Next Article in Journal
Controlling Agronomic Variables of Saffron Crop Using IoT for Sustainable Agriculture
Next Article in Special Issue
Changes of Bioclimatic Conditions in the Kłodzko Region (SW Poland)
Previous Article in Journal
Degradation of Organics and Change Concentration in Per-Fluorinated Compounds (PFCs) during Ozonation and UV/H2O2 Advanced Treatment of Tertiary-Treated Sewage
Previous Article in Special Issue
Real Estate Values and Urban Quality: A Multiple Linear Regression Model for Defining an Urban Quality Index
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Sustainability
Volume 14
Issue 9
10.3390/su14095601
sustainability-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Does Google Trends Show the Strength of Social Interest as a Predictor of Housing Price Dynamics?

by
Mirosław Bełej
Department of Spatial Analysis and Real Estate Market, University of Warmia and Mazury in Olsztyn, Oczapowskiego 2, 10-719 Olsztyn, Poland
Sustainability 2022, 14(9), 5601; https://doi.org/10.3390/su14095601
Submission received: 8 April 2022 / Revised: 29 April 2022 / Accepted: 3 May 2022 / Published: 6 May 2022
(This article belongs to the Special Issue Sustainable Cities and Regions – Statistical Approaches)
Browse Figures
Versions Notes

Abstract

:
A recently emerged sustainable information society has ceased to be only a consumer and has become a web-based information source. Society’s online behaviour is tracked, recorded, processed, aggregated, and monetised. As a society, we are becoming a subject of research, and our web behaviour is a source of information for decision-makers (currently mainly business). The research aims to measure the strength of social interest in the housing market (Google Trends), which will then be correlated with the dynamics of housing prices in Poland in the years 2010–2021. The vector autoregressive model was used to diagnose the interrelationships (including Granger causality) and to forecast housing prices. The research showed that web searching for the keyword “dwelling” causes the dynamics of dwelling prices and is an attractive alternative to the classical variables used in forecasting housing market prices.

1. Introduction

The human being functions in a complex and multidimensional socio-economic reality. General principles, rules or patterns of behaviour passed down from generation to generation in new conditions of the human environment (organisational, economic, informational) may be ineffective in the current reality. The evolution of society from hunting, agricultural and industrial to information society has changed the paradigms of thinking and required adaptation to the current conditions of functioning of societies. Almost 20 years ago, Hilty [1] recognised that sustainable development and the emerging information society are two significant visions that characterise the beginning of the 21st century.
Human economic activity, which has significantly disturbed the balance of the ecosystem, provided the basis for introducing the concept of sustainable development, which promotes a balance between socio-economic development and the natural environment. Combining the (equally important) relationships between economy, environment and man in one concept depart from traditional concepts, where these issues were treated separately (ceteris paribus). Sustainability is a concept [2,3] to improve and sustain a healthy economic, ecological and social system for human development. The best-known definition of sustainable development is proposed in the Brundtland Commission Report [4]. The report defines sustainable development as development that meets the needs of the present generation without compromising the ability of future generations to meet their own needs [2]. More detailed concepts of ‘sustainability’ and ‘sustainable development’ are presented in several studies [5,6,7,8,9,10,11].
Hilty [1] believed that at the beginning of the new millennium, we are on the cusp of an information-intensive economy, in which nothing can be done without heavy use of Information and Communication Technology (ICT). The memorandum “Sustainable Information Society” [12] has provided that ICT will penetrate his everyday life and affect people, society and the environment. However, the ever-broader usage of ICT does not automatically favour a sustainable, environmentally fair development. The memorandum stated [12] that:
  • Internet may offer excellent opportunities to develop into a society;
  • ICT offers great opportunities which could be distributed very unevenly in society;
  • It is becoming increasingly difficult to establish a relationship between cause and effect in the digital world;
  • ICT alone does not cause a reduction in the use of natural resources by production and consumption;
  • ICTs offer great potential for sustainable development, but the opportunity to reorient our activities toward a sustainable information society can be lost without discussion.
Today, in 2022, we are fully aware that the new society, which does not remember the times without the Internet, cannot imagine the world without the possibilities offered by modern information and communication technologies (ICT). It can be said that the present society, accepting the general idea of sustainable development, is now transforming itself into a modern and sustainable information society.
According to Thomas [2], sustainable development is focused on the ability of humans to meet human needs. If we become a sustainable information society, the question will bewhether this social development is equally applicable to housing markets, which meet basic human needs. Home (in the Markov hierarchy of needs) offers a feeling of safety [13], a sense of belonging [14], is a factor in the health of societies [15] and is also considered a significant human right [16]. Housing is one of the areas of the economy that has a significant effect on the level of satisfaction of social needs, dynamics of economic processes and effectiveness of developmental activities. Interconnections of housing development and the economy indicate that the former plays a significant role in elevating the country’s social, economic, and spatial cohesion [17,18,19]. Housing is a particular type of commodity [20] because it is a spatially stationary commodity, highly durable, costly, heterogeneous and physically changeable. Sustainable housing significantly influences the satisfaction of social needs and the efficiency of development activities. Improved housing increases the quality of life and contributes to the achievement of several sustainable development goals (SDGs), including those addressing health (SDG 3) and sustainable cities (SDG 11) [21].
A sustainable information society cannot exist without information about the housing market. The primary source of housing market data is the range of government, local government and private institutions that collect, process, aggregate and share it. The information that the public needs in buying a property, thanks to the development of ICT, can now be accessed directly on the internet. Thus, a prospective homebuyer, without leaving his/her home and using tools such as a laptop or smartphone, can analyse: the prices of similar properties, the surroundings of the real estate, the air pollution level, the sunshine level, the security level of the neighborhood (robberies, thefts) and even the traffic jam level in the vicinity of the chosen property. On the one hand, ICT allows society to obtain information about the market, but on the other hand, thanks to ICT, the same society (each person individually) becomes a provider and subject of data analysis (often unconsciously). Society’s web behaviour leaves a footprint that, when processed and aggregated, becomes data. As a result, the actions of one part of society in the sphere of information acquisition generate the possibility of analysing their behaviour for another part of society. Based on such data it is possible to conclude the current needs of society in terms of housing. An instrument to measure public sentiment can be Google Trends.
Internet search trend analysis is becoming increasingly popular in business and science to understand social awareness [22,23,24,25]. According to Yang [26], the activity of internet users at a given point in time is assumed to reflect collective behaviour and shows the interests, concerns and intentions of the observed population. As a freely accessible tool, Google Trends provides information on trends of keywords that people search for on Google [27]. It could be said that Google has dominated the search market. Therefore, it can be a valuable tool for assessing the popularity or public interest of a given product, topic or event. Presenting the absolute number of searches for a given type of topic would generate difficulties compared to other keywords. Therefore, Google Trends provides the relative search volumes of search terms on the web, by estimating the search volume for all searches for a given time and location. The relative search volume index (RSV) is generated by normalising the search results by the highest proportion of queries in the created time series. The wide applications of Google Trends in science and business are outlined in many scientific studies [28,29,30,31,32].
This paper shows social interest in the housing market in Poland (2010–2021) using Google Trends. In the sustainable information society era with widely available ICT tools, an individual searches for information about dwellings, houses or land just on the web. Measuring the significance of such interest in time may help diagnose the future, i.e., in forecasting the dynamics of housing prices. It seems reasonable to assume that the increased social interest in real estate on the web results after some time in a reaction through the purchase of a dwelling or a house. Of course, it is necessary to distinguish the potential interest of the real, just as the offer price of the transaction price is distinguished.
The research concerns internet searches for the word “flat” in Polish, which is a rather small percentage of similar research, where the English language dominates. The Polish residential market is currently very attractive to investors from all over the world. This is due to Poland’s stable and growing economy, still lower prices than in Western Europe and a stable upward trend in residential prices. Currently, there is a remarkable increase in individual investment in housing to protect against the loss of capital, due to the National Bank of Poland’s low interest rate policy. In such a situation, the presented research is very timely and helpful in the evaluation of investment decisions. The presented research may contribute to the discussion on potential drops in residential prices in Poland, an essential element of investors’ risk minimisation policy. Investros’ beliefs in the continuation of price increases is currently dominant. The important question, then, is whether such an assumption is still possible. This research allows us to partially change the prevailing assumptions on the housing market.
The adopted time horizon (2010–2021) of the research allows analysis of opposing phases of residential market activity in Poland. In the first phase of this period (2010−2013), we observed a decrease in housing prices after peaks reached in 2007−2008. This was a consequence of global megatrends connected with the so-called mortgage crisis. In the second phase of the analysis period (2013−2021), we observed a strong upward trend in residential prices, which was not stopped even by the appearance of the COVID-19 pandemic in Poland in March 2020. You can read more about the housing market in Poland in the extensive literature on the subject [33,34,35,36,37,38,39,40,41]. The use of Google Trends in such distinct phases (price decreases and increases) allows us to take a more in-depth look at the public’s actual behaviour in the housing market.
In this study, the vector autoregressive model VAR was used together with the analysis of Granger causality. All computations and their visualisation were carried out in the R software.

