Loading...

Using Google Trends in Real Estate Research

Definition, State of the Art, Strengths and Weaknesses, Threats and Opportunities for Real Estate Research based on Google Trends

Bachelor Thesis 2014 33 Pages

Business economics - Offline Marketing and Online Marketing

Excerpt

Table of Contents

List of Tables

List of Abbreviations

I. Introduction

II. Main Part
II. 1 Definition and History of Google Trends
II. 2 Summary of the Current State of Real Estate Research Based on Google Trends
II. 3 Discussion of the Usage of Google Trends for Real Estate Purposes
II. 3. 1 Weaknesses of Google Trends Based Real Estate Research
II. 3. 2 Strenghts of Google Trends Based Real Estate Research
II. 4 Discussion of Future Threats and Opportunities for Real Estate Research Based on Google Trends

III. Résumé

References

List of Tables

Table 1: “Predicting the Present with Google Trends”

Table 2: “Forecasting Housing Prices with Google Econometrics”

Table 3: “The Future of Prediction: How Google Searches Foreshadow Housing Prices and Sales”

Table 4: “´Geco` and its Potential for Real Estate Research: Evidence from the U.S. Housing Market”

Table 5: “Forecasting Residential Real Estate Price Changes from Online Search Activity”

Table 6: “´GECO’s Weather Forecast` for the U.K. Housing Market: To What Extent Can We Rely on Google ECOnometrics?”

Table 7: “Fear and Loathing in the Housing Market: Evidence from Search Query Data”

Table 8: “Detecting Mortgage Delinquencies with Google Trends”

List of Abbreviations

illustration not visible in this excerpt

I. Introduction

“ The Internet contains an enormous amount of information which, to our knowledge, classical econometrics has yet to appropriately tap into. Such information comes timely on a continual basis. It is particularly welcome at times of an economic crisis where the traditional flow of information is too slow to provide a proper basis for sound economic decisions. Not only has traditional (and typically official) statistical data a slow publi- cation scheme, these data also do not reflect well the structural changes in the econ- omy. ” 1

A very suitable example for this kind of failure of a traditional survey was the report of the American National Bureau of Economic Research in September 2010, which de- clared that the Great Recession had ended 15 months ago2. It is anything but surprising, that decision makers as well as researchers are constantly trying to overcome such struc- tural problems. As the internet is used by considerably more people than any survey could possibly capture, it did not take long for research to discover its enormous poten- tial. To illustrate the internet´s capability, one just needs to take a look at processed data volumes. In 2009, the social network Facebook handled more than 25 terabytes of data a day, which by the time was already more than what the largest survey in the world, the U.S. Census, produced in a decade3. In 2012 the daily amount of data processed by Facebook, had already gone up to 500 terabytes and has been constantly rising ever since4. However, prior to the launch of the internet service Google Trends in 2006, re- search based on internet data was still rather difficult, as most of the relevant data was not yet publicly accessible5. With the release of Google Trends, researchers were given the opportunity to examine the correlation of search frequency of particular keywords and i.a. the current economic conditions, as Google Trends graphically shows the popu- larity of a particular search term compared to the total number of searches on Google6. It did not take long until the real estate industry became aware of the service´s huge potential and began to use it for real estate purposes. This treatise addresses the ques- tion, whether Google Trends is a suitable instrument for supporting real estate research.

In the course of this paper, the current state of real estate research based on Google Trends will be investigated. Then a SWOT analysis of the potential of Google Trends for real estate purposes will be established in the subsequent paragraphs, followed by an overall conclusion. But first of all, Google Trends will be introduced in more detail in addition to a quick overview on the different research areas on the grounds of the ser- vice.

II. 1 Definition and History of Google Trends

Google Trends is a public web facility of Google Inc. that shows how often a particular search term is entered into the search engine Google, relative to the total number of searches done on Google since 20047. The result is presented in a graph that can option- ally be downloaded as a CSV-file, which contains more detailed information. Addition- ally, it is possible to compare up to five different search- terms within one single graph. The horizontal axis of this graph is the “Interest over time”, that begins in 2004 and shows the search query volume aggregated in monthly intervals. In the optional CSV- file, the weekly aggregated data can be viewed, whereby the latest data is released every Sunday8.

The vertical axis of the graph does not represent absolute search volume numbers, be- cause the data is being normalized and presented on a scale from 0-100. This modified search query data emerges from dividing the amount of requests for a particular search term by the entire global search volume in the relevant time span. The outcome will then be scaled by equating the maximum value of this quotient to date to 100. The rest of the data on the graph is being scaled accordingly9. In case Google does not have enough data for the particular search term in some periods, a 0 is shown on the graph10. If there is not enough search query data for a search term in general, the service simply displays “Not enough search volume to show graphs”. Below the main graph “Regional interest” and “Related searches” for the typed in search term can be seen in two addi- tional graphics. Moreover, it is possible to refine the main graph by country, year, cate- gory, such as real estate and type of web search, which branches out in web, image or news search, Google Shopping and YouTube search11.

