MATERIALES Y MÉTODOS
4.4 VARIABLES REGISTRADAS Y MÉTODOS DE ANÁLISIS
4.4.2 Propiedades químicas Constituyentes inorgánicos
This section aims to provide a snapshot of the conclusions reached by previous studies that examined whether a housing bubble existed in the UK. Technically, the UK housing market was in a bubble during the periods of 1986-1989 and 2001-2007. Given the critical advantage that our research has in terms of time, being that it a later study, I can compare the results of previous work on UK housing bubble detection to observe whether those models manage to correctly identify the known UK bubble cases. This analysis will help us to better understand the effectiveness of the existing models for diagnosing bubbles in the UK housing market. Details of the methodology used by the studies presented below can be found in Chapter 4. Cameron et al. (2006) addressed the question of whether there was a bubble in UK house prices in the 2000s. In their analysis, the authors defined a bubble as a systematic but temporary deviation of house prices from fundamentals. The model consists of a system of dynamic equilibrium-correction (inverted housing demand) equations, incorporating spatial interactions and lags and the relevant spatial parameter of heterogeneity. This model used data from the period of 1972-2003. The period between 1972-1996 was used to build the model’s estimation, while the period between 1997 and 2003 was used for forecasting purposes. The results suggest that in 2003, house prices were not substantially overvalued. Thus, no bubble is detected in the UK in the early 2000s. Also, the study proposed “moderate nominal falls in house prices in 2006-2007 are a possibility, especially in London and the South but not in the country as a whole” (p.23). However, revisiting the study of Cameron et al. (2006), Muellbauer and Murphy (2008) suggest that by mid-2007, UK house prices looked "slightly overvalued" (cited in Kuenzel and Bjornbak 2008).
By examining the successfulness of Cameron et al.’s (2006) model for detecting UK housing bubbles, one can assert that it did not prove to be correct in practice. In particular, the model failed to identify the UK housing bubble that was forming in the early 2000s. Also, the second finding of Cameron et al. (2006, p.23) that ‘‘moderate nominal falls in house prices in 2006-2007 are a possibility, especially in London and the South but not in the country as a whole’’ also proved to be incorrect. As to the study of Muellbauer and Murphy (2008), their findings that UK house prices looked slightly overvalued by mid 2007 are not inconsistent with the true market condition at that time. This is because house prices by mid 2007 were in the final bubble year, indicating a peaking bubble year rather than a slight overvaluation.
Also, the period prior to 2007, that is, 2001 to 2006, was also bubbly, but Muellbauer and Murphy’s (2008) model also failed to identify them as such.
Weeken (2004) applied a simple asset-pricing model (i.e. dividend discount model) to examine whether the observed rate of increase in house prices was unsustainable. Additionally, Weeken (2004) introduced a constant real rate growth model. The dividend values in Weeken’s (2004) model are implied by the rental values. Such data is available from real estate agencies and research institutes like Investment Property Databank (IPD). It is important to note Weeken’s (2004) finding that ‘housing dividends’ are difficult to estimate and that asset pricing models face several limitations. In relation to this, the study suggests that no clear results can be produced due to data and model limitations. However, by comparing the implied housing risk premium in 2004 (the study year) with that of the late 1980s, the author suggests that house prices were closer to sustainable levels in 2004 than was the case in the late 1980s. Thus, no bubble was detected for the early 2000s.
By critically analysing the accuracy Weeken’s (2004) results for the selected UK housing bubble cases, it seems that they do not capture the real picture. This leads us to conclude that the asset pricing model, to the extent that it was used by Weeken (2004), is unable to explain and correctly diagnose the housing bubble component in the UK housing market. However, Weeken (2004) also employed the indicator of house prices to net rentals (a concept close to the price to earnings ratio used for equity markets) by linking it to the mean reversion theory. After accounting for this, Weeken (2004) showed that in the 2000s, this indicator was well above its long-term average, thus concluding that 2004 house prices might be unsustainable. This argument is in line with our findings, which show that the concept of affordability in conjunction with the mean reversion theory could be helpful in diagnosing housing bubbles in the UK, particularly if used along with other variables in an algorithmic model.
