4.4 The main housing bubble detection methods
4.4.1 Descriptive approach
The descriptive statistic and ratio approach has been extensively used by Case and Shiller (2003), Baker (2002), Haines and Rosen (2007), Chung and Kim (2004) Hlaváček and Komárek (2009), Cadil (2009), Weeken (2004), Dolphin and Griffith (2011), Hou (2009) and Shen et al. (2005). Most of these studies used the descriptive approach along with other methods whereas Dolphin and Griffith (2011) showed evidence of bubbles utilising a mostly descriptive approach. Currently, application of the descriptive approach to housing bubble detection involves the graphical analysis of a single or several indicators and might also involve the following: i) descriptive statistics (i.e mean, standard deviation, maximum and minimum values); ii) correlation analysis; and iii) supportive analytical tools such control charts. The study by Dolphin and Griffith (2011) is an ex-post analysis for the detection of bubbles. The descriptive analysis applied in this study was limited to searching for unusual performances on the basis of the graphical trajectory of house prices, consumer spending growth vs. house prices, affordability indices, lending, rental indices, and house building completions. The main course of action used by Dolphin and Griffith (2011) was an analysis of the relationships between house prices and the aforementioned variables. The authors conclude that some metrics can be used on a descriptive level to indicate when a housing bubble is forming. Weeken (2004) applied a partly descriptive approach in conjunction with an asset pricing fundamental value approach to assess whether house prices were unsustainable. In the descriptive analysis part, historical series of house price to earnings ratio (HPE) were used. The results suggested that the HPE ratio was well above its long- term average. In the past, such a situation has coincided with periods in which house prices have fallen. However, Weeken concludes that data associating house prices to rental levels is
subject to several limitations. Similarly, Haines and Rosen (2007) combined a regression- based fundamental approach with descriptive analysis. On a descriptive level, the study focused on average house prices, the mortgage servicing index and mortgage rates. The analysis of these variables was primarily aimed at assessing the affordability levels of median households. The authors claim that mortgage payments and affordability indicators are essential variables that determine the prices households are willing to pay to purchase a house. Moreover, they used a descriptive statistics analysis on house prices, affordability index, income, unemployment, population density, construction cost and median age of the population. Shen et al. (2005) also employed a partly descriptive approach, utilising the descriptive statistics of house prices, disposable income, and vacancy rates. Hou’s (2009) study used a combination of different quantitative indicators. Initially, he compared historic house prices with per capita disposable income growth. He continued by examining characteristics of housing demand and supply. From the demand side, Hou (2009) analysed the population structure, while from the supply side he compared houses completed with houses sold. A comparative analysis looking at house prices, gross domestic product and per capita disposable income was also done. The rich analysis in Hou’s study was followed by an examination of the level of mortgage loans, price-to-income ratio and rent. Moreover, the statistical tool of a control chart was introduced to quantify housing bubbles. He applied the control chart to house prices to assess whether there was any unusual performance with regards to the rate of change of house prices. His results suggest that control charts provide a promising way to analyse housing bubbles and that simultaneous peak points for each fundamental indicator could be interpreted as evidence of a bubble. Case and Shiller (2003) also analysed the descriptive statistics of house-price-to-income ratio and house prices. They also presented descriptive regression results along with a graphical analysis of income per capita, expected rate of inflation and mortgage rates. Interestingly, Case and Shiller (1989) provided an analysis of the correlations between changes in house prices and other variables. Garg et al. (2009) argue that the main determinants of house price on a demand level are population size and growth, income, price of housing, cost and availability of credit, consumer preferences, investor preferences and price of substitutes. The quantity of new supply is determined by the cost of inputs (i.e. land, labour and building materials), the price of the existing stock of houses and the available technology for production. They further claim that existing home sales, housing starts, housing affordability, mortgage rates, GDP, and unemployment rate are important for analysing bubbly house price levels. Kuenzel and Bjornbak (2008) used an in-depth descriptive analysis to assess UK housing booms. Their
analysis involved graphical/descriptive analysis of demand-side and supply-side indicators. On the demand side, the variables included real average house prices, population growth rates, household size and type, real disposable income, mortgage rates, buy-to-let yields and house-price-to-rent ratio. On the supply side, the list included housing stocks, changes in housing capacity and vacancy rates. Having examined the potential drivers of house price growth, the study moved towards integrating these into an overall identification of housing market equilibrium. The study concluded that a downward adjustment of UK house prices could continue, as house prices were still above levels that could be judged as consistent with housing market equilibrium (long-term average). Cadil (2009) used a partly descriptive approach focusing on real house price growth and house price to earning ratio to assess bubbles in the Czech housing market. After applying a descriptive analysis of house prices and house-price-to-earnings ratio, he acknowledged that neither indicator provided any clear evidence of a housing price bubble. However, he concluded that basic indicators used for housing bubble identification are somewhat alarming. Chung and Kim (2004) used three estimating methods to assess a housing bubble in three Chinese regions. One of these approaches, a descriptive approach, is based on the assumption that bubbles are the variation in several macro-economic variables such as Gross Domestic Product and Consumer Price Index. The method can be summarised as follows:
Where is the amount of bubble, is the rate of change in actual prices, and is
the rate of change in GDP. The data for this approach cover the periods of the first quarter of 1997 and the first quarter of 2002. Relying on this approach, Chung and Kim (2004) found that the results appeared much more pronounced than what one might have expected and that all the three cities showed evidence of a price bubble.
Hlaváček and Komárek (2009) used a holistic descriptive analysis of several variables (along with other statistical methods) to assess the overall variability of the Czech housing market. The variables were distinguished on the basis of demand and supply. On the supply side, they included building plot prices, construction prices, and housing stock. On the demand side were marriages, divorces, natural population growth, net migration, unemployment rate, economic activity rate of the population, vacancies and labour force, average monthly wage,
Bt =ΔPt− ΔGDPt Bt(%)= Bt ΔPt =% Share of Bt Bt ΔPt ΔGDPt
rent per month and loans. Moreover, the study applied a graphical descriptive analysis of rental returns and price-to-income ratio. The descriptive analysis concluded that there were price bubbles in 2002/2003 and 2007/2008.