Lastly, the way a search term is typed into Google Trends is quite substantial for the final outcome. According to Google´s support page, there are four possibilities for en- tering a keyword, for example the search term real estate. The first alternative is to key in real estate. The final graph will show all Google search queries randomly related to real and estate. The second option is to enter “ real estate ” in, the result will then in- clude both parts of the specific search term in the right order and is the most frequently used alternative in research. Thirdly, it is possible to type in real+estate, which includes all search terms that either contain the words real or estate. Finally, it is also possible to enter real-estate; the relevant outcome will comprise search requests including the word real, but excluding the word estate12.

After having examined the most important technological aspects of Google Trends, a closer look at the history of the service will be taken. Google Trends was originally launched in May 200613, but in August 2008 Google released an additional web facility named Google Insights for Search, often abbreviated as I4S, a more sophisticated and advanced service for accessing search query data14. In September 2012, Google merged Google Insights for Search into Google Trends, to improve the mobile use of Google Trends as well as to give it, according to Google “a clean new interface to give you a clearer view of what’s on the world’s mind”15. As the merger took place less than two years ago from this writing, it grows apparent that some of the literature dealt with in this paper, used either Google Insights for Search data or the Google Trends version prior to the merger. The following paragraph gives a short overview of the most impor- tant research that has been done on the basis of Google Trends and I4S data so far, to demonstrate how versatile the service can be used for research purposes:

2009: Ginsberg et al. published the first academic work using Google´s search query data. By observing search requests related to the most frequent influenza symp- toms, they managed to identify regional illness hot spots in the U.S. This ap- proach resulted in the invention of Google Flu Trends 2, a service dedicated to the global tracking of epidemics. Their paper was fundamental and revolutionary to this new field of academic research16.

2009: Nikolaos Askitas of the Institute for the Study of Labor and Klaus Zimmermann of the University of Bonn used Google Insights for Search to predict unemploy- ment in Germany with four German keywords associated with unemployment. They also proofed that the chosen keywords correlated well with the unemploy- ment rates that were taken from monthly German survey data17.

2009: Torsten Schmidt and Simeon Vosen of the University of Ruhr found that Google search frequency forecasts consumer spending better than traditional consumer surveys, by comparing Google-based to survey-based methods of buying behaviour predictions18.

2011: In the Journal of Finance, Zhi Da, Joseph Engelberg and Pengjie Gao revealed the relevance of Google data as a direct and timely measure of investors’ attention. Furthermore, their paper outlined the general usefulness of web search data in fi- nancial applications19.

2012: Francisca Beer of the California State University, Fabrice Hervé and Mohamed Zouaoui of the University of Burgundy, introduced a new measure of French in- vestor sentiment to examine its impact on the stock market, based on the search volume of Google Trends data. They affirmed that investor sentiment supports a prediction of short- term market returns20.

2013: Pedro Latoeiro, Sofia B. Ramos and Helena Veiga of the University Carlos III of Madrid analyzed whether web search queries predict stock market activity in a sample of the largest European stocks. They claimed, amongst other things, that an increase in web searches for stocks on Google is followed by a temporary in- crease in volatility and volume21.

II. 2 Summary of the Current State of Real Estate Research Based on Google Trends

Subsequent to the overview on the various fields of research based on Google Trends, a closer look at the literature on real estate related topics resting upon Google Trends or I4S data will be taken. After a brief summary of each paper, further details will be given in the table below the corresponding paragraph.

Among the first ones to make use of Google Trends to examine its potential for predicting new house sales amongst other variables were Choi and Varian, two economists, who work for Google. They applied seasonal autoregressive models to predict the current home sales and prices in the U.S. Moreover, they found that models including relevant Google Trends data outperform models that omit these factors and the respective gain even proved to be quite substantial. Furthermore, they claimed that the search term “Real Estate Agencies” performs best to predict contemporaneous house sales - out of the 6 different search terms they tested- whereas an increase in search requests for “Rental Listings & Referrals” indicates decreasing home sales22.

Table 1: “Predicting the Present with Google Trends”23

illustration not visible in this excerpt

Kulkarni, Haynes, Stough and Paelinck, most of them employees of the George Mason University in Fairfax, U.S., also focused on forecasting house prices with Google I4S data. Unlike the other authors, they used search term categories instead of separate search terms for their investigation. Nevertheless, they successfully managed to forecast and update the Case-Shiller Indices, which are defined as “the leading measures of U.S. residential real estate prices”24. Using Granger causality regressions, the researchers found the Google search terms to predict housing prices quite well, even though the reverse causality could not be proven. Employing their 5 different search term catego- ries, consisting of various combinations of real estate-related search terms, the city level Google search index could be used to forecast the House Pricing Index of cities. Fur- thermore, they succeeded to predict the national level monthly Case-Shiller Index with the national level Google search index25.

