The Global Housing Prices Index for 2021

Introduction 

The housing price index is a leading indicator of the economic activities of the worlds. It has successfully predicted eight out of ten recessions that have taken place since 1945. Housing prices play a critical role in determining the phase of the business cycle. As the economy boom, the construction and housing industry also expands in response to the increased demand and therefore pushing the house prices upwards. In a recession, a decline in private income leads to a drop in the aggregate demand and house prices. The fall in the house prices is slower because the house owners are not willing to lower their costs and therefore, most of the adjustments are achieved from a decline in the volume of sales leading to a decline in the construction sector and house based employment. In contraction and recessions, the actual house prices fall rapidly as the general trends in inflation reduce the house prices including those with sticky nominal prices. House prices can be used to forecast output as it represents a sizeable portion of GDP. Similarly, since house prices represent a significant part of the combined wealth of an economy, any fluctuations can indicate the evolution of GDP and future direction of inflation. Accurate forecasting of housing prices is an essential tool to the monetary policy authorities and market participants (Plakandaras et al., 2014). 

Background 

The IMF global house price index for fifty seven countries has continued to rise since the 2008 recession. Data from the IMF indicates that the price index experienced a steady growth from 2000 before facing a decline in 2008 which lasted until 2011. From 2012, there has been s steady increase on each quarter indicating a positive year on year growth. According to the IMF, the rise in the house prices is not uniform in all countries. The fund categorizes its nations into three clusters depending on the level of growth in their respective housing index. The three include gloom, bust and boom and boom. Similarly, the prices do not increase throughout the individual countries, and there are differences in the national and city level. The Global housing watch follows the development in the housing sector around the world on a quarterly basis providing current data on the house prices and the measures used for valuation of housing markets like house price to income and price to rent ratios. Different houses have set up unique hedonic characteristics including the number of bedrooms, bathrooms, squared footage, location, proximity to school modernized, conditions and availability of garage. Additionally, properties are of different types including cooperative unit, condominium, and single family house. The characteristics of the units sold vary from time to time leading to a change in the representation of the sample population set that determines price increase and decreases. Keeping a standard population sample set will ensure that the real appreciation or depreciation is reported by ensuring that sample different sample set is not compared therefore resulting to misleading information. Using a simple moving average to determine the house prices will provide misleading information as a high end and a low-end house in two periods would indicate a decline in the price if changes in sales price are used as a measure of the price movement. However, the reality from the comparison can be that the prices might have increased if the sample set remained constant. This paper tries to address the above issue by using Hedonic regression and repeat sales method (IMF, 2017). 

Data 

IMF follows the development of housing markets in selected countries around the world on a quarterly basis to provide current data on the housing prices and other metrics for assessing the value of the housing market. The fund offers different global house price indexes that include; real house price index, house prices around the world, credit growth across the globe, house price to income ration and house price to rent ratio (IMF, 2017). 

Global price index 

It is an average of the actual prices in different countries. According to Knight Frank, the average global housing prices continue to rise since 2012. More nations have recorded increased growth over the years after the 2008 recession. Prices around the world have rebounded back in the past four years. However, the increase is different from one location to the other. The price increase is unusual for each country and therefore, the IMF has clustered the prices into three categories; gloom, bust, and gloom, and boom. Gloom includes countries in which the house prices substantially fell following the recession and have not recovered. Bust and boom include those states where the housing prices have rebounded after 2013 following a downward spiral from 2007 to 2012. The boom consists of those countries that had a modest decline in the housing prices in the same period and quickly recovered after the recession. The house prices different within each cluster and nation (IMF, 2017). 

Countries in the gloom category include China, Netherlands, Singapore, Slovenia, Brazil, Russia, Spain, Finland, France, Macedonia, Singapore, Cyprus, Greece, and Croatia. The bust and boom countries include the UK, US, Hungary, New Zealand, Ireland, Denmark, Japan, South Africa, Estonia, Latvia, Indonesia, Portugal, and Iceland. Countries in the boom category include Norway, Canada, Belgium, Chile, Malaysia, Sweden, Switzerland, Hong Kong, Israel, Australia, Mexico, Korea, and India, Slovakian Republic Colombia, Austria and Taiwan province of China. The figure below shows the global housing price index computed quarterly by the IMF from 2007 to 2017. From the figure, the global prices increased from 2000 to its peak in 2007 where the price drastically declined following the global recession that occurred from 2007 to 2012. It is evident that there was a sharp decline in the housing prices in 2007 and 2008 (IMF, 2017). 

Source; (Imf, 2017). 

Empirical framework 

Several methods are used to construct house price indexes. The different approaches can be classified into two groups; simple and econometric techniques. Procedures in the simple category include simple and weighted averages and the median. Econometrics include the hedonic method and repeat sales (Cullen, 2009). 

Hedonic regression 

The different characteristics in the selected samples are controlled statistically, and the selling prices are regressed on different variables that are unique to each unit including the area, number of rooms, lot size, interior space, condition and the status of construction. In this case, the price indexes are determined in many ways by employing the sensitivity of the coefficients from the regression. The estimated equation for each quarter is used to predict the value of a unit in every period. The characteristics of the houses remain constant in the period under estimation. Another approach includes the introduction of dummy variables into an equation to help in capturing changes in the price over time while ensuring that the characteristics are controlled. In the second alternative, the costs for the unique features are not allowed to vary over time unlike in the first approach, individual components change over time, and they also get their prices. For this technique, a substantial volume of data for the units sold and their unique characteristics is employed (Cullen, 2009; Garg, 2016). 

