61 ,23* 39,30 2,38 ,39* Step 2 Size -,13 ,01 -,45* -,13 ,01 -,41* -,13 ,01 -,46* Star 35,19 2,13 ,34* 23,78 4,59 ,23* 39,27 2,38 ,39* DaysDiff -,01 ,03 -,01 -,18 ,73 -,09** ,02 ,04 ,01 Step 3 Size -,13 ,01 -,47* -,13 ,01 -,40* -,14 ,01 -,49* Star 39,56 1,92 ,39* 32,10 3,85 ,31* 41,89 2,21 ,41* DaysDiff ,10 ,03 ,05* -,12 ,61 ,06 ,17 ,04 ,09* something is missing
34
Avail -11,32 ,43 -,42* -11,30 ,65 -,53* -11,25 ,59 -,36*
(4) N = 2858. R2 = .146* for step 1. R2 = .145 for step 2. R2 = .314* for step 3. * p < .01
(5) N = 679. R2 = .108* for step 1. R2 = .115** for step 2. R2 = .388* for step 3. * p < .01 ** p < .05
(6) N = 2179. R2 = .165* for step 1. R2 = .164 for step 2. R2 = .286* for step 3. * p < .01 ** p < .05
The results show that size has a negative effect on room price, regardless of whether the check-in date is a weekday of weekend day. Also, the results confirm the expectation that the number of available hotels for a specific check-in date has a negative effect on room price. The expectation that there would be a negative effect of the number of days left until a certain check in date on the room prices is only partly confirmed, since this effect is only significant on the weekend check-in sample. I find that for the whole sample of check in dates the rate can be estimated as follows;
Rate = B1 + B2size + B3star + B4daysdiff + B5avail (1)
This means the price for a 3 star hotel with 81 rooms 10 days before arrival with all hotels available equals; 131.24 + 81 * -.14 + 3 * 39.56 + 10 * .10 + 11 * -11.32 = € 115.06.
After establishing the overall price trends, in the next phase, regression analyses were used to estimate the trend of variation for each hotel and check in date combination. A positive trend indicates that on average the hotel’s prices for the observed period increased, while a negative trend indicates a decrease.
35
Figure 5: Average trends in the percentage rate change for the total arrival sample (panel A), the weekday arrival sample (panel B) and the weekend arrival sample (Panel C).
In Figure 5 panel A the line increases beyond the intersection point of hotel number 8, indicating a positive change in rate. The cluster of hotel numbers, in the order of 9, 1, 5, 10, and 6 have in common that they have had on average a positive price change in the 13 weeks prior to the average arrival date. Panel B shows the same structure for the weekday arrivals, except that hotel number 5 is missing. Panel C plots the weekend arrivals with a positive trend for a larger group of hotels. In Panel A-C, hotel number 6 appears to have the largest
percentage in rate increases.
Table 10 shows the output of the regression analysis in which hotel 6 is used as a
benchmark, since it showed the biggest positive average rate change over the observed period. This analysis helps us to zoom in on the hotels included in the sample individually and to determine whether the variations between them are significant for the independent variables
avail, daysdiff and dummy variable weekend. Additionally, a new independent variable
(lnRatet-1) is created to form a lag in the prices published of 1 week, which operationalizes the
concept of price reaction in this study. The results indicate that hotels 3 and 4 show an -2.00% -1.50% -1.00% -0.50% 0.00% 0.50% 1.00% 1.50% H11 H4 H7 H2 H3 H8 H9 H1 H5 H10 H6 -2.50% -2.00% -1.50% -1.00% -0.50% 0.00% 0.50% 1.00% 1.50% H4 H11 H7 H2 H5 H3 H8 H9 H10 H1 H6 -1.00% 0.00% 1.00% 2.00% 3.00% 4.00% 5.00% H3 H8 H4 H1 H2 H9 H7 H11 H6 H10 H5 Panel A Panel B Panel C
36 increase in price of 13.1% and 14.7% (p < 0.05) relative to hotel 6 compared to the prices a week earlier. Hotel 3 and 4 are both 4 star properties and are both relatively large with over 200 rooms, which could accounts for a large part of their similarity in pricing strategy, which differs from hotel 6. In terms of price reaction to a change in availability hotels 1, 8, 9 and 10 have prices that are 2.3 – 5.2% lower (p < 0.01). Interestingly, these hotels have relatively few rooms and therefore might respond more aggressively than hotel 6. In terms of time reaction it is logical that all the other hotels show lower prices than hotel 6, since hotel 6 showed the greatest average positive trend. However, in the weekend hotel 6’s rates were lower than those of hotel 3, 5 and 11. From this very focused analysis of individual firm behavior the main takeaway is that firms deploy different pricing strategies however, which are likely to be determined by their star rating and size.