2. Materials and Methods

2.1. Study Area and Data Description

The research was conducted on the Polish housing market in 2010–2021. The National Bank of Poland (BaRN) database was the source of data on housing prices. The research assumes average quarterly prices per square meter of a flat (Secondary Residential Market) weighted by the city’s residential market resources, averaged for the seven largest Polish cities (Gdańsk, Gdynia, Kraków, Łódź, Poznań, Warsaw and Wrocław). Data on public interest in housing were obtained from the Google Trends website. The options available on this website are insufficient for a more precise and accurate keyword search. Therefore, data collection has been carried out using the “gtrends” package in R software. First, the search had to be conducted in Polish, as the research concerned a Polish society. It has been assumed that the keyword is “mieszkanie”, which means a dwelling, a flat or an apartment in Polish. Secondly, the script in R allows you to select a search category to minimise errors. In the case of this work, code 1080 (Real Estate Listings) was chosen, which allows us to search for terms correlated with housing listings in real estate companies.

2.2. Methodology

Autoregressive vector models (VAR) explain the endogenous variables solely by their history and could include certain exogenous variables, such as trends and seasonal dummies, but they do not have to classify variables as endogenous or exogenous [42]. VAR models are a bridge between traditional econometrics and time series models. They are multiequation models whose expansions may be consistent with even incredibly detailed economic theory. At the same time, they make full use of information on processes that generate variables [43]. The basic form of the VAR model is:
x t = A 0 D t + A 1 x t − 1 + A 2 x t − 2 + … + A k x t − k + e t
where:
  • xt—vector of observations on the current values of all n variables model x = [ x 1 t   x 2 t   x 3 t …   x n t ] T ,
  • Dt—vector of deterministic components of equations,
  • A0—matrix of parameters with the vector variables Dt, not containing zero elements,
  • Ai—matrix of parameters with lagged variables of vector xt, not containing zero elements,
  • et—vectors of stationary random disturbances et = [et1 et2 et3 … etn]T.
As opposed to structural models, the VAR model allows a broad analysis of system linkages without maintaining the “ceteris paribus” principle. They do not explicitly require the identification of endogenous exogenous variables [44], no causal relationships between vector variables are excluded and no restrictions are imposed by design [45,46]. This assumes that every variable affects every other variable in the system. VAR models are helpful in many contexts [45,47,48,49,50]:
  • forecasting a set of related variables where no unambiguous interpretation is required;
  • examining if one variable is predictive of another (Granger causality tests basis);
  • impulse response analysis, which looks at the response of one variable to a sudden but instantaneous change in another variable;
  • decomposition of the forecast error variance, in which part of the variance of the forecast of each variable is attributed to the influence of the other variables.