Table 2: “Forecasting Housing Prices with Google Econometrics”26

illustration not visible in this excerpt

In the same year, Wu and Brynjolfsson added an important piece of work to the field of real estate related research by means of Google Trends. Similar to their predecessors, they also examined the relationship between five real estate related search terms and the future house price index and future home sales in the U.S. Using a simple seasonal auto- regressive model, they found positive correlations for both, nevertheless the correlation between search queries and the future house price index did not prove to be as strong as the correlation for home sales. As an explanation, the authors pointed to the ambiguity of search requests, which can come from either the demand or the supply side and hence may not have a clear influence on future home prices. Another implication the authors mentioned, is that search frequencies may actually be more effective for predicting the future than the present in this specific field of research. They explained this impact by the fact that a house buying process often takes months to complete27.

Table 3: “The Future of Prediction: How Google Searches Foreshadow Housing Prices and Sales” 28

illustration not visible in this excerpt

Two years after Wu´s and Brynjolfsson´s study, Käsbauer, Schäfers and Hohenstatt dedicated a paper to the analysis of consequences of Google data for house prices and housing transaction volume in the U.S. Their VAR model lead to the result, that Google I4S data predicts transaction volume quite well, and similar to Choi and Varian they also found “Real Estate Agency” to be the best predictor of transaction volume. Accord- ing to the authors, that is the case, because the aggregation level of the search term is not as vague as “Real Estate”, but not too precise all the same. The prediction of hous- ing prices on the grounds of Google query data could also be asserted, but due to occur- ring ambiguous effects, the found correlations between search frequency and housing prices were not as strong as for house sales, which confirm Wu´s and Brynjolfsson´s previous findings. However, the authors regarded the search term “Home Financing” to provide interesting insights into the relevant financial decisions. Besides, an important result of the authors´ Granger causality tests and OLS regressions was that housing market dynamics have an impact on Google Trends data, which then again influences real-world data itself. This interrelation appears, because internet users reveal their in- tentions to the search engine while browsing, but their intentions are driven by the real world and have an impact on it at the same time. In other words, people interested in buying a home are influenced by price-volume conditions. Consequently, their buying behaviour is dependent on and also a factor of influence on market occurrences. An- other goal of the study was to use Google Trends to overcome some of the problems of traditional surveys, which was successfully accomplished by their demonstration of the ability of real- time search query data to overcome the problem of informational time lags of traditional survey data29.

[...]


1 Askitas, N. and Zimmermann K. F. (2009a), p. 11

2 cf. Hubbard, D.W. (2011), p. 5

3 cf. Hubbard, D.W. (2011), p. 5

4 cf. Tam, D. (2012)

5 cf. Hubbard, D.W. (2011), p. 69

6 cf. Rouse, M. (2013)

7 cf. Google Support EN (2014)

8 cf. Google Trends (2014)

9 cf. Askitas, N. and Zimmermann, K. F. (2014), p.5

10 cf. Google Support EN (2014)

11 cf. Google Trends (2014)

12 cf. Google support DE (2014)

13 cf. searchengineland.com (2014)

14 cf. Schwartz, B. (2008)

15 cf. Lardinois, F. (2012)

16 cf. Ginsberg, J. et al. (2009)

17 cf. Askitas, N. and Zimmermann, K.F. (2009a)

18 cf. Schmidt, T. and Vosen, S. (2011)

19 cf. Da, Z., Engelberg, J. and Gao, P. (2011)

20 cf. Beer, F., Hervé, F. and Zouaoui, M. (2012)

21 cf. Latoeiro, P., Ramos, S.B. and Veiga, H. (2013)

22 cf. Choi, H. and Varian, H. (2009a)

23 cf. Choi, H. and Varian, H. (2009a)

24 us.spindices.com (2014)

25 cf. Kulkarni et. al. (2009)

26 cf. Kulkarni et. al. (2009)

27 cf. Wu, L. and Brynjolfsson, E. (2009)

28 cf. Wu, L. and Brynjolfsson, E. (2009)

29 cf. Hohenstatt, R., Käsbauer, M. and Schäfers, W. (2011)

Details

Pages
33
Year
2014
ISBN (eBook)
9783668452442
ISBN (Book)
9783668452459
File size
552 KB
Language
English
Catalog Number
v366524
Institution / College
University of Regensburg
Grade
1,3
Tags
using google trends real estate research definition state strengths weaknesses threats opportunities

Share

Previous

Title: Using Google Trends in Real Estate Research