Repeat sales method 

The data is constrained to the units sold for two or more times the period under estimation. It isolates the actual increase in housing prices without calling for detailed information on the characteristics of each unit. By controlling the number of the repeat sales, an accurate measure of housing price change can be obtained. In this approach, the quality is not measured by it is required to remain the same over time. In this method, only a small size of the sales data is used due to the limitation of using units that have been sold severally. Repeat sales are widely used in commercial and research studies but have received criticism that only offers relevant information on a type of housing unit and might not be applicable to the rest of the market (Cullen, 2009; Garg, 2016). 

Hybrid model 

A model that combines hedonic regression and repeat sales avails all sales transactions for use and identified issues in repeat sales can be addressed. Another approach involves assessed value method in which modified repeat sales are constructed by including the values of the houses as a variable combined with the sales. Another variant involves allowing the hedonic coefficients to vary with time. We can use the relationship between the different variables to determine the global housing price index. Using weighted values for each quarter and the date as an independent variable, we can evaluate a regression equation as follows (Garg, 2016). 

Weighted values =ᵝ 0 + ᵝ 1 quarter 1-4 2000 + ---- ᵝ 1 quarter 3 2017 

Results 

The summary statistics for the data are as follows; 

Mean	136.51	Maximum	159.11
Median	145.14	Minimum	99.390
Standard deviation	19.437	Skewness	-0.81283
C.V	0.14238	Ex Kurtosis	-0.81462
5% percentile	100.17	Inter-Quartile range	30.316
95%	158.13

The table below shows the values obtained from the analysis of the quarterly global housing price index from the IMF. The date is used as the independent variable, and the dependent variable is equally weighted. A total of 69 observations were made, and the following output was obtained from Gretl. 

Model 2: OLS, using observations 1-69 

Dependent variable: equally weighted 

	Coefficient	Std. Error	t-ratio	p-value
const	108.242	2.63178	41.13	<0.0001	***
dateq	0.807778	0.0653535	12.36	<0.0001	***

Mean dependent var	136.5139		S.D. dependent var	19.43742
Sum squared resid	7832.252		S.E. of regression	10.81200
R-squared	0.695140		Adjusted R-squared	0.690590
F(1, 67)	152.7730		P-value(F)	6.04e-19
Log-likelihood	− 261.1573		Akaike criterion	526.3145
Schwarz criterion	530.7828		Hannan-Quinn	528.0872

From the output, the coefficient for the constant value was 108.242, and for the date, it was 0.807778. The standard error for the constant was 2.63178 and 0.0653535 for the period and shows an estimate of the interval that the mean will fall. For this calculation, it was multiplied by 1,96, therefore, enabling the estimation of where 95% of the data will fall. The t-ratio is 41.13 and 12.36 for the constant and the date implying that the variable is statistically significant. The P value was less than o.0001 for the two variables. It indicates that there is a strong relationship between the variables if the null hypothesis rejected the existence of any such relationship. The mean of the dependent variable is 136.5139, and its standard deviation is 19.43742. The sum square residue is 7832.252 while the standard error of regression is 10.81200. The R square is 0.695140, and the adjusted R square is 0.690590. The F (1,67) is 152.7730, and the P-value (F) is 6.04. 

The standard error computed above s 

The following image shows the movement of the housing prices over the years under consideration. Note, the X shows the total number of quarters from 2000 to 2017. 

The above image shows a normal distribution for the global house price index from the data. The prices are typically distributed around the mean of 136.51. 

The image above was obtained from forecasting the values from the IMF data using a confidence interval of 95% t (67, 0.025) = 1.996 

Conclusion 

Data on global housing price index was obtained from the IMF and used to calculate the different statistics that helped to determine the accuracy of the models used to compute the housing prices data. From the regression analysis, there is a strong relationship between the variables and therefore the two can be used for prediction purposes. From the results, it is possible to use historical data from the housing index to determine future house prices. The statistics show that most of the data falls within a 95% confidence interval implying that it the quarterly unit prices falls within the second standard deviation. The data indicates that there are no outliers and therefore can be used to make predictions of economic activities and the levels in which businesses are in the business cycle. 

References 

Cullen, I. (2009). Constructing a Global Real Estate Investment Index.  Global Trends in Real Estate Finance, 100-113. doi:10.1002/9781444315301.ch7 

Garg, A. (2016). Statistical Methods for Estimating House Price Index.  Journal of Business & Financial Affairs,05 (04). doi:10.4172/2167-0234.1000231 

Imf. (2017). IMF Global Housing Watch. Retrieved November 13, 2017, from http://www.bing.com/cr?IG=0C4AB577456A4880ABC19E7ACB5BC68D&CID=1F5FC35D9A276BE738B2C8679B216AF5&rd=1&h=N4_ph23IIclgPyffLv2zX_CjUCRjcP0MJzsKUQaHxfs&v=1&r=http%3a%2f%2fwww.imf.org%2fexternal%2fresearch%2fhousing%2f&p=DevEx,5067.1 

Plakandaras, V., Gupta, R., Gogas, P., & Papadimitriou, T. (2014). Forecasting the U.S. Real House Price Index.  SSRN Electronic Journal . doi:10.2139/ssrn.2431627