3. Results

The time horizon of the research (Q1 2010−Q4 2021) resulted from the availability of data on housing prices, as data for Q4 2021 were not available until February 2022. The beginning of the research (Q1 2010) coincides with the stabilisation of the housing market after the massive price increases of 2007−2008. Google Trends analysis of keyword dwelling (“mieszkanie”−in PL) gave us the so-called relative search volume index (RSV). The RSV index represents the relative level of interest in housing on a scale from 0 to 100. Figure 1 shows the results for the keyword “mieszkanie” search (2010–2021) and the beginning of COVID-19 in Poland. The monthly RSV values were converted to quarterly values due to the quarterly housing price data.
Figure 1 shows the dynamics of the monthly RSV values. As we can see, from 2010 to 2016, the level of interest oscillated between 30 and 60 RSV, with a stable increasing trend. The 2017−2019 period shows significant public interest in housing, with RSV levels rising to almost 90. The peak in popularity of searches for the term “dwelling” in the Google search engine in August 2021. However, the smoothed RSV index values oscillate around the value of 80 RSV, falling to 57 RSV in December 2021. Smoothing was accomplished using a loess procedure. An interesting result is the complete lack of negative reactions of interest in dwellings after the start of the COVID-19 pandemic in March 2020. In the following months of COVID-19, there is an increase in dwellings interest. It is even more evident that there is a change in the long-term trend, from increasing to decreasing, when considering the last two years of analysis in weekly terms (Figure 2).
In the case of the weekly RSV index results for searches for the term dwelling in 2020−2021, the traditional seasonality associated with the seasons is visible. However, by the end of 2022, there is a successively increasing decrease in interest in dwellings in Poland. It should be noted that the purpose of this investigation was not to analyse all factors that may influence the prices of dwellings. For this reason, the further part will not include other crucial factors that influence the dynamics of housing prices. One of the most significant ones is the decisions of the Monetary Policy Council of the National Bank of Poland, which influenced the increase of the main interest rates and the increase in credit costs as well as reduced credit availability for the poorer part of society.
The monthly relative search volume (RSV) index for the keyword ‘‘mieszkanie’’ has been transformed into quarterly values aligned with Poland’s quarterly housing prices. Figure 3 shows the time series of the RSV index and housing prices per square meter (absolute and logarithmic values). The lack of a sufficiently long time series for the Polish real estate market is a fundamental problem that cannot be currently overcome. A priori it is assumed that there is no long-term relationship between the variables [51]. According to Adkins [52] and Brooks [53], a differencing (to achieve stationarity) can erase long-term information. The next part of the study used logarithms of the variables, as recommended by Canova and Drachal [51,54].
Figure 3 shows that RSV values continuously increase from 2010 to early 2021, after which they fall dramatically. Housing prices, determined as a weighted average of prices from the seven biggest Polish cities, fall to 5.5 thousand PLN (1200 EUR) per square meter by 2013. Since then, an uninterrupted increase in prices to almost 10 thousand PLN (2200 EUR) per square meter is visible.
In the first stage of the study, the OLS linear regression model (Table 1) was used for logarithmic values of quarterly time series of secondary market prices (SHM) and relative search volume (RSV).
The results of the OLS model indicate that RSV coefficient is positive and is statistically significant. This means that the increase in web searches for “dwelling” relates to a dwelling increase trend. Due to its structure, the OLS model imposes that the RSV coefficient affects the SHM coefficient. This study employs a vector autoregressive (VAR) philosophy, which is that structures should not be imposed and it should be decided which coefficients matter. In the first step of the vector autoregression model, the optimal lag level for the time series under study should be found. For this purpose, the VARselect function from the vars package in R was used. The function returns information criteria and the final prediction error to sequentially increase the lag order up to a VAR(p)-process, based on the same sample size [42]. VAR models are multidimensional models of the ARMA class [44], where the autoregressive (AR) process of lag length prefers a time series in which its current value is dependent on its first p-lag values [55]. The level of lags p is established based on four information criteria: ‘Aikaike’s information criterion (AIC), Hannan-Quinn criterion (HQ), Schwarz information criterion (SC) and final prediction error (FPE). The results of the calculations in Table 2.
Table 2 shows that the following results were obtained: AIC(4), HQ(4), SC(2) and FPE(4). Only the Schwarz (SC) information criterion obtains a minimum lag value (2). According to the literature [55,56,57], caution should be taken when applying the AIC, as it tends to select a large number of lags. Instead, the SC information criterion should be preferred for VAR models because, in small samples, AICs and FPEs generally achieve higher latency values than HQs and SCs [44]. In a further step, the VAR model (2) is therefore estimated, and the results are shown in Table 3.
The VAR(2) model was then tested: Portmanteau test, heteroscedasticity test, Jarque-Bera test, Testing for Structural Breaks in the Residuals. The p-value in Portman Test with 12 lags is greater than 0.05 (p-value = 0.3331), which suggests that there is no serial correlation in the VAR model, so that passes. The test for heteroscedasticity with 12 lags gives a p-value greater than 0.05 (p-value = 0.789), which means that the model does not suffer from heteroscedasticity and the Vars model passes that test. The Jarque-Bera test The Jarque-Bera test shows that the p-value is greater than 0.05 (p-value = 0.596), so we cannot reject the null hypothesis that the residuals are normally distributed. The results of testing for structural breaks in the residuals are shown in Figure 4, which shows that there are no points which exceed the two red lines (upper and lower confidence interval), so the VAR(4) model is stable. More on the essence of this test can be found in the paper [58].
The construction of the Var model is a basic, but also an initial stage of the analyses. Now, forecasts can be built, causality analysis can be done, impulse response functions, decompose error variance can be decomposed for structural VAR models or vector error correction models can be created [34]. A causality analysis of time series forecasting will be demonstrated in the current work.
The VAR philosophy, introduced earlier in this paper, does not explicitly impose causality and, as a result, allows us to study which variable influences which one. Therefore, in this step of the VAR(2) model, a causality analysis in the Granger sense will be carried out. Causality in Granger’s sense is the dependence of processes generating data. Variable X is a cause of variable Y, and including past values of variable X in the model predicting variable Y will increase prediction accuracy [59]. The Granger causality test in the VAR(2) model assumes the null hypothesis H0: RSV (relative search volume) does not have Granger-cause SHM (housing prices from secondary market) or H0: SHM does not have Granger-cause RSV. Additionally, the study used instant causality, where the null hypothesis was H0: No instantaneous causality between RSV and SHM or H0; no instantaneous causality between SHM and RSV. Table 4 shows the results.
The p-value of the Granger causality (RSV → SHM) is much smaller than 0.05, indicating that RSV Granger causes SHM. In other words, the movements in searching for the term “dwelling” in the Google search engine precede movements in housing price dynamics. As a result, we can say that the value of the RSV coefficient could be useful in predicting future changes in housing prices. In the case of immediate causality, the null hypothesis cannot be rejected, which confirms the general knowledge of the inertia property market. For an opposite relation to SHM → RSV, the null hypothesis cannot be rejected for both causalities in the Granger sense and immediate causality. In other words, the movements in dwelling prices do not precede movements in searching for ‘‘dwelling’’ in Google. According to the assumptions of this paper, the influence of other macroeconomic factors on the dynamics of housing prices is not analysed. In the context of the sustainable information society, it can be assumed that the society searching for the word “dwelling” on the internet is influenced by external factors in the form of a range of economic and social information. As a result, the search dynamics for the word “dwelling” depends on assessing the existing risk in the environment. We can refer to this as the RSV index (relative search volume). It allows one to evaluate the actual tendencies of the housing market, as the intensification of the search for this word on the web may indicate a real interest in the purchase of a flat in the future. On the one hand, the information society makes use of a range of available social and macroeconomic information via the web, while on the other hand, it provides information on the investment potential at a given time. Figure 5 shows the relative search volume (RSV) forecasts and the housing prices from the secondary market (SHM) in Poland for six months and one year of the forecast.
In the case of the shorter forecast period (1/2 year), we observe a stabilisation of residential prices, which is taking place for the first time in the last five years. The one-year forecast predicts a change from the previous upward trend in housing prices and the start of price falls in the second half of the year. At the same time, the RSV indicator is forecast to further reduce the interest in housing. In connection with this, a series of negative information reaching the society influences the decrease of interest in web searching for the term “dwelling”. These phenomena include rising interest rates at the National Bank of Poland, the increasing level of inflation and the increasing requirements of banks regarding so-called creditworthiness.

