• No results found

Bubble case studies: An empirical perspective

 

6.2.1 Descriptive Analysis

There is an obvious interest in conducting an empirical analysis of the last two UK housing bubbles. This sub-section aims to observe and compare the graphical trajectory of the selected bubble case studies and to examine them on the basis of persistence, magnitude and severity. For the purpose of analysing the selected case studies with regards to persistence, magnitude and severity, the study adopts the approach of Agnello and Schuknecht (2011). The persistence (i.e. duration) of each bubble case is calculated as the temporal distance from the beginning to the end of each event. The magnitude is defined as the size of price change from the beginning to the end of the bubble. Finally, the severity is computed by combining persistence and magnitude for each case , via a triangle in which the base represents persistence and height is the magnitude . Hence, the severity is computed as . This sub-section utilises data on UK nominal average house prices, as provided by Nationwide statistics, for the periods of 1986-1989 and 1990 and 2002-2007 and 2008. I also use long-term data for the period of 1980-2011 to make price comparisons between bubble and non-bubble periods. The nationwide average house price is the nominal average price of all property types (i.e. new houses, modern houses, older houses). This dataset has been converted to annual figures to reduce noise from short-term ‘‘interruptions’’ of long-term trends.

Figure 9.0 shows the direction of house prices during the bubble periods of 1986-89 and 2002-2007. The broken line denotes the 1980s bubble, while the solid line indicates the equivalent values for the 2000s bubble. It must be clarified that in Figure 9.0, the last year of

i

Dt Ai

each case has been ignored. This is because 1990 (i.e. the 5th year of the 1980s bubble) and

2008 (i.e. the 7th year of the 2000s bubble) are recorded as being the first bust years for each

case. Figure 9.0 also reveals that the 1980s bubble ended in its fourth year (1986-1989), while the second bubble lasted two years longer, ending in its sixth year (2002-2007). However, in the 1980s bubble, house prices rose exponentially, while in the 2000s case, prices slowly moved into a period of disinflation (i.e. lower annual rate of increase). In terms of price magnitude, the results in Table 11 reveal that the 1980s case saw a total magnitude of 63%, while for the 2000s the magnitude was 77%. As for the severity, the 2000s case was almost twice as severe as the 1980s case.

Figure 9. UK housing bubble case studies: 1986-89 and 1990-bust vs. 2002-07 and 2008-bust

   

   

Table 11. Persistence-Magnitude and Severity of UK bubble case studies

Years Persistence (Years) Magnitude (%) Severity 1986-1989 4 63 126 2002-2007 6 77 231 -10.00% -5.00% 0.00% 5.00% 10.00% 15.00% 20.00% 25.00% 0 1 2 3 4 5 6 7 8

UK Housing Bubbles 1980s and 2000s

6.2.2 Correlation Analysis

This sub-section aims to examine the correlation pattern between house prices and selected variables, with the view to enhance my empirical understanding on the UK housing bubble cases. In particular, this sub-section will look at correlation patterns within several time frames and market conditions, including the long term, normal periods and bubble periods. With this approach, I seek to observe whether any ‘special relationship’ exists during bubble periods and whether a similar correlation pattern exists between the two selected bubble cases. For this analysis, I use log-transformed data based on the natural logarithm for each data set. The dataset covers the period of 1983-2011 and includes the variables of house prices, debt-burden ratio, gross lending, housing completions, income and house price to earning ratio. House prices and debt-burden ratios are available through Nationwide Statistics. Gross lending is available from the Council of Mortgage Lenders, whereas the data for housing completions comes from the Department for Communities and Local Government. Income figures come from the Office for National Statistics, while the house price to earnings ratio is available through Halifax Statistics. A full description of these datasets is provided in Section 5.3.

The study’s Pearson correlation analysis involves the examination of five different time frames over the period of 1983-2011. Table 12.0 indicates the long-term correlation patterns of the selected variables. The first part (1) involves a long-term examination of the period of 1983-2011. The second part (2) covers the period between 1983 and 2011, with the difference being that it excludes the period of the bubble case studies. The third part (3) covers the years of 1983, 1984, 1985, 1996, 1997, 1998, 1999, and 2000. This time frame excludes both the bubble and recession periods, thus only taking into account the ‘normal’ market periods. The fourth (4) and fifth (5) parts of the correlation shown in Table 12.0 are related to the 1980s and 2000s bubble cases, respectively. It is worth clarifying here that the year 2001 has been taken into account only in the first part of Table 12. This is because there are different views in the literature regarding whether 2001 is considered the year that signalled the beginning of the 2000s UK housing bubble (Rapp 2009, Clark et al. 2010, Dolphin and Griffith 2011, Phang 2013, Kuenzel and Bjornbak 2008). Hence, to avoid any bias, the year 2001 has not been included in the analysis for parts 2, 3, 4, and 5 of Table 12. The results in Table 12 indicate that over the long term (i.e. 1983-2011), the variables of Gross Lending, Income and H/P to income ratio show a strong positive relationship with House Prices, whereas the D/B ratio exhibits a weak positive relationship over the long

