This is a literature review of the article “Disrupting the regional housing market: Airbnb in New Zealand” by Malcolm Campbell, Hamish McNair, Michael Mckay and Harvey C Perkins.

Introduction

I have chosen to review “Disrupting the regional housing market: Airbnb in New Zealand” by Malcolm Campbell, Hamish McNair, Michael Mckay and Harvey C Perkins. This article appeared in a 2019 issue of the Regional Studies, Regional Science journal.

The aim of the article was to investigate the spatial distribution of Airbnb listings across New Zealand to gain more understanding of how Airbnb may be affecting the hotel industry and the rental market.

I chose this article because after moving to a house that shares a driveway with three Airbnb listings, I became interested in the effects they can have on neighbourhoods and the development of legislation concerning them.

Variables

This article is quite brief and straightforward. In terms of variables, only two are used: Airbnb data from November 2018 obtained via code which scraped the Airbnb website, and census area unit (CAU) geographies along with their usually resident population counts from the 2013 Census.

Methods

The methods used in this article consist primarily of an interactive Shiny choropleth map coloured according to the number of Airbnb listings per 1000 usual residents, with hover text giving the CAU name, number of listings and number of usual residents.

This visualisation is supported by the authors providing the Moran’s I value of the relationship and its p-value; this is the only instance of true exploratory spatial data analysis (ESDA) in the entire article.

The data for the ten CAUs with the most listings is also presented in a simple table.

Findings and conclusion

The authors found that the spatial distribution of Airbnb listings in New Zealand follows “patterns of statistically significant mildly positive clustering”, as evidenced by the Moran’s I value of 0.33 with a p-value of ≤ 0.05 (Campbell et al., 2019, p. 141). Given the locations of these clusters – in both large urban centres and smaller regional towns popular with tourists – the authors conclude that New Zealand’s spatial distribution of Airbnb listings is a combination of these two distribution tendencies which have been identified in other studies.

They also note that for some CAUs, a high Airbnb listing count translated to a high number of Airbnb listings per 1000 usual residents. For example, the median number of listings per 1000 usual residents from the top ten CAUs shown in Table 1 is 75, yet Queenstown Hill came out with 204 listings per 1000 usual residents.

Variables evaluation

Beginning on a positive note, it was wise of the authors to discount meshblocks due to that spatial unit being too small for the maximum randomisation of the Airbnb listing locations. However, even when using CAUs instead of meshblocks, it is possible that listings located near borders ended up being counted in the wrong CAU.

Some more weaknesses centre around the code used to scrape the Airbnb data. As mentioned in the article, the method the code employs can lead to a 20% undercount of listings. Additionally, the code is no longer in working order nor does the author plan to fix it, meaning that anyone wishing to replicate this study would need to find an alternative way of getting the Airbnb data (Slee, 2019).

Regarding the use of CAUs, the authors said that this allowed them “to connect [the Airbnb data] to information about individuals, areas and households”, but they did not make any such connections in the report (Campbell et al., 2019, p. 141). This seems like a missed opportunity to me, as including more variables would have helped them to prove or disprove a relationship between Airbnb listings and the rental market or the hotel industry. For example, they could have used census variables such as dwelling type or tenure of household (specifically rent) and compared these counts to the number of entire home listings in each CAU; or they could have taken the weekly rent paid variable, averaged it for each CAU, divided it by seven and compared it to the average nightly rate charged by entire home listings in the same CAU (Stats NZ–Tatauranga Aotearoa, 2014). As for the hotel industry, relevant data is harder to find. Stats NZ’s Accommodation Survey only has data aggregated at regional and territorial authority levels and has been discontinued, but more accurate data could potentially be scraped from a hotel booking website (Stats NZ–Tatauranga Aotearoa, 2019).

Methods evaluation

One strength of the authors’ methods is that they bothered to standardise the Airbnb listing counts against the usually resident counts, unlike other studies I have seen (such as Zhang and Chen, 2019) where their Airbnb spatial distributions simply mirror population density maps.

Another strength was the use of an interactive map instead of a static one, as the ability to zoom plus the inclusion of hover text enabled the map to convey more information without visual crowding. I would have appreciated if the listings per 1000 people were given as a percentage in the hover text to make it easier to compare CAUs. Another potential improvement would be to produce a spatiotemporal animation which compares the spatial distribution of Airbnb listings over time.

The table with its basic data summary is easy to interpret and does not require as much statistical knowledge as a regression table. This, combined with the simplistic nature of other elements in the article, makes me suspect that perhaps the authors were aiming for a very general readership over a particularly academic or technical one.

One particularly glaring weakness is that the only ESDA technique included was the Moran’s I value. It seems like a missed opportunity to not have explored the local indicators of spatial association (LISA).

Findings and conclusion evaluation

I am not satisfied with the authors’ interpretation and perceived implications of their findings. Frankly, for an article titled “Disrupting the regional housing market”, it fails to provide any insight into how Airbnb is disrupting the regional housing market.

The studies they list as either examples of “a well-recognized inner-urban phenomenon” or “places with a high level of tourism provision” were all carried out at the city level while this study examined an entire country, which may explain why it yielded a combination of both tendencies (Campbell et al, 2019, p. 142).

Finally, I find it strange that their sole recommendation for future research centres on the effect that Airbnb listings have on councils’ ability to collect taxes, without any mention of the equally important effect that they have on residents’ ability to find (and keep) long-term rentals. This is a disappointing end to an article which has plenty of room for improvement.

References