4. Discussion

Research on the use of web searches for words such as “house”, “dwelling”, “rental”, “apartment” and “real estate” is steadily increasing in the world literature. This is a natural consequence of the emergence of a sustainable information society, whose natural environment is collecting, processing, and generating information. The information is available at any time and any place with the current availability of the World Wide web via smartphones. Searching for a dwelling, house, or rental costs can be done on a tram or bus, while waiting for a doctor’s appointment or at work. This way of obtaining information on the property market has become extremely popular in the last decade. Hence, there is a growing trend of research using data generated by Internet users, as each search engine operation (e.g., Google) is recorded, stored, aggregated and processed to produce new information. According to Matias [60], since launching Google Trends, we have seen a great deal of interest in what can be learned from search trends. Many studies have shown how to use search trends data for effective nowcasting in diverse areas, such as health, finance, economics, politics and more. This part of the discussion will focus exclusively on the real estate market, and to this extent, the subject is difficult to find in the Polish research literature. For example, the Bulczak study [61] on Google Trends for forecasting in the real estate market concerned not Poland but Great Britain from the years 2004−2014. As a result, the presented study may be one of the first and certainly one of the most up to date (the horizon of this research is 2010–2021) concerning the Polish residential market in the area of research. Due to the impossibility of comparing the obtained research results for the Polish residential market, the further scientific discussion will be based on international publications.
Limmios and You [62], in their work” Can Google Trends Improve Housing Market Forecasts”, state that their research derives from the essence of empirical economics and that incorporating the Google Trend data deteriorates the forecasting ability. In this research, the keywords “real estate agency” were adopted, whereas it is the keyword “dwelling” in this study. It seems that people looking for a dwelling to buy do not type into Google the phrase “real estate agency”, instead typing only what they are looking for, i.e., a dwelling, a house, a house or a plot of land, or they add the word “sell” or “rent”. At the same time, the studies presented were not based on the VAR model. Askitas [63], who used the keyword “buy and sell home” in Google Trends, presented a more similar research concept, and, as a result, confirmed the high correlation with the SP/Case-Shiller Home Price Index confirmed the benefits of forecasting using this method. The concept adopted in this work is also confirmed by the experience of Berach and Wintoki [64], who considered that the intensity of the search for real estate terms for a specific city can be treated as a proxy of the sentiment of the housing market for that city. In their study, they used the keyword “real estate”, which is not as unambiguous as a dwelling (“mieszkanie”−in polish). This research has another element in common with this study, as it uses data from the Housing Price Index provided by the Federal Housing Finance Agency (FHFA). This index shows quarterly price changes, as does the National Bank of Poland data adopted in this work. Interesting use of Google Trends as an indicator to analyse housing demand was presented by Huarng et al. [65]. This paper also searched Google Trends for a term referring to Real Estate Agency (as in work [62]). However, it did not refer to an unspecified type of real estate company but used the most popular non-real estate sales service in Taiwan. The keyword thus adopted may coincide with the potential demand in the housing market. This is a concept worth considering for subsequent work.
VAR models are widely used to study housing market price dynamics; however, it works simultaneously, with Google Trends and the VAR model incorporated in several studies. For example, in one study [63] based on freely accessible, real-time Google Trends data, the research analysed a statistically significant contribution to unemployment dynamics by GTU shocks in the United States with the VAR model. Davis and Heathcote [66] showed that housing investments are twice as volatile as other investments and are ahead of cyclical fluctuations, while non-residential investments constitute a lag variable. In the work of Cellmer et al. [49], a VAR was constructed, and Granger tests and impulse response analysis were performed using the impulse response function (IRF). As a result, it has been shown that the response of real estate prices to the impulse from the explanatory variables appears between the first and fourth quarters and expires after about three years. Xiao and Zhou [67] found that by building a VAR model, they can evaluate the effectiveness of regulatory policies in the Shanghai real estate market. In the following work, we analyse the dynamic spillover effects of shocks on the housing market, using a six-variate VAR model that was conducted in the subsequent work [48]. In this paper, housing market variables, housing price and housing trading volume are used, and as shock variables, housing demand shock, housing supply shock, interest rate shock, household loan shock, aggregate demand shock and aggregate supply shock are taken into account. Most of the available articles on the use of the VAR model in housing market analyses do not show a strong resemblance to the present study. This may indicate the originality and topicality of the research presented.

