Table 1. Average house prices in pound sterling and the numbers of homeless households in the 33 different London Boroughs, in the last quarter of 2002. Red highlights are the highest of figures while blue ones are the lowest.
The following two charts are visualisations of the independent (x axis) and the dependent variables, named Chart 1. and Chart 2. Central London has the most expensive boroughs, Kensington and Chelsea, Westminster and
Chart 1. Pie Chart Visualisation of Independent Variable
Camden. In these three boroughs the uncommon high pricing of housing stock is due to the location. This may have to be taken into account when looking at regression analysis, as these boroughs average house costs are much higher than usual. The highest numbers of homeless are found however in Haringey, Newham and Brent, which are situated on the outskirts of London. Which would not be in line with the theory that expensive housing cost links with high number of homeless. This is what I was trying to find but looking at the pie charts, the same coloured slices looked very different in size.
Chart 2. Pie Chart Visualisation of Dependent Variable
Variation in Data
Once the data was ready for analysis I used descriptive statistics methods. And because I had different measurement values in my calculations (pound sterling in house prices and number of households in homelessness figures) I could not make direct comparisons between the 2 sets of data. So first I calculated the means and standard deviations for both sets of values to then be able to compare the Coefficient of Variation for each set. The results of these are below in Table 2. and 3, and show that there is considerably greater variation in the figures of homelessness.
Table 2. Descriptive Statistics: House Prices in the Last Quarter of 2002.
Table 3. Descriptive Statistics: Homelessness in the Last Quarter of 2002.
I also made the histograms for both house prices and homelessness figures to establish the skewness, if any. Chart 3. illustrates that house prices data distribution is negatively skewed. While homelessness values are unevenly distributed, tailed both to left and right and so has a much higher levels of skewness (Chart 4.). This is also apparent from the skewness value of 0.637 (Table 3.) that is much lower compared with the same value of 1.956.
Chart 3. Histogram of House Prices
Chart 4. Histogram of Homelessness
Null Hypothesis- Describing the Relationship- The Strength of Correlation
I went on trying to find a link between house prices and homelessness using inferential statistics. My null hypothesis was that there is no relationship between the levels of house prices and homelessness. If any, then the strength of this could be tested, and my hypothesis rejected. I felt it would probably be a “long shot ” finding any kind of a link, as the factors of my dependent variable were so complicated. In Table 4. the regression statistics results are displayed. The key statistical measure to assess the strength of a correlation relationship is called the Product Moment Correlation Coefficient (‘ r ‘), it takes into account the amount by which each value differs from the mean of its own distribution, the standard deviation of the two distributions, and the number of pairs of values. If r were zero, there would be no correlation at all. The closer the correlation coefficient gets to +1 or -1 the stronger the correlation. The closer it gets to zero the weaker it is. A very low figure of the Multiple R value is equivalent to Product Moment Correlation Coefficient, which suggest a very weak, almost negligible relationship between the 2 variables. R-square in the second row of Table 4. represents the proportion of the variation in homelessness that is being explained by variation in the independent variable, house prices. In other words approximately 0.00009% of homelessness is caused by higher house prices in certain areas of London. Again a very tiny amount. The P-value for the X Variable is quite the opposite; it is a very high figure. Because this is the value indicating the probability of being wrong in rejecting my null-hypothesis, I could I argue that there is no relationship between higher numbers of homeless people and higher house prices.
Table 4. Regression Analysis
Table 5. shows the residual values for each borough. Haringey, Newham and Brent stand out with very large positive value meanwhile City of London, Merton and Bexley has huge negative values. In the latter three boroughs homelessness was much lower than expected, in the case of City of London, the reasons for this are pretty obvious. Firstly it is the smallest borough and has the lowest population. Therefore it has the lowest numbers of accommodation in “normal” housing, and so, much less chance for a household to become homeless. It is also the financial district of London and its image is important in economic terms so one would expect the local council to receive special instructions and finances from central government with regards to its homelessness and housing policies. At the other end of the spectrum, in Haringey, homelessness is much higher than expected. This is not a huge surprise as it ranks as one of the most deprived boroughs in the country, with 8.1 per cent of the population unemployed in January 2001, double the national average. Almost half of its 223,700 people come from ethnic minority backgrounds, which is probably part of the reason for the high number of homeless.
Table 5. Residual Values- Highlighted in red is the highest values while lowest are in blue.
Meaning of the Results
A scatter-chart visualisation then helped me further to become absolutely aware of the no-relationship theory being right. Chart 5. indicates this clearly especially when the linear trend-line was inserted.
Chart 5. Scatter Diagram
Conclusion
Due to the nature of the project with its time and resource limitations, it is only a snapshot and I could hardly give a real and meaningful account about the link between house prices and homelessness. Firstly I had to pick out one particular quarter of one particular year, which in itself limits any other periods to be accounted for and therefore seriously affecting the end result. So my samples here were not truly representative of the population. It would have been very useful to compare growth rates of house prices versus growth rates of homelessness over a long period of time, such as a decade. Finding house prices growth rates would have been easy as they are readily available and have a reliable source. However extracting the same data for the dependent variable is almost impossible and even if it were done it would be distorted. The main reason for this is how much the legislation changes affected data recording over the decades. There is also of course the issue of population growth in London, which would obviously have an effect on the number of homeless as the pressure on housing had been growing. Not to mention the periodic influx of refugees during conflicts in Balkans and from other war-stricken all over the world. The number of households re-housed after being classed homeless also does not include many hidden-homeless, who squat, rough-sleep or go on sleeping in friends’ houses for a considerable length of time. Homelessness has been showing a steady growing trend in London for the last decade and the same growing trend is obvious with regards to house prices. To link them up I would have needed over 30 pairs of values to apply regression analysis and give an accurate picture about how strong the link was. The lack of data and time meant that this was not feasible. I would have also preferred to have worked with the number of homeless people per 1000 head population per borough as it would have been a much more representative figure and real comparison between boroughs would have been possible. These were again not readily available and calculating the values out myself, would have taken a great deal of time. So limitations to such a task are in abundance one just needs to learn to over-look certain aspects of the study.
References
-
Centrepoint charity
-
Crisis is the national charity for solitary homeless people
-
Shelter is the UK campaigning charity for homeless and badly housed people.
-
Office of the Deputy Prime Minister’s web-site
-
The Land Registry in England and Wales.
-
School of Geography
-
Resource Information Service
-
_ The ESRC Research Centre for Analysis of Social Exclusion (CASE)