That pandemic shock in the online vacation rental market failed to happen; until it did!

Photo by Mariano Diaz on Unsplash

It was the first week of March. The once seemly distant spreading of coronavirus in China seemed closer, while images from the most recent outbreak in Italy started to pour into our televisions.

Monitors awash in red, desperate people on a buying spree, everyone locking themselves at home, seemed like the end of the world.

Meanwhile, in Rio de Janeiro, even amid lockdowns, movement restrictions, and social distancing, Airbnb hosts kept their listings up and running at the same price as before. And for April, May, June, July. In October market nosedived, the supply almost halved, and the listing prices were down by 13%.

With so many restrictions, anyone expected the vacation rental market to sour rapidly. Nonetheless, hosts were unwilling to de-list or reduce prices.

Considering this surprising behavior, I analyzed the price and supply before and during the pandemic. I investigated whether a price difference existed when segregating the listing by their titles in English or Portuguese. Finally, I checked which features matter for price determination.

The Data

We have monthly data from Airbnb by Inside Airbnb. For Rio de Janeiro, it skips data for some months. The time series starts in April 2018 and ends in December 2020.

The Pandemic

The figure below represents the median prices in Brazilian Real (BRL) or the supply of rooms along with the date of scraping. The first row contains all observations. In the second row, split between those English titled from Portuguese.

Median prices in Brazilian Real and supply count. Figure by the author.

We divide the pandemic into two starkly different periods: March to July and October to December.

In the first period, the expected shock in prices and supply never materializes. A reduction in price due to seasonal effect around July never emerges.

Nonetheless, we observe a price hike for English titled listings and an increase of supply for Portuguese.

For this first period, we can hypothesize that:

1. Hosts kept or increased prices due to the devaluation of the Brazilian Real (or inflation).
2. Hosts increased the supply of Portuguese titled rooms to domestic travelers due to international movement restrictions.
3. Hosts were afraid of losing income and tried their chances.
4. Hosts listing in English react more promptly to devaluations.

In the second period, the market bottomed, yet in December, the prices rebounded. Supply slumped 43% from 35.000 to 20.000. Price initially shrunk from 300 BRL to 260 BRL (13% decrease), although the price jumped in December to 350 BRL (34% increase from bottom).

Adding other hypotheses:

1. Price and supply decreased due to low demand.
2. A restriction relaxation increased demand amid a short supply, driving prices up in December.
3. Hosts tried to keep supply and prices during the first period. Yet, the lack of demand took its toll. Prices and supply collapsed after a delay.

If you can read this, you will be probably paying more for a room in Rio

If you ever come to Rio, rent a room in Airbnb, and know no Portuguese. Expect to pay an 80 BRL premium daily for the convenience [1]. And that is five fewer caipirinhas for you.

The median price difference between listings in English and Portuguese. Figure by the author.

We suppose hosts fluent in English cater to more affluent international visitors, and given the low English literacy of Brazilians, they can charge more.

English listing prices are stable throughout the year. We guess that international touristic demand might be steadier than domestic.

What makes a price?

As the adage says: location is everything, or almost. Latitude and Longitude correlate negatively with price, meaning that south-western prices are higher.

We expected the southern trend. The famous beaches of Copacabana and Ipanema locate in the Zona Sul region, along with other touristic hotspots. However, they are a bit eastern.

Spatial distribution of prices. Figure by the author.

It stands out that the Barra da Tijuca region — which is south, but western, and also rich but not that touristic — is pushing these higher prices to the west.

Another relevant correlation is that the more reviews a room has, the lower the price. Probably unsatisfied renters are more likely to give negative reviews. A sentiment analysis might confirm this hypothesis.

Listing with higher minimum nights tends to impact prices positively.

In fitting a model, the most relevant features are geographical (latitude and longitude), room type (apartment, room, etc.), and reviews [2].

Conclusion

The pandemic shock in the online vacation rental market in Rio de Janeiro occurred later than expected. Hosts tried to keep their room online at the same price but ended giving in. However, in December, the prices shot up again.

We observed that prices in English titled rooms are systematically higher than those in Portuguese [1]. Also, those English prices are less prone to seasonal effects.

Geography, room types, and reviews influence prices the most [2].

The findings are mostly observational and proving some of our hypotheses are beyond the scope of this article. Instead, they can be start points for future works.

[1] We tested the difference in prices between Portuguese titled listing from those in English using Anderson–Darling and Welch’s tests.
[2] We fit the data into a Random Forest model.
[3] The complete analysis is at my GitHub.