5. Conclusions

Since 2013, the housing market in Poland has shown an increasing trend, and housing prices in 2021 have exceeded the previous maxima of 2008. At the end of 2021, interest in the housing market expressed through Google searches for the term “dwelling” changed direction from increasing to decreasing. It may be argued that the decrease in the information society’s interest in the analysis of dwellings available for purchase in estate sales offices precedes the decrease in the real demand in the housing market. This may result in a change in the previous trend and a decrease in residential prices. It seems that the prices of dwellings in Poland at the beginning of 2022, have been significantly overestimated and are less and less adjusted to the income level of the society. This is because for an average net salary, one can buy half a square meter of a flat. The source of such high prices in 2021 has been unusually low-interest of main rates and very low bank deposit rate returns and rising inflation at the same time as profits from deposits from other financial investments, increasing interest in investing capital in real estate. Thus, the driving force behind recent increases in housing prices has been investment demand. Google Trends has shown a decrease in public interest in real estate companies. The adopted RSV index is an interesting alternative to the variables constituting the basis for the classical forecasting of housing market prices. The next research stage should be housing prices using the VAR model, taking into account the RSV index and other available variables.

Funding

This research received no external funding.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

Housing prices–Narodowy Bank Polski–Internetowy Serwis Informacyjny (nbp.pl).

Acknowledgments

The author expresses his sincere gratitude to the Journal Editor and the anonymous Reviewers who spent their valuable time providing constructive comments and assistance to improve the quality of this paper.

Conflicts of Interest

The author declares no conflict of interest.

