# RECORD 009443a37464c7bfc0c1892b4c485e939d19af7c
# TITLE: Testing Housing Price Drivers in Santiago de Chile: A Hedonic Price Approach
# YEAR: 2023.0  VENUE: Critical Housing Analysis  DOI: 10.13060/23362839.2023.10.2.558
# AUTHORS: Francisco Vergara-Perucich
# PAGES: 14  CANDIDATE TABLE PAGES (1-based): [7]
# ABSTRACT: Hedonic modelling is an empirical analysis technique that is widely used to review the housing market and unpack the main determinants of price. A correct determination of housing price drivers allows a weighted prediction of the value of each dwelling for sale or rent. This paper reviews the fundamental determinants of housing prices that have been published in the literature to see which ones have a better predictive fit for the case of the housing market in the city of Santiago, Chile. From a record of 456,000 property transactions, a dataset composed of 11 explanatory variables is elaborated. Through a semi-logarithmic regression, 4 variables that contribute to explaining the formation of housing prices in Santiago, Chile, are identified. The results indicate that the socioeconomic characterisation of urban areas where housing is traded, the price of copper on the London Metal Exchange, the mortgage interest rate, the age of residential buildings are the main drivers for the main drivers of prices in Santiago. Given the crisis of access to housing in Chile, the article provides relevant information for decision-makers in housing policy.


===== PAGE 1 (score 4, numeric-paren 0, table-words 0, var-words 4) =====

Volume 10 | Issue 2 | 2023 | 44-57 
Available online at www.housing-critical.com 
https://doi.org/10.13060/23362839.2023.10.2.558 
 
44 
 
Testing Housing Price Drivers in Santiago de Chile: A 
Hedonic Price Approach 
 
José Francisco Vergara-Perucich 
 
Universidad de Las Américas, Chile 
jvergara@udla.cl 
https://orcid.org/0000-0002-1930-4691 
 
 
Abstract: Hedonic modelling is an empirical analysis technique that is widely used to review the housing market 
and unpack the main determinants of price. A correct determination of housing price drivers allows a weighted 
prediction of the value of each dwelling for sale or rent. This paper reviews the fundamental determinants of 
housing prices that have been published in the literature to see which ones have a better predictive fit for the case 
of the housing market in the city of Santiago, Chile. From a record of 456,000 property transactions, a dataset 
composed of 11 explanatory variables is elaborated. Through a semi-logarithmic regression, 4 variables that 
contribute to explaining the formation of housing prices in Santiago, Chile, are identified. The results indicate that 
the socioeconomic characterisation of urban areas where housing is traded, the price of copper on the London 
Metal Exchange, the mortgage interest rate, the age of residential buildings are the main drivers for the main 
drivers of prices in Santiago. Given the crisis of access to housing in Chile, the article provides relevant 
information for decision-makers in housing policy. 
 
Keywords: hedonic economics; housing in developing countries; urban policy and planning; land markets and 
housing policy. 
 
 


===== PAGE 7 (score 18, numeric-paren 0, table-words 5, var-words 2) =====

Volume 10 | Issue 2 | 2023 | 44-57 
Available online at www.housing-critical.com 
https://doi.org/10.13060/23362839.2023.10.2.558 
 
50 
 
of various variables on housing prices in Greater Santiago, rather than focusing solely on the 
predictive capabilities of the hedonic price model. 
 
 
Results 
 
During the period under investigation, a total of 456,133 official transactions were analysed. 
Among these transactions, 44.5% were related to house transactions, 47% were apartment 
transactions, and 1.5% were land transactions. It is important to note that a significant number 
of the houses transacted were likely demolished to make way for future development. 
 
Table 3 provides a comprehensive summary of descriptive statistics for each variable. The 
average price per square metre of housing during the study period was found to be USD 1,882, 
while the price of land was USD 369. Additionally, Table 2 illustrates that the average size of 
a housing unit was 115.92 m2. This figure may seem high, but it is important to consider that 
the value is influenced by the initial phase of the analysis, during which the production of houses 
closely matched the sale of apartments. 
 
Overall, the data highlight the diverse nature of the transactions and shed light on the prices and 
sizes associated with different types of properties. 
 
Table 3: Descriptive statistics of the variables studied 
 
Variables 
Mean 
Median 
Std. dev. 
Skewness 
Kurtosis 
Ln housing prices 
3.887 
3.895 
0.240 
0.143 
-0.559 
IDSE 
0.762 
0.758 
0.015 
0.863 
-0.170 
m2 
115.920 
112.740 
21.271 
2.186 
8.970 
Sales 
3366.100 
3873.000 
1582.300 
-0.444 
-1.148 
Age of residential building 
50.710 
49.540 
9.270 
0.519 
-0.081 
Construction cost index 
4779.800 
4889.900 
632.100 
-0.243 
-1.313 
Wage index 
5.779 
5.846 
1.152 
-0.406 
1.076 
Population 
5361100 
5322700 
179700 
0.756 
-0.288 
Mortgage interest rate 
4.071 
4.140 
0.545 
0.720 
1.069 
Ln land price 
2.270 
2.243 
0.174 
0.457 
0.415 
Ln copper price on the 
London Metal Exchange 
1.077 
1.128 
0.233 
-0.807 
0.717 
Ln money supply 
8.338 
8.393 
0.325 
-0.360 
-1.192 
Blue-chip index (IPSA) 
4135.100 
4055.500 
745.030 
-0.092 
0.128 
Source: Author based on sources indicated in Table 1 and Table 2. 
 
Most of the values of skewness and kurtosis fall within the range of 2 and -2 (George and 
Mallery 2010). Among the values analysed, only the average square metre of a property (m2) 
exhibits high kurtosis, while most other variables demonstrate normality. Figure 1 displays the 
plots of the transformed data. The variables population, construction costs, and money supply 
exhibit a distinct deterministic trend, which was excluded from the regression model using the 
stepwise technique owing to the potential presence of a spurious relationship with the dependent 
variable. 
 
Figure 1: Variable plots 


===== PAGE 8 (score 2, numeric-paren 0, table-words 1, var-words 0) =====

Volume 10 | Issue 2 | 2023 | 44-57 
Available online at www.housing-critical.com 
https://doi.org/10.13060/23362839.2023.10.2.558 
 
51 
 
 
 
 
 
Source: Author’s calculations. 
 
Table 4 presents the estimations obtained through multiple linear regression. The model 
successfully explains 60.74% of the variance in the natural logarithm of the housing price, 
providing a statistically reliable fit for predictive scenarios. Importantly, there is no evidence 
of collinearity among the relevant variables. Additionally, the White test for heteroskedasticity