From listings to all-tenant rents: a probabilistic model
Summary
Focus
We study how to measure rents paid by all tenants using data from online rental listings. Asking rents are the prices of rental units on the market. They are available quickly and in great detail. But they differ from the rents that most tenants pay, which change only slowly.
We build a weekly all-tenant rent index for Switzerland. We start by cleaning millions of listings and removing extreme values. Then we adjust rents based on location and quality, such as canton, size and number of rooms. This gives us detailed information on developments in asking rents by region and apartment size. Next, we use a probabilistic model that links asking rent developments to all-tenant rent developments by modelling how past asking rents enter the rent stock. Finally, we model the effect of Swiss rules that allow changes to rents when the mortgage reference rate changes.
Contribution
We present a new way to measure rents in real time without using tenant surveys. This is important because rents are the largest item in the consumer price index in many countries. Better, faster information on rent trends helps central banks and other policymakers monitor inflation. Current rent data in the consumer price index arrives with a delay and is usually not available by region or apartment type. Our method uses only listing data and official statistics so can be adapted to other countries without detailed rent surveys.
Findings
Our weekly all-tenant rent index is very close to the official rent component in the consumer price index, especially after 2021. The main reason is the gradual transmission of asking rents into the broader rent stock as tenants move. Adjustments linked to the mortgage reference rate also matter, but they play a supporting role. Our index improves rent inflation nowcasting and reveals large differences across regions and apartment sizes. This gives policymakers a timely and detailed view of the rental market.
Abstract
Rents are the largest component of the Consumer Price Index (CPI) in many countries, making accurate and timely measurements of rental price developments essential for inflation monitoring and policy decisions. Market (asking) rent indices are often available in near real-time and with high detail, but differ substantially from the rents paid by the overall tenant population, as typically measured in the CPI. This paper proposes a model to bridge the gap between asking and all-tenant rents. First, using rental-unit listings for Switzerland, we construct timely, granular, and high-frequency indices of asking rents. Second, using a probabilistic model that accounts for the duration of tenants' stays, we estimate all-tenant rents based on historical asking rents. Additionally, we incorporate rent changes during ongoing tenancies. For Switzerland, this corresponds to adjustments permitted under Swiss tenancy law in response to changes in the mortgage reference rate and inflation. This allows us to provide weekly, real-time, and highly disaggregated estimates of all-tenant rents, which are highly correlated with the official quarterly survey-based rental index in the Swiss CPI. Our approach provides a tool for timely rental price monitoring and forecasting that can be adapted for use in other countries.
JEL classification: R21, E31, E37
Keywords: asking rents, rent indices, duration model, inflation