References

  1. Hilty, L.M. Sustainable Development and the Information Society. In Human Choice and Computers: Issues of Choice and Quality of Life in the Information Society; Brunnstein, K., Berleur, J., Eds.; Springer Science & Business Media: Berlin, Germany, 2002; pp. 305–315. [Google Scholar]
  2. Thomas, C.F. Naturalizing Sustainability Discourse: Paradigm, Practices and Pedagogy of Thoreau, Leopold, Carson and Wilson; Arizona State University: Tempe, AZ, USA, 2015; ISBN 132172571X. [Google Scholar]
  3. Mensah, J. Sustainable Development: Meaning, History, Principles, Pillars, and Implications for Human Action: Literature Review. Available online: http://www.editorialmanager.com/cogentsocsci (accessed on 1 January 2019).
  4. Schaefer, A.; Crane, A. Addressing Sustainability and Consumption. J. Macromark. 2016, 25, 76–92. [Google Scholar] [CrossRef] [Green Version]
  5. Giovannoni, E.; Fabietti, G. What Is Sustainability? A Review of the Concept and Its Applications. In Integrated Reporting: Concepts and Cases that Redefine Corporate Accountability; Springer: Berlin, Germany, 2013; pp. 21–40. [Google Scholar] [CrossRef]
  6. du Pisani, J.A. Sustainable Development—Historical Roots of the Concept. Environ. Sci. 2006, 3, 83–96. [Google Scholar] [CrossRef]
  7. Pawłowski, A. The Sustainable Development Revolution. Probl. Sustain. Dev. 2009, 4, 65–76. [Google Scholar]
  8. Sustainable Development: Principles, Frameworks, and Case Studies. Int. J. Sustain. High. Educ. 2011, 12, 434–438. [CrossRef]
  9. Mozaffar, Q. Sustainable Development: Concepts and Rankings. J. Dev. Stud. 2001, 3, 134–161. [Google Scholar]
  10. Berke, P.R.; Conroy, M.M. Are We Planning for Sustainable Development? J. Am. Plan. Assoc. 2000, 66, 21–33. [Google Scholar] [CrossRef]
  11. Jabareen, Y. A New Conceptual Framework for Sustainable Development. Environ. Dev. Sustain. 2006, 10, 179–192. [Google Scholar] [CrossRef]
  12. Gohring, W. The Memorandum” Sustainable Information Society. In Sh@ring—EnviroInfo; Minier, P., Susini, A., Eds.; Editions du Tricorne: Geneva, Switzerland, 2004. [Google Scholar]
  13. Protopopova, A. Effects of Sheltering on Physiology, Immune Function, Behavior, and the Welfare of Dogs. Physiol. Behav. 2016, 159, 95–103. [Google Scholar] [CrossRef]
  14. Museus, S.D.; Yi, V.; Saelua, N. The Impact of Culturally Engaging Campus Environments on Sense of Belonging. Rev. High. Educ. 2017, 40, 187–215. [Google Scholar] [CrossRef]
  15. Dovie, D.A. Assessment of How House Ownership Shapes Health Outcomes in Urban Ghana. Societies 2019, 9, 43. [Google Scholar] [CrossRef] [Green Version]
  16. Leckie, S. Housing as a Human Right. Environ. Urban. 1989, 1, 90–108. [Google Scholar] [CrossRef] [Green Version]
  17. Case, K.E.; Quigley, J.M.; Shiller, R.J. Comparing Wealth Effects: The Stock Market versus the Housing Market. Adv. Macroecon. 2005, 10, 5. [Google Scholar] [CrossRef]
  18. Lis, P. Wahania Cykliczne Rynków Mieszkaniowych. In Aspekty Teoretyczne i Praktyczne; Wydawnictwo Adam Marszałek: Toruń, Poland, 2012. [Google Scholar]
  19. Ball, M. Housing Policy and Economic Power: The Political Economy of Owner Occupation; Routledge: London, UK, 2017; Volume 828, ISBN 1-135-83596-9. [Google Scholar]
  20. Galster, G. William Grigsby and the Analysis of Housing Sub-Markets and Filtering. Urban Stud. 1996, 33, 1797–1805. [Google Scholar] [CrossRef]
  21. United Nations. Transforming Our World: The 2030 Agenda for Sustainable Development. Available online: https://www.un.org/ga/search/view_doc.asp?symbol=A/RES/70/1&Lang=E (accessed on 1 September 2016).
  22. Li, C.; Chen, L.J.; Chen, X.; Zhang, M.; Pang, C.P.; Chen, H. Retrospective Analysis of the Possibility of Predicting the COVID-19 Outbreak from Internet Searches and Social Media Data, China. Eurosurveillance 2020, 25, 2000199. [Google Scholar] [CrossRef]
  23. Verma, M.; Kishore, K.; Kumar, M.; Sondh, A.R.; Aggarwal, G.; Kathirvel, S. Google Search Trends Predicting Disease Outbreaks: An Analysis from India. Healthc Inf. Res. 2018, 24, 300–308. [Google Scholar] [CrossRef]
  24. Sun, S.; Wei, Y.; Tsui, K.-L.; Wang, S. Forecasting Tourist Arrivals with Machine Learning and Internet Search Index. Tour. Manag. 2019, 70, 1–10. [Google Scholar] [CrossRef]
  25. Wu, S.; Liu, J.; Liu, L. Modeling Method of Internet Public Information Data Mining Based on Probabilistic Topic Model. J. Supercomput. 2019, 75, 5882–5897. [Google Scholar] [CrossRef]
  26. Yang, S.; Santillana, M.; Kou, S.C. Accurate Estimation of Influenza Epidemics Using Google Search Data via ARGO. Proc. Natl. Acad. Sci. USA 2015, 112, 14473–14478. [Google Scholar] [CrossRef] [Green Version]
  27. Kardeş, E.; Kardeş, S. Google Searches for Bruxism, Teeth Grinding, and Teeth Clenching during the COVID-19 Pandemic. J. Orofac. Orthop. Fortschr. Der Kieferorthopädie 2021, 1–6. [Google Scholar] [CrossRef]
  28. Nuti, S.V.; Wayda, B.; Ranasinghe, I.; Wang, S.; Dreyer, R.P.; Chen, S.I.; Murugiah, K. The Use of Google Trends in Health Care Research: A Systematic Review. PLoS ONE 2014, 9, e109583. [Google Scholar] [CrossRef] [Green Version]
  29. Hamulka, J.; Jeruszka-Bielak, M.; Górnicka, M.; Drywień, M.E.; Zielinska-Pukos, M.A. Dietary Supplements during COVID-19 Outbreak. Results of Google Trends Analysis Supported by PLife COVID-19 Online Studies. Nutrients 2020, 13, 54. [Google Scholar] [CrossRef] [PubMed]
  30. Lee, H.S. Exploring the Initial Impact of COVID-19 Sentiment on US Stock Market Using Big Data. Sustainability 2020, 12, 6648. [Google Scholar] [CrossRef]
  31. Lengyel, A. Tourism, Meditation, Sustainability. APSTRACT Appl. Stud. Agribus. Commer. 2016, 10, 81–92. [Google Scholar] [CrossRef]
  32. Önder, I.; Gunter, U. Forecasting Tourism Demand with Google Trends for a Major European City Destination. Tour. Anal. 2016, 21, 203–220. [Google Scholar] [CrossRef]
