How useful is listings data for research?

Jens Kolbe; Rainer Schulz; Martin Wersing; Axel Werwatz

How useful is listings data for research?

Jens Kolbe^* (Corresponding Author), Rainer Schulz, Martin Wersing, Axel Werwatz^* (Corresponding Author)

^*Corresponding author for this work

Research output: Working paper

Abstract

Compared to other asset classes, information on transactions of residen-tial real estate is scarce and available only with delay. Listing informa-tion from web-platforms is abundant and timely. Is listings data usefulfor research? We examine this question and nd that distributions ofask and sale prices di er signi cantly, both because of characteristicscomposition and implicit pricing. Estimates of the average willingnessto pay from ask data can be widely o when compared with estimatesfrom sale data. Ask data is also not useful to predict prices of individualhouses and su er from large error variances. Quality-controlled ask andsale price indices show similar trends and we nd that an ask price indexcan be used for nowcasting. Overall, our analysis shows that ask datahas limited potential for research, and is no substitute for sale data.

Original language	English
Publisher	Humboldt-Universität zu Berlin
Publication status	Published - 2020

Publication series

Name	FORLand Working Papers
Publisher	DFG Research Unit 2569 FORLand, Humboldt-Universität zu Berlin
No.	2020-19

Keywords

hedonic modelling
nowcasting
price prediction
stochastic dominance

Access to Document

https://edoc.hu-berlin.de/bitstream/handle/18452/21783/FORLand-2020-19.pdf?sequence=1&isAllowed=yLicence: CC BY-NC-ND

Cite this

@techreport{99b9ff78bc584e57a413e47151f0366f,

title = "How useful is listings data for research?",

abstract = "Compared to other asset classes, information on transactions of residen-tial real estate is scarce and available only with delay. Listing informa-tion from web-platforms is abundant and timely. Is listings data usefulfor research? We examine this question and nd that distributions ofask and sale prices di er signi cantly, both because of characteristicscomposition and implicit pricing. Estimates of the average willingnessto pay from ask data can be widely o when compared with estimatesfrom sale data. Ask data is also not useful to predict prices of individualhouses and su er from large error variances. Quality-controlled ask andsale price indices show similar trends and we nd that an ask price indexcan be used for nowcasting. Overall, our analysis shows that ask datahas limited potential for research, and is no substitute for sale data.",

keywords = "hedonic modelling, nowcasting, price prediction, stochastic dominance",

author = "Jens Kolbe and Rainer Schulz and Martin Wersing and Axel Werwatz",

year = "2020",

language = "English",

series = "FORLand Working Papers",

publisher = "Humboldt-Universit{\"a}t zu Berlin",

number = "2020-19",