Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
This article was downloaded by: [Raya Vilchez, Josep Maria][Consorci de Biblioteques Universitaries de Catalunya] On: 18 November 2010 Access details: Access Details: [subscription number 789296667] Publisher Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 3741 Mortimer Street, London W1T 3JH, UK Applied Economics Letters Publication details, including instructions for authors and subscription information: http://www.informaworld.com/smpp/title~content=t713684190 The effect of time on hotel pricing strategy Josep Maria Rayaa a Departament of Economia i Empresa, Escola Universitaria del Maresme (Universitat Pompeu Fabra), Barcelona, Spain First published on: 18 November 2010 To cite this Article Raya, Josep Maria(2010) 'The effect of time on hotel pricing strategy', Applied Economics Letters,, First published on: 18 November 2010 (iFirst) To link to this Article: DOI: 10.1080/13504851.2010.532091 URL: http://dx.doi.org/10.1080/13504851.2010.532091 PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf This article may be used for research, teaching and private study purposes. Any substantial or systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. Applied Economics Letters, 2010, 1–5, iFirst The effect of time on hotel pricing strategy Downloaded By: [Raya Vilchez, Josep Maria][Consorci de Biblioteques Universitaries de Catalunya] At: 17:59 18 November 2010 Josep Maria Raya Departament of Economia i Empresa, Escola Universitaria del Maresme (Universitat Pompeu Fabra), C/Ernest Lluch, s/n 08302 Mataro, Barcelona, 08005, Spain E-mail: [email protected] Tourist product distribution over the Internet is encouraging companies to implement dynamic pricing policies. The aim of this article is to present an empirical model of the dynamics of room prices in tourist resorts on the Catalan coast. We estimate a discrete time duration model for the probability of a price change occurring at any particular time and a count model for the number of price changes occurring over the period. The results suggest that the largest marginal effects are caused by a change in the location, the hotel category and the market share. I. Introduction The pricing strategy used by a hotel has an immediate impact on the hotel’s yield. The increase in tourist product distribution over the Internet is encouraging companies to implement dynamic pricing policies, changing their prices over time depending on bookings up until that moment and competitive pressure. In tourism, most studies in the empirical literature on hotel price determinants have been conducted using hedonic price models. This type of model (Rosen, 1974) shows how heterogeneous products are composed of various characteristics and the implicit marginal price can be known by estimating a model (hedonic price model) that accounts for the price of a product in terms of its characteristics. For the case of hotel prices, the studies by Coenders et al. (2003), Espinet et al. (2003), Rigall (2004) and Uriel and Ferri (2004) are of particular interest. The hedonic price model was used by Hartman (1989) to identify pricing strategies for luxury hotels. In contrast, few empirical studies analyse hotel prices from the viewpoint of competition (Davies and Downward, 1998; Davies, 1999). Among them, Baum and Mudambi (1995) show the short-term inelasticity of supply in the tourist trade and the uncertainty of demand. All these papers use economic theory to build pricing models in which the effect of time on pricing is not introduced. There is, however, an ample literature that addresses the temporal dimension of hotel pricing, albeit indirectly. Dynamic pricing of hotel rooms is a standard practice of yield management (Kimes, 1989), which has become widespread with the increase in the use of the Internet as a means of tourist product distribution. This practice involves price discrimination between consumers, the chosen indicator of price sensitivity being how early the booking is made (Gallego and van Ryzing, 1994; Zhao and Zheng, 2000; Zhou et al., 2005). A higher occupancy rate is considered to induce higher prices, up until a particular time at which, if the occupancy is not 100%, the price will decrease because of the perishable nature of the tourist product. The aim of this article is to present an empirical model of the dynamics of room prices in a sample of hotels in different tourist resorts on the Catalan coast. The article is structured as follows. First we will present the methodological framework of the article and then the data used to make the empirical estimates. This will be followed by a discussion of the main results, and finally by our conclusions. Applied Economics Letters ISSN 1350–4851 print/ISSN 1466–4291 online # 2010 Taylor & Francis http://www.informaworld.com DOI: 10.1080/13504851.2010.532091 1 J. M. Raya 2 Downloaded By: [Raya Vilchez, Josep Maria][Consorci de Biblioteques Universitaries de Catalunya] At: 17:59 18 November 2010 II. Methodological Framework To model the probability of not continuing in a certain state (constant pricing) in which the individual has spent t periods, based on a group of variables, the most suitable method is survival or duration models in discrete