[Paper Review] What do online listings tell us about the housing market?
The paper analyzes Italian online housing ads to quantify duplicates, corrects them with machine learning, and demonstrates how listings can inform demand, supply, liquidity, and price dynamics.
Traditional data sources for the analysis of housing markets show several limitations, that recently started to be overcome using data coming from housing sales advertisements (ads) websites. In this paper, using a large dataset of ads in Italy, we provide the first comprehensive analysis of the problems and potential of these data. The main problem is that multiple ads ("duplicates") can correspond to the same housing unit. We show that this issue is mainly caused by sellers' attempt to increase visibility of their listings. Duplicates lead to misrepresentation of the volume and composition of housing supply, but this bias can be corrected by identifying duplicates with machine learning tools. We then focus on the potential of these data. We show that the timeliness, granularity, and online nature of these data allow monitoring of housing demand, supply and liquidity, and that the (asking) prices posted on the website can be more informative than transaction prices.
Motivation & Objective
- Evaluate limitations of online housing listings as data sources for the housing market.
- Quantify the impact of duplicate ads on measurements of supply, demand, and prices.
- Develop a machine learning pipeline to identify and merge duplicate ads into true listings.
- Validate the deduplicated dataset against official statistics and local price measures.
- Show how listing data can be used to monitor demand, supply, liquidity, and price dynamics.
Proposed method
- Use Immobiliare.it weekly ad snapshots (2016-2018) to build a dataset of ~1.4 million ads and ~950k housing units.
- Apply text analytics (doc2vec) and structured feature distance measures to assess similarity between ads.
- Implement C5.0 decision trees to classify when two ads refer to the same dwelling, with separate models for same vs. different users.
- Construct ad clusters to aggregate duplicates into single listings and create a deduplicated final dataset.
- Validate deduplicated data by comparing delistings, sales, and price measures against official sources (OMI, Italian Housing Market Survey).
- Analyze the determinants and dynamics of duplicates and evaluate implications for measurement error and policy usefulness.
Experimental results
Research questions
- RQ1How large is the measurement error introduced by duplicate ads in online housing listings?
- RQ2Can machine-learning deduplication reliably recover true housing units from ad-level data?
- RQ3What can online listings tell us about housing demand, supply, liquidity, and price dynamics compared with traditional sources?
Key findings
- Duplicates cause substantial but uneven measurement error; effects are smaller at higher aggregation but larger for local markets.
- About 77% of dwellings have a single associated ad, 13% have two duplicates, and 10% have more than two; duplicates account for a sizable share of ads for some homes.
- New deduplicated listings align with official statistics in terms of delistings, sales, and price levels; deduplication reduces bias, especially at local levels.
- Sellers/brokers post multiple ads more when asking prices are high or demand is weak, indicating non-random duplication tied to conditions and incentives.
- Online listing data can proxy demand via webpage visits, predict time-on-market and price revisions, and provide timely signals for market tightness and price cycles when properly deduplicated.
Read this paper
Open the full text in the Nubint reader and ask about any passage as you read.
Read this paper on NubintThis review was generated by AI from the original paper. Please check the paper before citing it.