  33. Brzezicka, J.; Wisniewski, R. Translocality on the Real Estate Market. Land Use Policy 2016, 55, 166–181. [Google Scholar] [CrossRef]
  34. Cellmer, R.; Kobylińska, K.; Bełej, M. Application of Hierarchical Spatial Autoregressive Models to Develop Land Value Maps in Urbanized Areas. ISPRS Int. J. Geo-Inf. 2019, 8, 195. [Google Scholar] [CrossRef] [Green Version]
  35. Bełej, M. Synergistic Network Connectivity among Urban Areas Based on Non-Linear Model of Housing Prices Dynamics. Real Estate Manag. Valuat. 2018, 26, 22–34. [Google Scholar] [CrossRef] [Green Version]
  36. Rącka, I.; Palicki, S.; Krajewska, M.; Szopińska, K.; Kempa, O. Changes on the Housing Market of the Downtown Area in Selected Polish Cities. Real Estate Manag. Valuat. 2017, 25, 79–90. [Google Scholar] [CrossRef] [Green Version]
  37. Bieda, A. Parametric Model of Real Estate Valuation for Land Located in Different Land-Use Zones. Geomat. Environ. Eng. 2017, 11, 17. [Google Scholar] [CrossRef] [Green Version]
  38. Kokot, S.; Gnat, S. Simulative Verification of the Possibility of Using Multiple Regression Models for Real Estate Appraisal. Real Estate Manag. Valuat. 2019, 27, 109–123. [Google Scholar] [CrossRef] [Green Version]
  39. Głuszak, M. Multinomial Logit Model of Housing Demand in Poland. Real Estate Manag. Valuat. 2015, 23, 84–89. [Google Scholar] [CrossRef] [Green Version]
  40. Cellmer, R.; Trojanek, R. Towards Increasing Residential Market Transparency: Mapping Local Housing Prices and Dynamics. ISPRS Int. J. Geo-Inf. 2019, 9, 2. [Google Scholar] [CrossRef] [Green Version]
  41. Tomal, M. House Price Convergence on the Primary and Secondary Markets: Evidence from Polish Provincial Capitals. Real Estate Manag. Valuat. 2019, 27, 62–73. [Google Scholar] [CrossRef] [Green Version]
  42. Pfaff, B. VAR, SVAR and SVEC Models: Implementation Within R Package Vars. J. Stat. Softw. 2008, 27, 1–32. [Google Scholar] [CrossRef] [Green Version]
  43. Kusideł, E. Modele Wektorowo-Autoregresyjne–VAR–Metodologia i Zastosowania; Dane Panelowe i Modelowanie Wielowymiarowe w Badaniach Ekonomicznych; Suchecki, B., Ed.; Absolwent: Łódź, Poland, 2000; Volume 3. [Google Scholar]
  44. Milo, W.; Łapińska-Sobczak, N.; Malaczewski, M.; Szafrański, G.; Ulrichs, M.; Wośko, Z. Stabilność Rynków Finansowych a Wzrost Gospodarczy; Wydawnictwo Naukowe PWN: Warszawa, Poland, 2010; ISBN 8371716087. [Google Scholar]
  45. Clements, M.P.; Mizon, G.E. Empirical Analysis of Macroeconomic Time Series: VAR and Structural Models. Eur. Econ. Rev. 1991, 35, 887–917. [Google Scholar] [CrossRef]
  46. Zivot, E.; Wang, J. Vector Autoregressive Models for Multivariate Time Series. Modeling Financ. Time Ser. S-PLUS® 2006, 385–429. [Google Scholar] [CrossRef]
  47. van Dam, A.; Frenken, K. Variety, Complexity and Economic Development. Res. Policy 2020, in press. [Google Scholar] [CrossRef] [Green Version]
  48. Min, S.-O.; Lee, Y.-S. Korean Housing Market Dynamics: A VAR Analysis with Sign Restrictions. J. Real Estate Anal. 2019, 5, 1–13. [Google Scholar] [CrossRef]
  49. Cellmer, R.; Bełej, M.; Cichulska, A. Identification of Cause-And-Effect Relationships in the Real Estate Market Using the Var Model and the Granger Test. Real Estate Manag. Valuat. 2019, 27, 85–95. [Google Scholar] [CrossRef] [Green Version]
  50. Bose, E.; Hravnak, M.; Sereika, S.M. Vector Autoregressive (VAR) Models and Granger Causality in Time Series Analysis in Nursing Research: Dynamic Changes among Vital Signs Prior to Cardiorespiratory Instability Events as an Example. Nurs. Res. 2017, 66, 12. [Google Scholar] [CrossRef] [Green Version]
  51. Drachal, K. Causality in the Polish Housing Market: Evidence from Biggest Cities. Financ. Assets Investig. 2018, 9, 5–20. [Google Scholar] [CrossRef]
  52. Adkins, L. Using Gretl for Principles of Econometrics; Oklahoma State University, Department of Economics and Legal Studies in Business: Stillwater, OK, USA, 2014. [Google Scholar]
  53. Brooks, C. RATS Handbook to Accompany Introductory Econometrics for Finance; Cambridge Books: London, UK, 2008. [Google Scholar]
  54. Canova, F. Vector Autoregressive Models: Specification, Estimation, Inference, and Forecasting. In Handbook of Applied Econometrics Volume 1: Macroeconomics; John Wiley & Sons, Ltd.: New York, NY, USA, 1999; pp. 53–110. ISBN 9781405166423. [Google Scholar]
  55. Liew, V.K.-S. Which Lag Length Selection Criteria Should We Employ? Econ. Bull. 2004, 3, 1–9. [Google Scholar]
  56. Billah, B.; Hyndman, R.J.; Koehler, A.B. Empirical Information Criteria for Time Series Forecasting Model Selection. J. Stat. Comput. Simul. 2005, 75, 831–840. [Google Scholar] [CrossRef]
  57. Khan, D.M.; Yahya, N.; Kamel, N. Optimum Order Selection Criterion for Autoregressive Models of Bandlimited EEG Signals. In Proceedings of the 2020 IEEE-EMBS Conference on Biomedical Engineering and Sciences (IECBES), Langkawi Island, Malaysia, 1–3 March 2021; IEEE: Piscataway, NJ, USA, 2021; pp. 389–394. [Google Scholar]
  58. Zeileis, A.; Leisch, F.; Hornik, K.; Kleiber, C. Strucchange: An R Package for Testing for Structural Change in Linear Regression Models. J. Stat. Softw. 2002, 7, 1–38. [Google Scholar] [CrossRef] [Green Version]
  59. Granger, C.W.J. Investigating Causal Relations by Econometric Models and Cross-Spectral Methods. Econometrica 1969, 37, 424. [Google Scholar] [CrossRef]
  60. Matias, Y. Nowcasting with Google Trends. In Proceedings of the International Symposium on String Processing and Information Retrieval, Jerusalem, Israel, 7–9 October 2013; Springer: Berlin, Germany, 2013; p. 4. [Google Scholar]
  61. Bulczak, G.M. Use of Google Trends to Predict the Real Estate Market: Evidence from the United Kingdom. Int. Real Estate Rev. 2021, 24, 613–631. [Google Scholar] [CrossRef]
  62. Limnios, A.C.; You, H. Can Google Trends Improve Housing Market Forecasts? Curiosit. Interdiscip. J. Res. Innov. 2021, 1, 21987. [Google Scholar] [CrossRef]