time. In this type of model, time is measured in discrete intervals. For the case that concerns us, the state of the price of a hotel room is observed, from time t = 1 to time t = Ti, whether the hotel keeps the price constant. At Ti, the event is completed and a sequence of states is thus obtained for that period. The survival time in a certain state (constant pricing or dynamic pricing), Ti, is a discrete random variable. The hazard function, which includes the probability of exiting a state in the period t conditional on spending t periods in this state ðhit Þ, is hit ¼ PrðTi ¼ tjTi tÞ ð1Þ Once a new binary variable is defined as yit = 1, if the hotel i reaches the situation of success (completes the event, i.e. modifies its price) at time t and yit = 0 otherwise, the likelihood function for the whole sample can be written as log L ¼ ti n X X yij log i¼1 j¼1 ti n X X hij 1 hij þ ti n X X logð1 hij Þ i¼1 j¼1 yij log hij þ ð1 yij Þ logð1 hij Þ i¼1 j¼1 ð2Þ The above expression has exactly the same form as the standard likelihood function for a binary choice model in which yit is the dependent variable and in which the data have been reorganized in hotel-period form (i.e. as many observations per hotel as periods to success). So it turns out that there is a simple way of estimating these discrete time probability models following Jenkins (1995). That is, reorganizing the data in hotel-period form, choosing the functional form hit and estimating the model using any one of the wellknown binary choice models. Second, a count model will be estimated for the number of times the hotel has changed its price over the period observed. Count models (Poisson and negative binomial) are used when the dependent variable is discrete and non-negative and has a right-asymmetrical distribution. The dependent variable usually includes few values. Estimating this type of model for this type 1 of variable solves the inconsistency problems of Ordinary Least Squares (OLS) estimation (as the dependent variable is non-negative and discrete) or multinomial logit estimation (intended for when the number of options is not very high). Furthermore, the Poisson model is a case that is embedded in the negative binomial one in which the mean and the conditional variance of the dependent variable are assumed to be equal.1 Thus, the probability of y taking the value l, conditional on x, where b are the parameters to estimate, is l exp½eðxbÞ · ½eðxbÞl P y¼ ¼ x l! ð3Þ So, given a random sample, we can obtain the likelihood function: LðbÞ ¼ n X fyi xi b exi b g ð4Þ i¼1 III. Data The sample consists of 111 hotels in 3 tourist resorts on the Catalan coast: Calella, Santa Susanna and Lloret de Mar.2 In each of these hotels, the price of a double room in the first week of August 2008 was monitored throughout the period running from the second week in January to the last week in July. As can be seen in Fig. 1, 30.67% of the hotels changed their price in the course of the 30 weeks they were observed. In addition, 16% changed the price once, 18.67% twice and 34.67% more than twice (up to a maximum of six times). The variables used in the study are as follows. Dependent variables Changes: dependent variable of the duration model that takes the value 1 if the hotel changed the price in a given week with respect to the previous week, and 0 if it did not. Number of changes: dependent variable of the count model, comprising the total number of price changes made by each hotel over the period analysed. Explanatory variables Share: number of hotel rooms with respect to the total number of rooms in that town. Town: Lloret de Mar, Santa Susanna or Calella The assumption of equality between the mean and the variance which gives rise to the Poisson model was rejected, for our data, at the significance level of 5% by means of the usual test for the alpha parameter (H0: a = 0). 2 The total hotel supply in these 3 towns consists of 127 hotels. The effect of time on hotel pricing strategy 3 35 30 25 20 Number of changes 15 10 Downloaded By: [Raya Vilchez, Josep Maria][Consorci de Biblioteques Universitaries de Catalunya] At: 17:59 18 November 2010 5 0 1 2 3 Fig. 1. 4 5 6 7 Number of price changes made by hotels over the period Category: number of stars of the hotel (two-, threeand four-star hotels). Week: week in which the price is under observation. Starting price: the price set in the second week of January for occupancy of a double room in the first week of August. IV. Results Table 1 presents the descriptive statistics of the variables of the study. It shows that the price was changed in 3.41%3 of the observations. In addition, the mean number of price changes made by one hotel was 2.01. Regarding the explanatory variables, the average market share of each hotel was 3.64%, whereas the average price observed in the first week (i.e. the second week of January) was E130.25. Santa Susanna accounted for 26.66% of the hotels in the sample, Lloret de Mar 37.33% and Calella the remaining 36%. Finally, 6.75% of the hotels had two stars, 66.26% three stars and 26.98% four stars. Table 2 shows the results of estimating a duration model for the probability of a price change occurring, without any change having taken place previously. Note that a larger market