An Empirical Analysis of House Price Bubble: Research Methods

An Empirical Analysis of House Price Bubble: Research Methods There is no consensus as to which method is the best to estimate house price bubbles. Some researchers have used a ratio approach; some have relied on a user cost approach (asset-market approach) model (see Levin & Wright, 1997), whilst still others have used the Vector error correction model (VECM). Thus, no single method seems to have universal approval for investigating the phenomenon. Direct marketing

Asset prices are determined by both demand and supply factors. Levin and Wright suggest the most common demand factors used to study house prices are income, inflation and interest rate. Many researchers also include construction cost as an important component of the supply side variables in studying house prices. Meen successfully used construction cost, interest rate, income, inflation and after-tax interest rate in his empirical model to explain the mortgage rationing impact on U.K. housing market in a mortgage rationing period (1978 to 1980) as compared to a period when mortgage rationing was absent (1981 to 1987). Malpezzi et al. also identified construction cost as a major determinant of house prices in their study on house price index determination for 272 U.S. metropolitan areas. Similarly, Case and Shiller treated construction cost as an important element in studying house price bubbles in four states in the U.S.
When studying the house price in the short run, researchers tend to ignore the impact of supply factors on house price dynamics because of the assumption that house supply does not move in a short period. For example, using quarterly data, Black, Fraser & Hoesli test the actual house prices relative to the house value in U.K., using only demand factors (income, inflation and interest rate). Some researchers use only inflation to capture the cost of supply (for example Coleman et al., 2008). However, the cost of supply in China’s housing market changes over time; the inflation factor alone cannot capture the supply factors perfectly. Inflation is calculated by the changes in a standard basket of goods that cannot adequately estimate the supply costs (such as material and labor costs) dynamic. The present study employs both demand factors (income, inflation and interest rate) and supply factors to capture house price movement in Beijing.
In order to estimate both the long term trend and short term dynamics of house prices in Beijing, this study adopts Coleman et al.’s model based on the VECM. The model consists of the housing demand and housing supply equations, reproduced below.Equation 3 examines the long term trend and short-run dynamics of Beijing house prices. Yearly data from 1998 to 2010 was used to investigate the long-run trend and quarterly data from 2005:Q1 to 2010:Q4 to investigate short-run term dynamics of the house price in Beijing.
The Beijing house price index was utilised to measure the change in house prices, Beijing GDP as the income variable, the consumer price index (CPI) as an inflation variable and construction cost as the cost of supply. These four series data sets were obtained from the statistics department of Beijing government. The interest rate variable was obtained from the Bank of China website.
Previous studies of house prices suggest that GDP is a good proxy measure of income. For example, Green tested the relationship between GDP and house price, finding that it was a good predictor of residential investment. Similarly, Gauger and Synder examined the relationship between residential investment, money supply, interest rate and GDP using a VECM model in both pre-regulation and post-regulation sub-periods. The authors found a positive correlation between residential investment and GDP. The GDP in Beijing maintains a high growth rate of about 10 per cent, which is approximately the same as the trend for house price growth. GDP also positively correlates with house price (Pillay & Rangel, 2005). Therefore, this study has used GDP as proxy for the income variable.