  63. Askitas, N. Trend-Spotting in the Housing Market. Cityscape 2016, 18, 165–178. [Google Scholar] [CrossRef]
  64. Beracha, E.; Wintoki, M.B. Forecasting Residential Real Estate Price Changes from Online Search Activity. J. Real Estate Res. 2013, 35, 283–312. [Google Scholar] [CrossRef]
  65. Huarng, K.-H.; Hui-Kuang Yu, T.; Rodriguez-Garcia, M. Qualitative Analysis of Housing Demand Using Google Trends Data. Econ. Res. Ekon. Istraž. 2020, 33, 2007–2017. [Google Scholar] [CrossRef] [Green Version]
  66. Davis, M.A.; Heathcote, J. Housing and the Business Cycle. Int. Econ. Rev. 2005, 46, 751–784. [Google Scholar] [CrossRef] [Green Version]
  67. Xiao, L.; Zhou, X. Research on the Influence of Regulatory Policies of Shanghai Real Estate on the House Price—Based on the Empirical Study of VAR Model. In Proceedings of the 2018 3rd International Conference on Education, E-learning and Management Technology (EEMT 2018), Bangkok, Thailand, 29–31 October 2018; Atlantis Press: Amsterdam, The Netherlands, 2018; pp. 140–145. [Google Scholar]
Sustainability 14 05601 g001 550
Figure 1. Google web searches for the keyword “mieszkanie” in Poland (2010–2021). The black dots indicate the monthly RSV level, the red line indicates the smoothed RSV values, the vertical purple line indicates the beginning of COVID-19 in Poland and the green vertical line indicates the highest level of interest in flats.
Figure 1. Google web searches for the keyword “mieszkanie” in Poland (2010–2021). The black dots indicate the monthly RSV level, the red line indicates the smoothed RSV values, the vertical purple line indicates the beginning of COVID-19 in Poland and the green vertical line indicates the highest level of interest in flats.
Sustainability 14 05601 g001
Sustainability 14 05601 g002 550
Figure 2. Google web searches for the keyword ‘‘mieszkanie’’ in Poland (2010–2021). Black dots indicate monthly RSV levels, the red line indicates smoothed RSV values and the vertical purple line indicates the start of COVID-19 in Poland.
Figure 2. Google web searches for the keyword ‘‘mieszkanie’’ in Poland (2010–2021). Black dots indicate monthly RSV levels, the red line indicates smoothed RSV values and the vertical purple line indicates the start of COVID-19 in Poland.
Sustainability 14 05601 g002
Sustainability 14 05601 g003 550
Figure 3. Quarterly time series of relative search volume (RSV) and secondary market housing prices (SHM) in Poland (2010–2021): (a) absolute values of RSV and SHM; (b) logarithmic values of RSV and SHM.
Figure 3. Quarterly time series of relative search volume (RSV) and secondary market housing prices (SHM) in Poland (2010–2021): (a) absolute values of RSV and SHM; (b) logarithmic values of RSV and SHM.
Sustainability 14 05601 g003
Sustainability 14 05601 g004 550
Figure 4. The results of the structural breakdown testing in the residuals of VAR(2).
Figure 4. The results of the structural breakdown testing in the residuals of VAR(2).
Sustainability 14 05601 g004
Sustainability 14 05601 g005 550
Figure 5. Forecasting of quarterly time series of relative search volume (RSV) and housing prices from the secondary market (SHM) in Poland (2010–2021): (a) 1/2-year forecast; (b) 1-year forecast.
Figure 5. Forecasting of quarterly time series of relative search volume (RSV) and housing prices from the secondary market (SHM) in Poland (2010–2021): (a) 1/2-year forecast; (b) 1-year forecast.
Sustainability 14 05601 g005
Table
Table 1. OLS regression results for logarithmic time series of secondary market prices (SHM) and relative search volume (RSV).
Table 1. OLS regression results for logarithmic time series of secondary market prices (SHM) and relative search volume (RSV).
CoefficientStd. Err.t-Valuep-Value
Intercept6.8770.32120.852<0.001 ***
Log(RSV)0.4620.0795.801<0.001 ***
R20.42
Adj R20.41
F-statistic33.65 on 1 and 46 df, p-value < 0.001 ***
Note: *** p < 0.001.
Table
Table 2. Criteria for selecting the lag level in the VAR model.
Table 2. Criteria for selecting the lag level in the VAR model.
Criterion/Lags12345
AIC(n)−1.591995 × 101−1.629301 × 101−1.642528 × 101−1.654289 × 101−1.645356 × 101
HQ(n)−1.573676 × 101−1.604875 × 101−1.611996 × 101−1.617650 × 101−1.602611 × 101
SC(n)−1.541329 × 101−1.561746 × 101−1.558084 × 101−1.552956 × 101−1.527134 × 101
FPE(n)1.224709 × 10−78.487537 × 10−87.516122 × 10−86.792835 × 10−87.605687 × 10−8
Table
Table 3. VAR (2) model estimation.
Table 3. VAR (2) model estimation.
CoefficientStd. Err.t-Valuep-Value
RSV.l11.8885480.14335513.1749.61 × 10−16 ***
SHM.l1−0.2098840.219489−0.9560.3450
RSV.l2−0.8756530.143046−6.1213.88 × 10−7 ***
SHM.l20.1567330.2177380.7200.4760
const0.4129260.2173991.8990.0651
sd10.0217280.0077612.8000.0080 **
sd20.0081830.0083060.9850.3308
sd30.0098070.0077481.2660.2133
R20.994
Adj R20.993
F-statistic858.1 on 7 and 38 DF, p-value = 2.2 × 10−16
Note: ** p < 0.01, *** p < 0.001.
Table
Table 4. Causality testing for VAR(2).
Table 4. Causality testing for VAR(2).
Granger CausalityInstant Causality
F-Testp-ValueChi-Squaredp-Value
RSV → SHM10.1780.0001212 ***0.0084690.9267
SHM → RSV1.95890.14810.0084690.9267
Note: *** p < 0.001.
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Share and Cite

MDPI and ACS Style

Bełej, M. Does Google Trends Show the Strength of Social Interest as a Predictor of Housing Price Dynamics? Sustainability 2022, 14, 5601. https://doi.org/10.3390/su14095601

AMA Style

Bełej M. Does Google Trends Show the Strength of Social Interest as a Predictor of Housing Price Dynamics? Sustainability. 2022; 14(9):5601. https://doi.org/10.3390/su14095601

Chicago/Turabian Style

Bełej, Mirosław. 2022. "Does Google Trends Show the Strength of Social Interest as a Predictor of Housing Price Dynamics?" Sustainability 14, no. 9: 5601. https://doi.org/10.3390/su14095601

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop