COVID-19, social distancing and guests' preferences: impact on peer-to-peer accommodation pricing

ABSTRACT

This paper investigates how guests' preferences in peer-to-peer (P2P) accommodations changed during the COVID-19 summer season. Specifically, we test the importance of attributes that better allow for preserving the social distancing. To this end, we adopt a semi-parametric hedonic pricing model. We take the city of Madrid as a compelling case study of an important tourist destination severely hit by the crisis. We show that guests' marginal willingness to pay for social distancing characteristics has changed from August 2019 to August 2020. In particular, we find that whereas listings with kitchen amenities increase 15.2 percentage points their premium price in August 2020 with respect to the previous year, the marginal willingness to pay for size-related characteristics decreased by 2.7 percentage points. Results are robust to sample and time composition. This study provides meaningful findings of a shift in guests' tastes towards social distancing attributes on P2P accommodations.

KEYWORDS:

• COVID-19 pandemic
• consumer behaviour
• peer-to-peer accommodation
• hedonic modelling
• generalized additive models

Disclosure statement

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

Notes

1 For an updated literature review, please refer to Kuhzady et al. (2020).

2 We perform a Principal Component Analysis (PCA) for the number of rooms, beds and the capacity of the listing where the linear combination (first dimension) explains the 85% of the overall variance of those three variables.

3 We do not include the coefficients of the non-linear control variables in Table 2 since there are not only one coefficient attached to each non-linear variable but many, and therefore are better shown through the partial plots. For the sake of exposition, we reproduce the partial plots only for the baseline specification (Madrid, Aug 2020–Aug 2019) and we do not show the location variable coefficients due its large number. Yet, the same results hold for the other models.

4 Following Halvorsen and Palmquist (1980), to interpret correctly the coefficient attached to a dummy variable as a semielasticity in a semilogarithmic expression, we must apply the following transformation to the coefficient $g=\left(\right(\exp \left(\text{β}\right)-1)\times 100)$. To obtain the semielasticity corresponding to August 2020 those corresponding to the isolated variable and the one interacting with time must be added.