The question behind the headlines
Online platforms that let people rent out homes or rooms for short stays grew quickly in many cities. Alongside that growth came a recurring claim: that short-term rentals push up housing costs, reduce the number of homes available to long-term residents, or change the character of a neighbourhood. Others argue the opposite or something more modest, noting that hosts earn income, visitors spread spending beyond hotel districts, and housing markets respond to many forces at once.
The disagreement is partly about values and partly about evidence. This article focuses on the evidence: how researchers try to connect rentals to local change, and why the connection is harder to establish than it first appears.
What gets measured
Studies tend to use a handful of measures.
- Listings. The number of properties offered for short stays in an area, sometimes broken down by whether the whole home is offered or just a room, and whether the host has multiple properties.
- Activity. How often listings are actually booked, since a listing that is rarely occupied differs from one rented most nights.
- Housing outcomes. Rents, sale prices, vacancy and the number of long-term tenancies.
- Neighbourhood characteristics. Changes in shops, population, noise complaints or resident opinion.
Each has problems. Listing counts do not show whether the property would otherwise have been rented to a long-term tenant, used by its owner or left empty. A listing may be active for a few weeks a year, which is quite different from a full-time commercial rental.
Where the data comes from
Researchers often cannot rely on official records alone. Some places require registration of short-term rentals, but compliance varies, and many hosts may not register. As a result, studies have commonly used data gathered from the platforms’ public pages, sometimes called scraped data, or data supplied by commercial analytics firms.
This is platform data, with the usual blind spots. It captures only the platforms covered. It may miss listings on other sites or arranged privately. Calendar data can show a date as unavailable without revealing whether a guest booked it or the host simply blocked it. Changes in how the platform displays information can also alter what researchers see. Broader issues around platform data and its blind spots apply directly.
Why correlation does not settle the matter
A simple finding might be: neighbourhoods with more short-term rentals also saw faster rent increases. That is a correlation, and it leaves several explanations open.
- Confounding variables. Attractive, central, well-connected areas draw both visitors and new residents. Demand for housing there may rise regardless of rentals. Confounding variables of this kind can create an association without any causal link.
- Reverse causation. Hosts may choose to list properties where visitor demand and prices are already strong, so the arrow may run from neighbourhood popularity to rentals rather than the other way. The idea of reverse causation is worth keeping in mind.
- Timing. Rentals, tourism and housing prices may all respond to the same wider change, such as an economic upswing or new transport links.
- Selection. Places where platforms first took hold may differ from those where they did not, which complicates comparisons of the two groups.
This is why a careful reader treats statements like “rentals caused rents to rise” as claims that need more than a comparison of two maps.
Stronger research designs
Researchers use various strategies to get closer to cause and effect, though none is perfect.
Comparing before and after. Looking at an area before and after rentals grew can help, but other changes in the same period might explain the pattern.
Comparison with similar areas. Matching neighbourhoods that look alike but differ in rental activity can reduce some bias, although unmeasured differences may remain.
Using rule changes. When a city introduces a limit, registration requirement or ban, the change can work as a natural experiment, since it affects some areas or property types and not others at a particular time. Comparing outcomes before and after, against places without the rule, can give clearer evidence. Even then, people and businesses may adapt, and the rule may coincide with other changes.
Using indirect instruments. Some studies use factors that influence rental growth without directly affecting housing prices, such as the timing of platform launches or the number of tourist attractions. These designs depend on assumptions that are debated.
Results from such studies often depend on the city, the time period and the type of rental, so findings in one place may not carry over to another. Reasonable researchers can reach different conclusions about size and even direction of effects, and a single study is rarely decisive.
Local change is more than prices
Neighbourhood change includes less measurable things: who lives nearby, whether neighbours know each other, how busy streets feel and which businesses stay. These are harder to capture. Resident surveys help but reflect who answers and how questions are asked. The wider conversation on gentrification narratives shows how stories about neighbourhoods can run ahead of data, with an area described as transformed when evidence is thin.
There is also a difference between aggregate effects and individual experience. A modest citywide effect could still be sharply felt in a particular building or street, and a visible concentration of rentals can shape perceptions even when its overall impact on rents is small.
Reading claims about rentals
A few questions help.
- Does the claim describe correlation or a tested cause?
- Is the data from official records, platform pages or a commercial source, and what might it miss?
- Is the comparison between similar places and similar times?
- Are other explanations, such as general housing demand, discussed?
- Does the finding relate to the same city and type of rental as the one being discussed?
Short-term rentals may well affect some neighbourhoods in some ways. The honest summary tends to be that the effects are plausible, vary by place, and are harder to isolate than public debate usually suggests.