term. Paradoxically, over the same time frame, I observe that Housing Completions are negatively correlated with House Prices. By excluding the bubble years over the period of 1983-2011 (part 2), the results reveal no significant changes. The only notable change is that the variable of Housing Completions has changed from -0.22 to -0.78, indicating an even more negative correlation. The third part (3) of Table 12 excludes the recessionary years from the second part (2). Based on these results, I find that the D/B ratio; Housing Completions and the H/P to Income ratio are negatively related to House Prices, while Gross Lending and Income are strongly correlated to House Prices. However, after focusing on the bubble years for each case, the correlation results (between house prices and the rest of the variables) are vastly different, indicating that ‘‘something happened’’ in these two bubble periods. For instance, in the fourth part (4) of Table 12, that is, the 1980s bubble, all of the variables tend to move in the same direction as house prices, with D/B ratio, Income and H/P to Income ratio displaying a particularly high positive correlation. As for the 2000s bubble represented in part five (5), the results reveal strong positive correlations between house prices and all the selected variables. Generally, the results in Table 12 lead us to conclude that it is not by coincidence that the selected variables tend to move in the same direction as house prices during bubble periods. This pure-positive tendency can be justified by the occurrence of the phenomenon it self.

Table 12. Correlations between house prices and other variables during various durations

House Price House Log

Price Log D/B Ratio Log Gross Lending Log Housing Completion Log

Income H/P-to- Log Income 1 Log (1983-2011) N 1 29 .35 29 .88 29 -.22 29 .96 29 .70 29 2 Log (1983/84/85/90/91/92/93/94/95/96/97/98/ 99/2000/08/09/10/11) N 1 18 .32 18 .83 18 -.78 18 .96 18 .67 18 3 Log (1983/84/85/96/97/98/99/2000) N 1 8 -.62 8 .99 8 -.33 8 .99 8 -.81 8

Bubble Case Studies

4 Log (1986/87/88/89) N 1 4 .96 4 .35 4 .46 4 .98 4 .99 4 5 Log (2002/03/04/05/06/07) N 1 6 .99 6 .80 6 .96 6 .98 6 .99 6

6.2.3 Theoretical validity of the Correlation results

By employing a compressive Pearson product-moment correlation coefficient in Sub-section 6.2.2, the study found that during bubble periods, D/B ratio, Gross Lending, Housing Completions, Income and House Price to Earning ratio exhibit a tendency to move in the same direction (i.e. a positive correlation) as House Prices. This sub-section is devoted to providing theoretical support for the above-mentioned finding by examining whether any existing theoretical framework of housing bubbles supports the empirical evidence presented in Section 6.2.2.

House Price vs. Gross Lending: Allen and Gale (2000) and Holcombe and Powel (2009) argue that a housing price bubble is part of a more general credit bubble and thus cannot exist without the existence of the latter. On a statistical basis, during housing bubbles, the index representing values of Gross Lending in housing is expected to show signs of a positive change (i.e. an increase). Therefore, given that house prices are always changing in a positive direction during bubbles, it is plausible to expect a positive relationship between House Prices and Gross Lending during housing bubbles.

House Price vs. Housing Completions: Mueller and Pevnev (1997) and Baum and Hartzell (2012) note that building development activity in housing markets increases during a boom. Developers always have greater incentive to invest in the construction sector (i.e. supply) when prices are rising because business profit-maximisation is inherent in rising prices when costs are held relatively stable. In further support of this argument, Holcombe and Powel (2009) and Baum and Hartzell (2012) note that the quantity of houses built often helps in the measurement of bubbles and can be described as a quantity dimension of bubbles. Housing completions are inherently prone to show an increase (positive change) during housing bubbles. The same applies to house prices. Therefore it is plausible to expect that during housing bubbles, prices and housing completions would move in the same direction.

House Price vs. Housing Affordability Indicators (D/B ratio and H/P to Income ratio): An increased affordability index value is translated as a decrease in housing affordability, while a decrease in the affordability index value implies an increase in housing

affordability. During housing bubbles, prices increase disproportionately to changes in income. As a result, affordability decreases (Case and Shiller 2003, Tsai 2013). Therefore since a decrease in affordability is translated as a positive change in the affordability indices, and since during bubbles, prices are always increasing35, it is logical to expect that house

prices will move in the same direction as the affordability indices during bubble periods.

House Price vs. Income: According to the study of Case and Shiller (2003) on the US housing market, there is a remarkable tendency for income to show positive performance along with house prices during housing bubbles. As for the UK housing market, the studies of Dolphin and Griffith (2011) and Kuenzel and Bjornbak (2008) have provided evidence that income tends to move in a positive direction during housing bubbles, as house prices do.

   

6.3 Can housing price-rate-of-change explain the presence of