Leo Saenger

How do the seasonal patterns of tourism shape Venice? How do different communities of tourists drive changes in the built environment? And what does all this mean for the city’s post-COVID future?

Abstract

Even before COVID-19, Venice was fighting another epidemic: touristification. Since WWII, the city has replaced over 100,000 inhabitants with an increasing number of AirBnBs, souvenir shops, and high-end restaurants. COVID-19 briefly took nearly all tourists out of Venice, but as they return, where will they locate? Moreover, how can COVID occasion a shift in how tourism occurs in the city center – and what might that shift entail? I combine two novel datasets, using reviews and location data from Tripadvisor and booking and customer data from AirBnB to show changes in tourist activity over time and the post-COVID transformation of Venice. While Italian tourism made up some of the initial post-COVID shift, changes in the distribution of tourist amenities – such as the closing of many AirBnBs in the city center – continue to reduce the importance of the historic center relative to the outer areas of the islands and highlight already-ongoing changes in Venice’s previously-neglected areas. I argue this juncture represents a unique opportunity to “de-center” tourism in Venice and make life more sustainable for locals without removing much-needed tourist revenues for small businesses in each community. 

Key Questions

Before 2020, some 30 million tourists visited Venice each year. From massive crowds to cruise ships towering over the skyline, the sheer congestion during peak months can make for a disappointing (and expensive) visit. Still, tourists worldwide take long trips to visit Venice’s unique architecture and get a chance to ride down the canals. As tourism has boomed, the local population has not — Venice’s population is now only a fraction of its historic totals, a paltry 50,000 compared to 170,000 just after WWII

In our main investigation, we developed a granular index of tourism to demonstrate that the built environment and the design of the city has contributed to the over-concentration of tourism within specific districts. Still, understanding how to alleviate this problem is challenging — as more tourists begin to return to the city, where will they locate? Moreover, how can COVID open up a potential shift in how tourism occurs in the city center — and what might that shift entail?

Data

TripAdvisor reviews and profile data was gathered from 273,000 reviews and 505 restaurants (out of some 1,600 on TripAdvisor) that make up a broad range of ethnicities and languages: in our main dataset, we have over 2,075 unique locations from over 85% of profiles from TripAdvisor. Location data was matched using each user’s supplied location and OpenStreetMap’s Geocoding API, and ambiguous locations (e.g. “Portland”, which could refer to Oregon or Maine) were removed from the sample. 

Restaurants within the sample include those that appeared in a search for all restaurants in Venice in February 2020, such that we do not over-sample restaurants that experienced extra success after March (e.g. those with space for outdoor dining). Review data was somewhat incomplete prior to 2014 (TripAdvisor doesn’t show 100% of reviews after a certain number of pages of reviews are shown), so reviews before that point are dropped. 

Data on AirBnB and other short-term rentals by owner was retrieved from AirDNA, a service which scrapes the AirBnB and other listings within individual cities and collects data on their occupancy rates, number of beds/reviews, and other information. AirDNA also collects data on AirBnBs which are no longer “active”, that is, they are not listed publicly on the website. I consider these AirBnBs to be “deactivated”, since this means that they were not used at all during that month (and thus could be used for other purposes).  

Flights of Fancy: Tourist Origins

Aside from seasonal shifts in previous years, the most striking shift here is toward the end of our data, where a complete collapse in visitors around March 2020 brings tourism levels to as low as 3% of the previous year’s values (during the lockdown), and at best only recovers to approximately 20% of the previous year’s value by late July and early August. Particularly striking is the complete lack of tourist information from Asia after March, and almost complete drop off of tourism from South America and North Africa. 

The map below shows origin points from visitors to Venice over time, with points sized by the average number of visitors from that area of the world. Clusters change in size as you zoom in and out of the map. Line thickness here represents the number of data points from that specific location (non aggregated) and can be clicked to view more information. Use the time-slider at the bottom of the map to explore changes driven by seasonality as well as by the March 2020 tourism collapse. 

Of particular interest here is also the resurgence in Italian reviews, alongside German, Dutch, and Swiss tourists, which seem to make up most of the “recovery” following July. Although just a relative shift (absolute tourism has decreased substantially), this substantiates the idea that domestic tourism, for Italians, is starting to replace going abroad, especially as dynamic travel restrictions prove difficult. Surprisingly though, the shift toward Europe does not strictly privilege those closer to Venice — a high number of tourists from the UK and Northern Europe suggest more is at play here. 

Regardless, these visualizations show a turn inwards and a shift in the demographics of the tourists that usually make up the majority of Venice’s residents — and in the next visualization, we see how that shift in composition is reflected in the relative popularity of areas to visit within Venice. 