Another interesting study that focused on UK housing bubble detection was that of Black et al. (2006). They proposed an advanced procedure using a time-varying present value approach to assess whether house prices are deviating relative to fundamental house prices in the UK. An adaptation of the vector autoregressive (VAR) methodology was also used along with other tests such as the Ljung-Box test statistic and Wald test statistics to enhance the statistical dynamic of their results. Black et al.’s (2006) model made use of UK housing data from Q4 1973 to Q3 2004. It covered the variables of house prices, disposable income and retail price index. The house price data came from Nationwide while data for real disposable income and retail price index (RPI) come from the Office of National Statistics. The housing
data were deflated by the RPI in order to convert the prices to real terms. The results suggest that the time-varying present value model does not fit well with actual prices, displaying periods of over and under valuation of prices. In particular, the study found departures between actual and computed fundamental prices for the mid-to late 1980s. Regarding this time frame, the study failed to identify the first year of the 1980s bubble (i.e. 1986) since actual prices were close to the fundamental values in that year. As for the 2000s bubble, the model’s results indicated that by the end of the sample time period (September 2004), there was a 25% gap between actual price and estimated fundamental price. Nevertheless, for the early formative years of the 2000s case (i.e. 2001, 2002, and 2003), their results showed that prices were ‘‘under’’-valued, despite the current widespread view that 2001, 2002 and 2003 were bubbly years and thus that house prices over-valued. To the best of my knowledge, the results produced by Black et al.’s (2006) model were the most successful compared to the rest of the UK studies. Nevertheless, it must be highlighted that the model failed to identify the year 1986 as the first bubble year of the 1980s bubble, nor did it identify the early years of the 2000s bubble. This leads us to conclude that while this approach may be useful for raising some warning when the market is experiencing a bubble, it may not successfully produce bubble warnings during the development of the phenomenon.
The next model is that of Zhou and Sornette (2003), which also focused on identifying and testing whether the UK and US housing markets were experiencing an unsustainable speculative bubble using a pure econophysic approach. This approach was created for application to financial assets rather than housing, which is significantly different in nature (Geraskin and Fantazzini 2011). In particular, Zhou and Sornette (2003) used the Log- Periodic Power Law (LPPL) to identify signals of faster-than-usual performance. These signals had been found to be consistent predictors of previous crashes in financial markets. The housing data used in the study was based on the Halifax housing indices and covered the period between December 1992 and April 2003. The paper concluded the following: (1) The US real estate market was in a “rational expectation” regime, and thus was not in a housing bubble.
(2) The UK real estate market was exhibiting an ultimately unsustainable speculative bubble in 2000-2003.
(3) The analysis points to the end of the UK bubble around the end of 2003, with either a crash or a change of direction by moving into a deflation regime.
As for the validity of the results obtained via the LPPL approach, I argue the following. First, the study was partly correct regarding the second finding, since the UK housing market is now believed to have been in a bubble in 2001, 2002 and 2003, although the year 2000 is not considered in the literature or our model to be a bubble year. Secondly, their prediction that UK housing bubble of the 2000s would end in a crash in 2003 or change direction (into a deflation mode) has proven wrong. Specifically, in 2003, UK house prices did not crash, nor did they transition to a deflation regime. Instead, prices continued to rise, albeit less rapidly, and can therefore be seen to have transitioned to a disinflation regime rather than to deflation. Although it is beyond the scope of this section to discuss findings obtained for other markets, it is worth mentioning that Zhou and Sornette’s (2003) model was also wrong in terms of its first finding. This is because it is currently well accepted that the US housing market was in a bubble in the 2000s. Despite this, the LPPL approach has gained popularity over the last years, as it has provided several successful bubble warnings in finance markets. However, the level of effectiveness of LPPL for the housing market remains ambiguous.