In the final stage of the analysis the hotels in the clusters that were formed based on their price trends for A) the whole sample, B) the weekday sample and C) the weekend sample were compared to the hotels that were not included in the sample. Hotels 1, 5, 6, 9 and 10 form cluster I, hotels 1, 6, 9 and 10 form cluster II and all hotels except hotel 3 and 8 form cluster III. These clusters are used in the next phase of the analysis in order to determine to what extent hotels’ rate changes trigger reactions from competing hotels and whether this reaction is different for hotels belonging to the cluster and hotels not belonging to the cluster for any check in date, weekdays or weekend days. These clusters represent hotels with similar pricing strategies for a certain booking date, e.g. the week days.
37
Table 10: Regression output. Dependent variable is natural logarithm of the rate.
Price reaction H1 -0.189 (0.158) H2 -0.100 (0.350) H3 -0.434*** (0.152) H4 -0.543*** (0.193) H5 0.000 (0.000) H7 -0.192 (0.164) H8 -0.196* (0.113) H9 0.000 (0.000) H10 -0.272* (0.153) H11 0.170 (0.245) lnRatet-1 0.022 (0.026) H1 X lnRatet-1 0.046 (0.044) H2 X lnRatet-1 0.141 (0.107) H3 X lnRatet-1 0.131** (0.059) H4 X lnRatet-1 0.147** (0.068) H5 X lnRatet-1 0.019 (0.051) H7 X lnRatet-1 0.042 (0.046) H8 X lnRatet-1 0.044 (0.045) H9 X lnRatet-1 -0.014 (0.048) H10 X lnRatet-1 0.073* (0.042) H11 X lnRatet-1 0.082 (0.062) Star 0.104 (0.208) lnSize -0.073 (0.108) Intercept 4.893*** (0.454) Obs 501 R2 (adj R2) 0.342 (0.313) F-statistic 11.836 p(F) 0.000 Availability reaction H1 -0.008 (0.115) H2 0.430*** (0.093) H3 -0.075 (0.118) H4 0.211** (0.105) H5 0.260*** (0.099) H7 0.000 (0.000) H8 0.420*** (0.087) H9 -0.019 (0.157) H10 -0.111 (0.104) H11 0.440* (0.230) Avail -0.060*** (0.007) H1 X Avail -0.027** (0.010) H2 X Avail -0.005 (0.010) H3 X Avail 0.010 (0.012) H4 X Avail -0.018 (0.011) H5 X Avail 0.008 (0.009) H7 X Avail 0.018* (0.011) H8 X Avail -0.052*** (0.009) H9 X Avail -0.040*** (0.011) H10 X Avail -0.023*** (0.008) H11 X Avail -0.028 (0.020) Star 0.000 (0.000) lnSize -0.305*** (0.071) intercept 7.030*** (0.381) Obs 2,858 R2 (adj R2) 0.555 (0.551) F-statistic 168.228 p(F) 0.000 Time reaction H1 0.062 (0.048) H2 0.245*** (0.040) H3 -0.103*** (0.039) H4 -0.063* (0.036) H5 0.096** (0.038) H7 0.000 (0.000) H8 -0.158*** (0.040) H9 0.000 (0.000) H10 0.150*** (0.045) H11 0.199*** (0.068) DysLeft -0.001** (0.001) H1 X DaysDiff 0.003*** (0.001) H2 X DaysDiff 0.002** (0.001) H3 X DaysDiff 0.001 (0.001) H4 X DaysDiff 0.000 (0.001) H5 X DaysDiff 0.003*** (0.001) H7 X DaysDiff 0.001 (0.001) H8 X DaysDiff 0.002* (0.001) H9 X DaysDiff 0.003*** (0.001) H10 X DaysDiff 0.001 (0.001) H11 X DaysDiff -0.000 (0.001) Star 0.421*** (0.068) lnSize -0.218*** (0.029) intercept 4.402*** (0.153) Obs 2,858 R2 (adj R2 0.367 (0.363) F-statistic 78.361 p(F) 0.000
Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1
38 Table 10 Continued. Weekend reaction H1 0.089* (0.047) H2 0.337*** (0.038) H3 -0.136*** (0.038) H4 -0.105*** (0.039) H5 0.151*** (0.042) H7 0.000 (0.000) H8 -0.062 (0.040) H9 0.000 (0.000) H10 0.112** (0.044) H11 -0.013 (0.071) weekend -0.109*** (0.035) H1 X Weekend -0.018 (0.048) H2 X Weekend 0.005 (0.049) H3 X Weekend 0.135*** (0.050) H4 X Weekend 0.089* (0.050) H5 X Weekend 0.128*** (0.048) H7 X Weekend 0.075 (0.052) H8 X Weekend -0.034 (0.049) H9 X Weekend -0.008 (0.049) H10 X Weekend -0.047 (0.047) H11 X Weekend 0.243*** (0.060) Star 0.320*** (0.067) lnSize -0.231*** (0.030) Constant 4.892*** (0.148) Obs 2,858 R2 (adj R2) 0.381 (0.377) F-statistic 83.244 p(F) 0.000
39
Fout! Ongeldige bladwijzerverwijzing. shows that the hotels in cluster I had 10.3%
lower rates than hotels not included in the cluster. Other significant differences between the clusters were found for availability and weekend, but not for time left before arrival. Tables 12, 13 and 14 with the regression output are appended.
Table 11: Price reaction to the clusters of hotels. See Figure 5 Panel A-C.
(1) (2) (3) clusterHotels_I 0.286*** (0.075) clusterHotels_II 0.150* (0.081) clusterHotels_III 0.228** (0.089) lnRatet-1 0.148*** 0.131*** 0.097*** (0.020) (0.018) (0.032) clusterHotels_I X lnRatet-1 -0.103*** (0.025) clusterHotels_II X lnRatet-1 -0.083*** (0.025) clusterHotels_III X lnRatet-1 -0.025 (0.035) Star 0.286*** 0.230*** 0.367*** (0.045) (0.046) (0.039) lnSize -0.166*** -0.177*** -0.197*** (0.023) (0.024) (0.023) Constant 4.325*** 4.670*** 4.204*** (0.156) (0.193) (0.130) Observations 501 501 501 R-squared 0.219 0.213 0.233 adj. R-squared 0.211 0.206 0.225 F-statistic 27.680 26.870 30.019 p(F) 0.000 0.000 0.000
Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1
clusterHotels_I: See Figure 5 Panel A; clusterHotels_II: See Figure 5 Panel B; clusterHotels_III: See Figure 5 Panel C.
The results of the analysis of pricing behavior indicate that different strategic clusters of hotels can be identified within this particular competitive set. The regression analyses allowed to statistically differentiate between these groups based on their overall price trends for different sets of arrival dates. Therefore we cannot reject the hypothesis that hotels’ pricing decisions are influenced by their competitors’ pricing decisions. This does not imply that the hotels in the competitive set actually do follow each other’s prices, but it is likely that the clusters of hotels that were identified in the data are similar properties and represent a
40 certain level of quality in terms of size, star rating, etc. and hence rely on the same approach to pricing.
7 Conclusion and discussion
Recently there have been reports of uncompetitive conduct in the hospitality industry. Examples include the investigations into price fixing practices and consumer deception by some of the larger online travel agencies. The aim of this study was to provide insights in the level of concentration in both the international travel agent industry and the Dutch hotel market and the applicability of the SCP paradigm to the aforementioned industries. I find that both industries can be labeled as competitive and that there are no high levels of
concentration, measured by either the Herfindahl-Hirschman Index or the n-firm
concentration ratio. The author does note that there might be a shift in concentration in the coming years and suggests further separating the NAICS industry codes to account for the differences in playing fields for the travel agent industry, e.g. online and offline. The travel agent industry can be defined as an oligopoly due to its high barriers to entry, while the market structure in the Dutch hotel market can be defined as monopolistic competition. In the next step this study showed how 11 hotels in Amsterdam belonging to a competitive set compete on prices for 28 check-in days in the period between May 26th and June 22nd. As far as I am aware this kind of inter-firm analysis with respect to actual pricing behavior has never been performed for the hotel industry. In general I find that availability, star rating, size and time left before check-in affect prices as expected. However, for the latter effect the sample showed a clear distinction between weekdays and weekend check-in dates, with the effect not being present in the former. The inter-firm reaction to price changes is small but significant and provides ground for further investigation using a larger sample of hotels.
