Business Forecasting Spss - Eviews Project

Business Forecasting SPSS - Eviews Project

Word Count: 1822

(Excluding References and Appendix)

Introduction

The objective of this project is to analyze the relationship between Mix-adjusted Average House Prices in United Kingdom and London, using fundamental approach of forecasting, producing Regression analysis. The data used for this project has regular monthly frequency, with February 2002 as a start date, and December 2015 as the end date.  

I have decided to analyze this data, as I find it interesting. This is mostly due to very significant difference in prices between whole UK and the area of London. Such a substantial difference is not very common among other countries and their capitals.

Literature Review

There is lot of different variables that can help forecast/predict the house prices in the United Kingdom.

To begin with, inflation has definitely impact on the house prices. However, house prices have outstripped inflation significantly, over passed 20 years in the United Kingdom.  Inflation has increased by 73%, but some of the regions in the United Kingdom have seen increase of more than 1,000% in the house prices, over this period of time.

As the supply of houses is relatively inelastic in the short run, as well as it is hard to obtain planning permits in the United Kingdom, it can be considered quite inelastic in the long run too. Thus it can be argued that, there is a significant shortage of housing, which explains why house prices have risen much faster than inflation and earnings.

This can be caused by other economic factors; such are rise in the GDP. Annual GDP growth causes the real disposable income to rise as well. Thus, giving more opportunities to buyers. And only naturally as the demand increases the price increases as well. In addition to this, as the real disposable income increases and the interest rates are continuously low, people are more willing to take mortgages.

Mortgages increase the house prices as well, after the deregulation of the banking sector, the industry has became more competitive, and as a consequence offering more mortgage products, thus increasing the prices even further. However, number of these products disappeared during the 2008, crisis due to shortage of finance in the market.

Furthermore, unemployment should be considered as significant factor affecting house price too, as with higher employment, the consumer confidence rises too, thus buyers are more willing to invest/spend.

Apart from economic factors, variables such as demographics or migration should be taken into serious consideration too. For example, increases in life expectancy, increase in divorce rates, or less marriages as well as children leaving homes earlier, are all factors increasing demand for the houses, therefore the price.

The percentage of foreigners living in United Kingdom is increasing from year to year, at the moment it is around 8 million. This significantly influences the demand for housing too.

And lastly, some observers have presented evidence that suggest house prices in United Kingdom over the past decade exhibited bubble behavior - that is, by the definition of a bubble, house prices may have risen to higher levels than could be justified by fundamentals. Others, however, disagree, and claim there is little credible evidence that home prices rose faster than could be justified by underlying determinants over the past decade.

Methods

As mentioned in the introduction, using fundamental approach of foresting, Regression analysis will be produced. These are the methods used to produce this analysis:

1. Bivariate Correlation

The first step, when constructing linear regression model is to check whether there is correlation between dependent and independent variable. As, if there was no correlation between the variables, there would be no point of trying to construct the Linear Least Squares Regression model.

Pearson’s product moment correlation coefficient is a measure of linearity; therefore it will be used in producing this analysis. If the Pearsonian coefficient will be close the 1, it indicates significant relationship, if it will be close to 0, it indicates very small or no relationship, thus absence of linearity.

1. Linear least squares regression for bivariate data

If there is adequate linearity present between the variables, it is possible to find the linear least squares regression line. This line minimizes the sum of squared errors, or the difference between the observed and predicted values of the dependent variable.

1. Assumptions underlying linear least squares regression

There are series of assumptions that underline the least squares regression model; most of these assumptions refer to errors/residuals from the regression analysis. Diagnostic tests need to be done, in order to confirm that the regression model is statistically sound.

• Residuals should be normally distributed.
• Residuals should have constant variance over time (homoscedasticity).
• Residuals should be uncorrelated between observations (absence of auto-correlation)
  

Findings and Data Analysis

Data have been exported from Office of National Statistics, for Great Britain in order to build this regression model. As mentioned in the introduction, the data has regular monthly frequency and the data range is from February 2002 to December 2015.

Dependent variable (y) represents mix-adjusted average house prices in the United Kingdom and the independent variable (x) represents mix-adjusted average house prices in London.

(Mix Adjustment divides the property market into 'cells'. Each cell contains properties with similar characteristics in similar locations. House price data can then be allocated to the different cells in order to assess the average price of each cell, i.e. of properties with different characteristics. After the mean values have been weighted, the average cell price is the mix-adjusted price.

The Office of National Statistics house price index, for example, creates as many as 100,000 different cells, providing a very detailed price analysis for different property types.

This method solves the problem of the change in property sales composition in a similar way to the Hedonic Regression model. There would be a cell for detached houses as well as one for terraced ones. This way, an increasing number of sales of detached houses would not influence the mean price of each cell. However, differences in weighting can have a substantial effect on the outcome of the analysis.)^[1]

1. Bivariate Correlation

Bivariate Correlation procedure can be obtained in SPSS, by commands “Analyze > Correlation > Bivariate”, Pearson model is the default, as is the two-tailed test.

It can be easily evaluated from the table^[2] produced by SPSS, that there is significant positive correlation (.978) between the variables.

Indicating that if the house prices in London increase, there will be increase in the house prices in the whole UK too.

This can be explained by the fact, that house prices in London are included in the overall house prices for the UK.

1. Multiple Linear Regression Model

Next step is running multiple linear regression via SPSS by “Analyze > Regression > Linear”. Moreover, “stepwise” method has been chosen, to only consider the variable, which is capable of increasing the accuracy.

Model Summary^[3] shows high coefficient of determination (R2 = .956), this is the first sign of a good prediction model, as it indicates that 95.6% of the variation is captured by this model.

Coefficients table^[4] represents values of the estimated linear least squares regression line parameters, or the coefficient and the constant. In this case, it illustrates that if house prices in London increase by one unit, the house prices for rest of the United Kingdom increase by 0.425 units.

1. Jarque-Bera test for normality

This is the first test conducted in Eviews, with hypothesis:

H0: residuals are normally distributed

H1: residuals are not normally distributed

The values in the graph^[5] don’t seem to be normally distributed. Further, hypothesis null is rejected, as probability = 0.000001 < 0.05. Thus, the residuals are not normally distributed.

This is due to the fact, that there is lot of different independent variables that might affect the house prices in UK. These include, for example, economic indicators such as, inflation or GDP growth.