The Odd One Out: Asset Uniqueness and Price Precision

  • Thies Lindenthal (University of Cambridge)
  • Carolin Schmidt (ZEW, Mannheim)


Based on applied machine learning (ML) techniques this paper suggests that round prices are not purely random events but are linked to liquidity and the uniqueness of the asset. First, using residential transaction data from the UK, we show that the availability of information from comparable sales influences the odds of observing a sale at a round price. Second, we explore ways to play to the strengths of deep neural networks and incorporate computer vision approaches and building-level imagery. Adding information on a building’s vintage and the typology of its direct surroundings to the training data boosts the predictive power of the suggested ML classifiers.

When a house is “the odd one out”, its value will be relatively difficult to establish which implies that sales prices suffer from a relatively low signal-to-noise ratio. Automatic appraisal systems or index estimations could improve their accuracy by incorporating our findings.

Machine Learning, Building Vintage and Property Values

  • Thies Lindenthal (University of Cambridge)
  • Erik B. Johnson (University of Alabama)


This working paper introduces an algorithm that collects pictures of individual buildings from Google Street View, trains a deep convolutional neural network (CNN) to classify residential buildings into architectural styles(vintages) and estimates the impact of vintage on sales prices for the universe of residential housing transactions for the city of Cambridge, UK, between January 1995 and June 2017.

The contributions of the paper are to 1) introduce a general algorithm for capturing building-specific images; 2) illustrate the efficacy of transfer learning using Google’s Inception-v3 model to classify images; and 3) provide a basic hedonic estimate of the impact of architectural style on housing sales prices. Preliminary estimates indicate a price premium of 11.6% for Georgian, 8% for Early Victorian, 22% for Late Victorian/Edwardian, 6.8% for Interwar, 6.1% for Contemporary, and 7.8% for Faux Victorian relative to the Postwar architectural style. We are currently extending the data framework to enable data collection and analysis for the entire UK.

Some aspects of this work has been cast into a mobile app:

The development of APIs to utilise machine learning models in various applications, substantial improvements to the process of collecting training data, data management and the presentation of any results has been greatly supported by a Centre for Digital Built Britain “Mini Project”.

Object Detection on Google Street View

  • Thies Lindenthal (University of Cambridge)
  • Mike Langen (Maastricht University)


We propose a new method to collect structural property characteristics, using image recognition and machine learning. Based on a training dataset of 10,000 Google Street View images, our algorithm is able to detect and extract property characteristics, such as the number of floors, building style, windows, garden, etc. By explicitly focusing on extracting property-level characteristics without land registry information as a requirement, our algorithm easier to implement and more precise than previous algorithms when it comes to property applications, making it an attractive choice for urban studies.

Given the ubiquity of Google Street View, our approach is transferable into many markets, allowing researchers to enrich existing datasets with hedonic information or collect new data in a cheap and efficient manner.

Public Transport, Noise Complaints, and Housing: Evidence from Sentiment Analysis in Singapore

  • Yi Fan (National University of Singapore)
  • Ho Pin Teo (National University of Singapore)
  • Wayne Xinwei Wan (University of Cambridge)


This paper investigates the effect of a new bus route on subjective noise complaints of residents and the influence of noise on housing prices. To overcome the challenge of mapping noise data with subjective emotion, we use a novel data source—text-based noise complaint records from residents in a town in Singapore—and apply natural language processing (NLP) tools to conduct sentiment analysis. To address the endogeneity concern regarding the bus route, we use a hypothetical least-cost path as an instrument for the existing bus route. We find that living closer to the bus route for every 100 meters increases noise complaints by around 10 percentage points, and the effect is more severe on medium floor levels (5th- 8th floors) and near bus stops (within 100 meters). We further link noise with housing prices and discover a price reduction of 3% with a 1-scale-point increase in noise complaints. This implies that bus noise o sets 17.8% of the benefit from convenience, which sheds light on the importance of noise insulation in transit-oriented developments.