Hodjatollah Mirzaei; Narges Razban; Teymor Mohamadi; Habib Morovat
Abstract
Housing price shocks of one region may spread to the housing market of neighboring regions or geographical areas bounded by political border and lead to the formation of price shocks in shock-receiving areas. The housing policies may not be effective when implemented regionally and separately ...
Read More
Housing price shocks of one region may spread to the housing market of neighboring regions or geographical areas bounded by political border and lead to the formation of price shocks in shock-receiving areas. The housing policies may not be effective when implemented regionally and separately if there is a confirmed network connection between the housing markets of regions. Price shocks to a housing market spreads with a delayto interconnected housing markets, ultimately resulting in the diffusion of the price shock across the entire of the housing network. This researchaims to investigate the housing network between selected cities (centers of the country's provinces) using the VAR model and Forecast Error Variance Decomposition (FEVD). The results of this research confirm the existence of a network connection between the housing markets of the country's provinces, and unlike previous studies, the results show that it is not only the city of Tehran that spreads price shocks to other regions, but also cities such as Karaj, Shiraz, and Arak spread price shocks to other cities. In addition, the results suggest that the recent price jump, since 2019 has significantly increased the density of the housing network in the country. Based on this, price shocks are expected to be distributed more quickly throughout the country.
Introduction
In addition to the fact that economic characteristics, macroeconomic policies, and external factors affect housing prices, housing price shocks in neighboring geographical areas also spread to housing prices in each region and can lead to the formation of price changes in the price-accepting region. Therefore, it is essential to investigate the network connection between the housing markets of the geographical regions within a country. This research aims to explore the network connections and dynamics between housing markets in provincial centers, as well as the relationships between all pairs of centers to form a comprehensive housing market network for the country. Specifically, the study seeks to identify: (a) the centers of the provinces whose housing price disturbances are most contagious to other provinces and (b) the centers of the provinces that are most affected by the housing price disturbances of other provinces should be identified.
Methods and Material
The study utilized data from the Statistics Center spanning period from 2009 to 2011.
The research methodology employed the vector autoregression (VAR) model. To address the identification problem in the model, the centers of the provinces were classified into four groups:
1: Tehran, Alborz, Mazandaran, Isfahan, Gilan, Khorasan-Razavi, Qom, Qazvin and East Azerbaijan.
2: Fars, Khuzestan, Golestan, Hormozgan, Bushehr, Zanjan and Hamedan.
3: Semnan, Yazd, Lorestan, North Khorasan, Kerman, South Khorasan, Kohgiluyeh and Boir Ahmad, Markazi and Kurdistan.
4: West Azerbaijan, Ardabil, Ilam, Kermanshah, Sistan and Baluchistan, and Chaharmahal and Bakhtiari.
In the network connection approach proposed Diebold and Yilmaz (2014), the vector autoregression model or VAR has been used.
In a country with three geographical regions A, B and C:
(1)
The VAR system of equations has three equations for housing prices in areas A, B, and C. The housing price in each region such as A at the current time (t) is a function of the price of the same region in previous periods ( ), and the price of other regions in previous periods ( and ) (k=1, 2, .. K). The effectiveness of the price of region A from the price of the same region and regions B and C in the previous periods are measured by β-11k, β-21k, and β-31k coefficients, respectively. The number of optimal breaks in equation (1) is determined by the Schwartz criterion.
In order to form a network connection and to check the amount of shock propagation from region i to j, variance analysis of prediction error is used. In this regard, Diebold and Yilmaz (2014) introduced four indicators:
(1)Shock received from others: The shock received by each region from other regions
FC=
(2)Shock sent to others: The shock sent by each region to other regions
OC=
(3)Total connections per network: average total shock per region
TC=
(4)Net communication or NC: the net shock sent by any region to other regions
NC=
Correlation between regions based on variance analysis
Shock received from other areas
areas
Shock sent to other areas
Results and Discussion
The reliability test of Becker et al. (2007) was conducted for all provinces, which was found to be significant in all cases.
Based on the results of VAR model and variance analysis:
First group: Isfahan and Qom are the biggest receivers and Mashhad is the weakest recipient. Karaj and Tehran are the biggest senders of shocks and Qom and Isfahan are the weakest senders.
Second group: Gorgan and Hamadan are the most important and Bushehr is the weakest recipient. Shiraz and Zanjan are the most important shock transmitters , while Bandar Abbas and Bushehr are the weakest.
Third group: Semnan and Sanandaj are the most important shock receiver, and Bojnord is the weakest receiver; Arak and Yazd are the most important sender of shocks and Sanandaj is the weakest sender .
Fourth group: Ardabil and Kermanshah are the most important senders and receivers of price shocks, respectively. The calculation of the total communication index in the housing network shows that the first group has the densest and the fourth group has the thinnest housing network.
In order to investigate the evolution of the housing network (changes in density over time), the Galtan's regression logic was used indicating an increase in the density of the housing network in the centers of the provinces over time.
Conclusion
The dynamics of real housing price changes demonstrate three distinct patterns.. During the years 2009 to 2012, the price of most centers decreased and remained almost constant from 2012 to 2018, and then all the centers had a sharp price increase. As a result:
(1) Karaj, Tehran, Shiraz, Arak and Ardabil sent the most price shocks;
(2) Isfahan, Gorgan, Semnan and Kermanshah received the most price shocks,
(3) the strongest housing network was observed between the cities of Mashhad, Sari, Qom and Tabriz, Isfahan, Karaj, Tehran, Qazvin and Rasht and
(4) the housing network among the provincial centers has increased during the years (2009 to 2010).
In times when the living conditions in the cities that are significant senders the shock become difficult, other cities within the network have become centers of population attraction and can change their roles. Consequently, it is advisable for housing market policies to consider the mutual influence between city prices. By doing so, when market price jumps occur, the extent of shock transmission from these driving centers can be somewhat controlled, thereby mitigating market excitement and excessive growth in prices. As an example, the policies on the supply side can be such that the supply in the shock-sending areas is facilitated. Preventive measures such as supporting the supply of semi-finished units, facilitating the conditions for issuing permits, facilitating access to construction loans, activating pre-sale tools, etc., should be adopted in leading areas so that when price jumps occur, shocks will be sent from these regions to other regions at a slower rate. it is advisable for housing market policies to consider the mutual influence between city prices. By doing so, when market price jumps occur, the extent of shock transmission from these driving centers can be somewhat controlled, thereby mitigating market excitement and excessive growth in prices.
Acknowledgment
In the end, we would like to express our gratitude to Dr. Nasser Khiabani, Dr. Ali Nasiri-Aghdam, Dr. Mirhossein Mousavi, and Dr. Taleblo, for their invaluable contributions to this paper.