share and a higher starting price increase the probability of a price change in any given week. With regard to market share, this effect seems to indicate that hotels with a larger market share are more professionalized and probably make more use of tools such as yield management. As far as the starting price of the room is concerned, a higher starting price also may be indicative of a more aggressive price discrimination policy (expecting to gain the maximum possible revenue from guests who book late) that therefore requires a dynamic price policy to also meet targets in terms of occupancy rate. Revenue management is executed more closely on average by hotels that price above their competitive set than by those that price below their competitive set. Table 1. Descriptive statistics of the variables of the study Table 2. Duration model (dependent variable: probability of a price change occurring (transition)) Variable Transition Number of changes Share Town Lloret de Mar Santa Susanna Category Two stars Three stars Four stars Starting price 3 Mean SD 0.034 2.01 0.036 0.181 1.906 0.030 0.266 0.373 0.442 0.483 0.067 0.662 0.269 130.25 0.251 0.472 0.443 30.77 Coefficient Category Three stars Four stars Town Lloret de Mar Santa Susanna Share Starting price Week Constant log L Z-Statistic 0.733 1.630 1.27 2.59 0.937 1.031 11.007 0.017 0.029 -6.234 -198.79 2.10 2.15 1.66 4.43 1.71 -6.34 In fact, there was a price change in 7.95% of the weeks, because, as mentioned earlier, 34.67% of the hotels made more than one price change. J. M. Raya Downloaded By: [Raya Vilchez, Josep Maria][Consorci de Biblioteques Universitaries de Catalunya] At: 17:59 18 November 2010 4 Furthermore, as the weeks pass without any price change having occurred, a price change becomes more probable. This is a consequence of the perishable nature of the tourism product, given that as the weeks go by and revenues have already been maximized by charging higher prices to guests who book ahead, the objective becomes the maximization of the occupancy rate. With regard to category, four-star hotels are observed to have a lower probability than two- and three-star hotels of making a price change at any given time, without a change having already taken place. The fact that the supply of accommodation in the area is very specialized in two- and three-star hotels may account for this effect, as four-star hotels have a customer segment with a more inelastic demand, and hence require a less dynamic price policy. Lastly, in comparison with Santa Susanna, hotels located in both Calella and Lloret de Mar show a higher probability of making a price change in any given week. Again, the explanation can be found in the characteristics of the area, as Santa Susanna is more specialized in family holidays than the other two destinations, and as such enjoys a more stable and inelastic demand in relation to price. Qualitatively very similar conclusions can be reached by observing Table 3, in which we present the results of an explanatory model of the number of price changes made by a hotel over the period analysed. Thus, a higher starting price or being located in Lloret de Mar or Calella as opposed to Santa Susanna increases the number of price changes made over the period, whereas four-star hotels (in relation to two- or three-star hotels) register a lower number of price changes over the period. The only difference is that now market share is not a significant variable. Probably the greater professionalization and stronger market power of these hotels enables them to rapidly capture customers with inelastic demand and, with all the available information, reach the optimal price and so minimize the number of price changes necessary to capture the rest of the clientele. Table 3. Count model (dependent variable: number of price changes) Coefficient Category Three stars Four stars Town Lloret de Mar Santa Susanna Starting price Share Constant log L Z-Statistic 0.389 1.347 1.19 3.35 0.838 0.920 0.164 4.285 -1.719 -129.31 2.08 1.82 3.39 0.84 -2.21 Table 4. Marginal effects Transition Duration model (Table 2) Category Three -0.0205 stars Four -0.0456 stars Town Lloret de 0.0341 Mar Santa 0.0341 Susanna Share 0.0030 Starting 0.0004 price Week 0.0008 Count model (Table 3) -0.6693 -2.3160 1.8005 1.8329 0.0736 0.0283 – With the aim of obtaining quantitative effects, the marginal effects of the two models are presented in Table 4. The first column shows the effect of each explanatory variable on the probability of a price change occurring at a particular time, without one having taken place previously (0.0288 on average). Thus, the market share of the hotel being 1% larger increases the probability of a price change occurring by 0.0031, whereas the starting price of the room being E1 higher increases this probability by 0.0005. One week elapsing increases the probability of a price change occurring at any given time by 0.0008. The hotel being located in Calella or Lloret de Mar has a greater effect on the probability of a price change occurring (0.0341), as this effect more than doubles the average probability of a price change occurring. Lastly, we obtain the greatest effect for accommodation in a four-star hotel (as opposed to in a two- or three-star