The question is: do have the travel agency - website profit power over the hotels, is the capability of the hotels to compete, to be profitable, impaired or not by the concentration / structure of the travel agency ?
41
8 Limitations
No research is perfect and therefore I would like to address the limitations of the current study. The first limitation lies in the classification of industries that is used on a global scale to assess markets. During my research I noticed that the industry codes are defined very broadly, leading to pollution and preventing clear analysis. It would be very interesting to have a clear distinction between travel agents that operate online (B2C) and travel agents that operate offline (mostly B2B), however that was not possible given the tight timeframe. The second limitation lies in the sample size of the Dutch hotels. Since the available databases contained financial data only 62 hotels in the Netherlands, the results presented here might not present an accurate overview of the level of competition in the industry. Moreover, because of the sensitive nature of a firm’s publications of its financials there might be deliberate
distortion of data. Deliberate distortion is most likely to occur in industries where the threat of potential entrants is high and there are few existing rivals (Li, Lundholm & Minnis, 2012, p. 5).
Regarding the hotel pricing analysis, the sample size consisted of only 11 hotels, mainly caused by budget restraints on the author’s account. This means that the results and conclusions presented in this study might not represent the market or industry as a whole, but I feel that the results provide a unique insight in actual firm behavior and that the limited sample size is offset by an extensive observation period of 13 weeks combined with a relatively large sample of check in dates. In an ideal situation however, it would be better to increase the sample size. Also, there were no five star hotels included in the study and the results might be different for other hotel classifications, regions, months, and so on.
42
9 References
Abrate, G., Fraquelli, G., and Viglia, G. (2012), Dynamic pricing strategies: Evidence from European hotels. International Journal of Hospitality Management, 31(1), 160-168.
Bailey, J.P. and Bakos, Y. (1997), An Exploratory Study of the Emerging Role of Electronic Intermediaries, International Journal of Electronic Commerce, Vol. 1, No. 3, 7-20
Bain, J. (1956), Barriers to New Competition: Their Character and Consequences in Manufacturing Industries, Harvard University Press, Cambridge, MA
Barthwal, R. R. (2004). Industrial Economics: An introductory textbook. 2nd edition. New Dehli: New Age International Publishers
Bikker, J.A. and van Leuvensteijn, M. (2008), Competition and efficiency in the Dutch life insurance industry, Applied Economics, 40:16, 2063-2084
Boone, J. (2008), A New Way to Measure Competition, The Economic Journal 118 (531), 1245-1261
Boone, J., Griffith, R. and Harrison, R. (2004), Measuring competition, Encore Meeting Clemons, E. K., Hann, I. H., & Hitt, L. M. (1998). The nature of competition in electronic markets: An empirical investigation of online travel agent offerings, The Wharton School of
the University of Pennsylvania.
Collins, M., and Parsa, H. G. (2006). Pricing strategies to maximize revenues in the lodging industry, International Journal of Hospitality Management, 25(1), 91-107.
Conner, K. (1991), A historical comparison of resource-based theory and five schools of thought within industrial organization economics: Do we have a new theory of the firm?
Journal of Management, 17, 121-154
Davies, B. (1999), Industrial organization: The UK Hotel Sector, Annals of Tourism
Research, Vol. 26, No. 2, 294-311
De Kuijper, M. D. (2009), Profit Power Economics: A New Competitive Strategy for Creating Sustainable Wealth, New York: Oxford University Press
Evans, D. S. (2006), Invisible Engines: How Software Platforms Drive Innovation and Transform Industries, The MIT Press, Cambridge, Massachusetts.
