Price dispersion across online platforms: evidence from hotel room prices in London (UK)

ABSTRACT

This paper studies the widespread price dispersion of homogeneous products across different online platforms, even when consumers can easily access price information from comparison websites. We collect data for the 200 most popular hotels in London (UK) and document that prices vary widely across booking sites while making reservations for a hotel room. Additionally, we find that prices listed across different platforms tend to converge as the booking date gets closer to the date of stay. However, the price dispersion persists until the date of stay, implying that the law of one price does not hold. We present a simple theoretical model to explain this and show that in the presence of aggregate demand uncertainty and capacity constraints, price dispersion could exist even when products are homogeneous, consumers are homogeneous, all agents have perfect information about the market structure, and consumers face no search costs to acquire information about the products. Our theoretical intuition and robust empirical evidence provide additional insights into price dispersion across online platforms in different institutional settings. Our study complements the existing literature that relies on consumer search costs to explain the price dispersion phenomenon.

KEYWORDS:

• Price dispersion
• online platforms
• law of one price
• hotel booking

JEL CLASSIFICATION:

• D40
• D81
• L83
• L86

Acknowledgments

The authors are grateful to Frank Verboven, Jan De Loecker, Jo Van Biesebroeck, Luis Vasconcelos, for invaluable help, advice, and encouragement. We sincerely thank the editor and the two anonymous referees for their detailed comments which led to significant improvements in the content and exposition of the paper. We also sincerely thank the seminar participants at KU Leuven and at the University of Essex for their valuable comments and helpful suggestions. Lastly, we thank Sunhyung Lee for helpful discussions.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/00036846.2023.2275219

Notes

¹ See Baye et al. (2006), Baye et al. (2006), Escobari (2012), Brynjolfsson and Smith (2000) for some of the evidences available in recent literature.

² The coefficient of variation is given by the ratio of the standard deviation to the average price and captures the extent of variability about the average price.

³ See De Silva et al. (2019)- a study of price dispersion over time and across geographical location, Chakrabarty and Kutlu (2014) a study of price dispersion and inter-firm, inter-flight and frequency competition, Xing (2010)- a study of price dispersion for online branches of Multi-Channel Retailers and online-only retailers for recent references, Choi et al. (2019)- a study of price dispersion for gasoline in 25 regions of Seoul, Korea in response to asymmetric information between retailers and consumers.

⁴ We used the web services provided by data extraction platform grepsr.com to scrape the data for this study.

⁵ A hotel might offer more room types (such as a suite or a family room) that may accommodate more than two guests. We did not collect information for those rooms in order to keep the set of products homogeneous.

⁶ Skyscanner uses consumer reviews, hotel star ratings (2-star, 3-star, 4-star, and 5-star), and other details from booking and text reviews to rank the hotels by popularity.

⁷ Note that, for the date of stay 7 Nov 2017, we have data from 14 dates of booking prior to the date of stay. Similarly, for the date of stay 13 Nov 2017, we have data from 17 dates of booking prior to the date of stay. For all other 5 dates-of-stay, we have data from 15 days prior to the date of stay. Therefore, we have 106 regressions in total.

⁸ In this model, the only factors contributing to price dispersion are capacity constraint and demand uncertainty.

⁹ Note that if the law of one price holds on the date of stay, then $C{V}_t$ should be close to zero. This implies that the regression coefficient of $C{V}_t$ on $C{V}_{t-1}$ would be very close to zero. As Figure 5 shows, the confidence interval for the regression coefficient of $C{V}_t$ on $C{V}_{t-1}$ is bounded away from zero and is close to 1, suggesting that price dispersion persists.

¹⁰ Note that in model 7, we include hotel id-date of stay-room type fixed effects. Hence, we can not include controls such as the number of hotel reviews, hotel star rating, and hotel review rating, as those variables are absorbed by these fixed effects.