hotel), for which the probability of a price change occurring decreases by 0.0457. The second column of Table 4 presents the marginal effects obtained through the estimates of the count model. With an average prediction of the number of price changes of 1.72, similar results are obtained. Thus, the increase in the number of price changes is 1.80 (accommodation in Calella as opposed to Santa Susanna), 1.83 (accommodation in Lloret de Mar as opposed to Santa Susanna), 0.028 (E1 increase in the starting price) and -2.32 (accommodation in a four-star hotel as opposed to a two- or three-star hotel). V. Conclusion The aim of this article was to present an empirical model of the dynamics of room prices in a sample Downloaded By: [Raya Vilchez, Josep Maria][Consorci de Biblioteques Universitaries de Catalunya] At: 17:59 18 November 2010 The effect of time on hotel pricing strategy of hotels in different tourist resorts on the Catalan coast, with a view to obtaining the impacts of their explanatory factors. To this end, I use survival or duration models in discrete time when the dependent variable is the time at which a price change occurs and a count model (negative binomial) for the number of times the hotel has changed the price over the period observed. The results highlight that a higher starting price (indicative of a more aggressive price discrimination policy) increases the number of price changes made over the period (and the probability of a price change occurring at any given time). However, although a larger market share increases the probability of a price change occurring in a particular period (reflecting a greater use of tools such as yield management), this variable does not affect the number of price changes made. Lastly, as the weeks pass without any price change, the probability of a price change being made becomes greater, thus bearing witness to the perishable nature of the tourist good. To summarize, the results obtained provide empirical evidence of how hotels apply techniques with the aim of offering their guests the right product and selling it at the right price at the right time. This is supported by price changes and by the importance of the starting price and the time elapsing before these changes take place. Furthermore, this behaviour varies from hotel to hotel, depending on its category and location, more dynamic pricing being found when demand is more elastic, either because of the destination being undifferentiated and therefore faced with keen competition or because the hotel belongs to a category in which price competition is very high. Lastly, the fact that the data are for 2008, a year marked by a serious economic crisis and a reduction in demand for tourist services worldwide, prevents us from discriminating what part of the results are caused by the use of yield management and what part corresponds to supply responding to falling demand (what is known as ‘the dilemma of the empty room’). 5 Acknowledgements This research has benefited from research grant ECO2008-06395-C05-01 from the Spanish Ministry of Educational Science. References Baum, T. and Mudambi, R. (1995) An empirical analysis of oligopolistic hotel pricing, Annals of Tourism Research, 22, 501–16. Coenders, G., Espinet, J. and Saez, M. (2003) Predicting random level and seasonality of hotel prices: a latent growth curve approach, Tourism Analysis, 8 15–31. Davies, B. (1999) Industrial organization: the UK hotel sector, Annals of Tourism Research, 26, 294–311. Davies, B. and Downward, P. (1998) Competition and contestability in the UK Package Tour Industry: some empirical observations, Staffordshire University, Division of Economics Working Paper 98.3. Espinet, J. M., Saez, M., Coenders, G. and Fluvia, M. (2003) Effect on prices of the attributes of holiday hotels: a hedonic prices approach, Tourism Economics, 9, 165–77. Gallego, G. and Van rysing, G. (1994) Optimal dynamic pricing of inventories with stochastic demand over finite horizons, Management Science, 40, 999–1020. Hartman, R. (1989) Hedonic methods for evaluating product design and pricing strategies, Journal of Economics and Business, 41, 197–212. Jenkins, S. (1995) Easy estimation methods for discrete time duration models, Oxford Bulletin of Economics and Statistics, 57, 129–38. Kimes, S. E. (1989) The basics of yield management, Cornell Hotel and Restaurant Administration Quarterly, 30, 15–9. Rigall, R. (2004) Hisendes locals i turisme: tres assaigs, unpublished PhD thesis. Universitat de Girona. Rosen, S. (1974) Hedonic prices and implicit markets: product differentiation in pure competition, Journal of Political Economy, 1, 34–55. Thrane, C. (2007) Hedonic price models and sun-and-beach package tours: the Norwegian case, Journal of Travel Research, 43, 302–8. Uriel, E. and Ferri, J. (2004) Aplicación del enfoque hedónico para medir la evolución del precio de los hoteles en España, Papeles de Economı´a Española, 102, 141–59. Zhao, W. and Zheng, Y. S. (2000), Optimal dynamic pricing for perishable assets with nonhomogeneous demand, Management Science, 46, 375–88. Zhou, Y. P., Fan, M. and Cho, M. (2005) Online purchasing strategies for dynamically priced perishable products, Department of Management Science, Washington Business School.