Trade vs. daily press: the role of news coverage and sentiment in real estate market analysis

ABSTRACT

Each week, thousands of newspaper articles on real estate topics are read by market participants. While the market is comparatively intransparent, readers hope to find valuable information. This raises the question of whether this investment of time pays off and whether different types of newspapers are an equivalent source of information. This paper examines the relationship between news-based coverage of real estate topics respectively news-based market sentiment and total returns of the asset classes of residential, office and retail. Using methods of natural language processing, including word embedding, topic modelling and sentiment analysis, three sentiment indicators for each asset class can be derived from 137,000 articles of two trade and two daily newspapers. Our results suggest that trade newspapers outperform daily newspapers in the prediction of future total returns and that the generated sentiment indicators Granger-cause total returns. Moreover, the results indicate that daily newspapers report more negatively on rising returns in the residential market than the trade press. To the best knowledge of the authors, this is the first study to quantify news coverage and sentiment for the main real estate asset classes through means of textual analysis, and to assess different sentiments in trade and daily press.

KEYWORDS:

• News sentiment
• textual analysis
• real estate asset classes
• GloVe
• seeded LDA

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. LSA aims to reduce matrix dimension in contrast to LDA, which is focused on solving the topic modelling issue.

2. Newspapers‘ circulation in Q4-2021 according to IVW (Informationsgemeinschaft zur Feststellung der Verbreitung von Werbeträgern e. V.).

3. The standard deviation of the article length for the IZ (IM/FAZ/HB) is 238 (431/391/469).

4. The total number of articles is 118,645 (3,601/10,497/3,805) of IZ (IM/FAZ/HB).

5. Temporal disaggregation is referred to as the process of deriving high frequency data from that of low frequency. Furthermore, macroeconomic and real estate-specific control variables have either been tested (e.g. unemployment rate, wages, population growth, building permits, construction cost indices) but did not lead to the improvement of estimation results, or data was not available on asset class level (e.g. construction turnover).

6. Pre-processing of the articles involves the removal of punctuation marks, numbers, non alphabetical and special characters, and stop words. ‘Stop words’ relate to frequently occurring words that have no relevance to the content of a text, such as ‘the’ or ‘and’. For this large text corpus, a general German stop word list is extended by frequent words in the real estate industry context. Furthermore, illustrations, tables, English articles, and editorial shortcuts are also excluded. The data is tokenised for the ensuing tasks. This process divides the text into units (tokens) such as phrases, words, and other meaningful entities. In this case, the text corpus is segmented by words.

7. Matrix factorisation methods decompose large matrices that capture statistical information of a text corpus and generate word representations of low-dimensional latent space in order to reduce computation time, while the concept of context window methods is to predict linguistic patterns as linear relationships between the word vectors based on local context windows and to perform better on word analogy tasks.

8. For instance, $w_{1,2,3}$ represents the third word of the second sentence of the first paragraph.

9. Despite the fact that the standard deviations of NC and NS differ, the differences are minor and do not call for a normalisation of the indices. Also, through the multiplicative linking of the indices, the values would become relatively small and could thus limit the informative power.

10. See Appendix 3 for the descriptive statistics of the generated indices. All results in this paper were estimated in German and have been translated. Expressions consisting of two or more words in English appeared as single words in German.

11. The time series of IM, FAZ and HB have been exponentially smoothed, to reduce their higher variance which are due to the lower circulation and longer articles compared with the IZ. Furthermore, indices for trade and daily newspapers are based on the mean of the underlying newspapers.

12. The remaining 28.06 % is not classified to one of the proposed asset classes.

13. Significant at 1 % level.

14. All correlations are at least significant at 5 % level.

15. See Appendix 3 for the descriptive statistics of the generated indices.

16. See Appendix 3 for the descriptive statistics of the generated indices.

Additional information

Funding

The project was supported through a financial assistance by KPMG Germany.

Notes on contributors

Franziska Plößl

Franziska Plößl is a research assistant and doctoral candidate at the International Real Estate Business School - University of Regensburg. Franziska studied Real Estate Business (MSc) at the University of Regensburg and International Real Estate (MSc) at Florida International University, graduating at the best of her class. She worked as a student assistant for architects and project developers, and also gained practical experiences in real estate finance and valuation in the banking sector at Sparkasse. Franziska has published on various topics, including CREM, nursing homes, megatrends, and sentiment analysis in the real estate industry.

Nino Martin Paulus

Nino Paulus first completed his Bachelor’s degree at the University of Regensburg. This was followed by a Master of Science in Real Estate Management with a focus on real estate investment and financing as well as real estate management and development at the International Real Estate Business School (IRE|BS) and a Master of Science in Real Estate at the University of Hong Kong. Nino gained valuable practical experience in the real estate sector, in particular as economic assistant to the management at DV Immobilien Management in Regensburg. He also worked as a research assistant at the Chair of Financial Accounting and Auditing at the University of Regensburg under the direction of Prof. Dr. Axel Haller. Since April 2021, Nino has been working as a research assistant at the Chair of Real Estate Management under the direction of Prof. Dr. Wolfgang Schäfers at the IRE|BS Institute for Real Estate Management at the University of Regensburg.

Tobias Just

Tobias Just is professor of Real Estate at the University of Regensburg and a Managing Director at the IRE|BS Immobilienakademie, the institute responsible for the executive education programmes of IRE|BS. Tobias studied economics in Hamburg and Uppsala (Sweden). In 1997, he started to work at the University of the Federal Armed Forces Hamburg. His PhD dissertation was awarded the university science prize in 2001. For more than 10 years, Tobias worked at Deutsche Bank Research. He headed the unit’s sector and real estate team and was a member of Deutsche Bank’s group-wide Environmental Steering Committee. In 2006, Tobias was a Research Fellow at the American Institute of Contemporary German Studies at the Johns Hopkins University in Washington DC. Tobias published more than 200 papers in professional and academic journals and books. Until 2021 he was President of the German Society of Property Researchers and is the editor of the ZIÖ — the German Journal of Real Estate Research. In 2018, Tobias became a Fellow of the RICS by nomination.