Room for Improvement: AirBnB and Tourist Destinations 

AirBnB has dramatically increased in popularity over the past five years, from just 10,000 reservations each month in the historic center in 2015’s peak season to nearly 60,000 each month in the peak of 2019 — more than the entire population of the historic center (53,000)! Although the pandemic caused a steep dip in net reservations, it’s worth noting that the total number of reservations within the Historic Center recovered to off-peak levels in 2019 by the end of 2020, approximately reaching 2016 peak season volumes. An initial read seems to suggest that those on the mainland and within the outer islands experienced a more tepid recovery, though as we’ll see the story is a bit more complex than the top-line numbers might suggest.

Part of the story comes from a remarkable consolidation of revenues among a few “superhosts”. It’s no wonder residents are becoming increasingly hostile toward AirBnB: in 2015, the top 5% of hosts (by total revenue) took home 27.4% of total revenues, while in 2020 the top 5% of hosts – just 67 people – took home 44% of all revenue earned in Venice. So too among the very top: in 2015, the top 1% made up just 6.6% of total revenues, but in 2019 that very top – just eight hosts – took home nearly 27% of the total revenue. The type of small, part-time AirBnB host the company champions took an even more severe hit in 2020 – yet our records show that the median earnings among the top 5% of hosts stayed roughly consistent, increasing the consolidation of the market within the hands of the top players further.

This is not to say that those heavily invested in AirBnBs did not take heavy losses — many did — but the ones whose properties survived managed to keep revenues remarkably steady. For example, while the median host among the top 5% (by total revenue) took home $156,855 per AirBnB in 2020 YTD, the median host in the bottom 95% of hosts earned just a fraction of that, just $5,229 so far this year. 

This collapse among the majority of smaller AirBnBs has contributed to a sharp uptick in AirBnBs closing for good (potentially to be converted to longer term housing). The net number of AirBnBs that de-listed entirely in the historic center reached 500 in August, with similar changes (greater than 100 in the mainland and 50 in the outer islands) elsewhere. These new apartments open up the potential for more residents to move back to Venice and suggest coming affordability increases, especially given how low tourism continues to persist.

Within the Historic Center, recovery has been very uneven, with reservation numbers that previously moved in lockstep significantly diverging with a few less popular neighborhoods experiencing dramatic changes in tourist volumes.

Heterogeneity in Nationalities Across Venice

The following maps show differences in dominant nationality of visitors among different regions of Venice. Aside from showing areas where Italians dominate, it also shows clearly the different “seasons” of tourists that come in during a typical year, with high regularity (especially within restaurants). AirBnB data may be more accurate in this instance, since we have more reliable information on user locations and a potentially wider user base (we know it is only tourists, not locals, and there’s less selection into reviewing — everyone is counted in the AirDNA database regardless of if they left a review or not). Still, the two charts are remarkably consistent in their central message — both in origins and in the post-COVID recovery. 

The AirBnB map merits further consideration, especially since it more clearly shows activity on the outer islands. We can see that not only does most lodging off of the historic center return to its pre-pandemic levels, some areas even see a relative increase in tourist rates — relative to a strict decrease within the more previously popular areas. This lends further credence to the “substitution toward the less popular areas” effect. It’s notable, however, that no such obvious trend is observable within our restaurant data, which suggests that tourist sites might have yet to “catch up” with those on the mainland.

Conclusions and Next Steps

A wealth of micro-data suggests that Venice is moving toward a more decentralized city, with tourists choosing to explore previously less-trafficked areas on the outer islands and former AirBnBs opening up for new purposes within the city center. Still, as we found in our previous investigation, a lack of tourist attractions and transportation linkages within these outer areas predicts that they won’t accumulate the same tourist volumes as the city center. This might be a feature, not a bug. Declining tourist volumes and new rooms for residents in the most densely trafficked areas might make them more appealing and livable for Venetians and tourists alike. On slightly more popular islands, residents could get an economic boost while heeding the lessons of the historic centre and avoiding over-touristification. 

Regardless of what comes next, a few facts are clear: restaurants will have to alter their business plans, and AirBnBs – the vast majority of which are run by hosts with low revenues and few other properties – will have to turn over their keys. Doubtless, many souvenir vendors, retail shops, and other tourist amenities will have to significantly pivot in a new direction as well. Loss of livelihood is never something to be celebrated. But when all is said and done, the Venice that emerges from the pandemic will be reshaped in significant, potentially lasting ways. It will be easy for residents to simply return back to the past, but a second opportunity exists for the taking. By outlining its contours, we can open up the possibility for a different future, and a potentially reshaped Venice.

Limitations

We only observe data from tourists, who greatly outnumber locals both in Venice and on the online sites that were scraped for this analysis. For this reason, we likely undercount amenities within the less trafficked areas of Venice — they’re less likely to be searched on Google, for instance, and thus more likely to be omitted from the top results within our searches. For this reason, we should be relatively cautious about interpreting a real lack of amenities from low reviews, or vice versa, on the outer reaches – since we had less data available there, it wasn’t possible to fully validate if the same relationships that hold on the mainland also apply.