THE DETERMINANTS OF HOUSE PRICES IN THE TWO-TIER HOUSING MARKET IN GUERNSEY

Executive Summary

With an area of just 24km2 and a population of over 65,000, there is an underlying shortage of property in Guernsey, and this, combined with the island’s large financial sector, its low taxes and island life, has put additional pressure on the price of housing. Guernsey is a particularly interesting case because of its unique two-tier housing market system, which has a ‘local market’ for locals and an ‘open market’ for non-locals.
This report analyses a sample of housing transaction data for Guernsey from January 2001 to June 2007 to establish the key determinants of prices in both of Guernsey’s house markets. Specifically, this report examines the risk and return characteristics of the open and local housing markets in Guernsey to determine which market has better capital growth potential, less uncertainty, and greater resistance to external shocks. Separate regression models were generated for each market, both of which showed GDP and interest rates, as hypothesised in the literature, are the key determinants of prices in both of Guernsey’s house markets. Additionally, prices in the local market were significantly affected by the total number of people in the workforce, the percentage of financial sector workers, and (inverse relationship) and the number of property transactions (positive relationship). Based on the data, it appears that the open market, while being more volatile, has better growth potential, less uncertainty, and may be more resistant to external shock, particularly because of the extreme supply constraints and the high demand.

Introduction

An accurate forecast of house prices is important to potential homeowners, as well as other real estate market participants, such as developers, investors, and financial institutions (Limsombunchai et al., 2004). In the exceedingly long run, investing in the house market is an excellent strategy but this market can be volatile and this means that buyers sometimes do not get anywhere near the long run return on this asset within their lifetime. Farlow (2004a) notes that a difference in effective real return of a few percentage points can significantly affect lifetime consumption and therefore buyers need to be careful of buying an overvalued asset.
The extensive literature on the UK housing market in this decade shows strong growth in prices and high volatility, reflecting (1) the increasing constraint on supply, making the supply of housing more price-inelastic and (2) lower response of demand to price signals compared to changes in income (Cheshire, 2004, Farlow, 2004a, b). This is particularly the case in local markets, since local supply constraints limit the response of prices to changes in the economic environment (Barker, 2003). This is one reason recent research has investigated regionally disaggregated data (cf. Bhattacharjee and Jensen-Butler, 2004). Indeed, Meen and Andrew (1998) argue that such regional analyses are essential to getting a more complete understanding of the dynamics of the UK housing market.
The case under study is Guernsey, a small-island, self-governing dependency of the Crown of England. With an area of just 24km2 and a population of over 65,000, it is not surprising there is an underlying shortage of property in Guernsey, which has put pressure on housing prices. Additionally, Guernsey large financial sector provides good employment opportunities, and its low taxes and island life have ensured Guernsey’s popularity and put additional pressure on the price of housing. Guernsey is a particularly interesting case, not only because of the local supply constraints, but also because of its unique two-tier housing market system, which has a ‘local market’ for locals and an ‘open market’ for non-locals.
The vast majority of properties belong to the former category, with only about 7% of the housing stock being available to non-locals. Properties in both categories have seen phenomenal increases in under a decade, with house prices in the local market more than doubling over six years, averaging approximately to £350,500, while open market house prices have risen by over 50% over five years to an average of £980,00 in 2005.
Meen and Andrew (1998) review the academic literature relating to regional property price variations and highlight a lack of unanimity regarding the appropriate theoretical basis for determining the factors affecting house prices, illustrating the complexities involved in house price modelling. The data for this study are average local and open market monthly house prices covering the period from January 2001 to June 2007. The data are provided by Linda Boxall, who is a local resident and can thus invest in either the local or open market. This report examines the risk and return characteristics of both housing markets to determine which has better capital growth potential, less uncertainty, and greater resistance to external shocks so as to help Linda to decide in which market to invest.
The report is laid out in three substantive sections as follows. The next section describes the various attributes of the data. This is followed by a presentation of the interpretation of the models used. The paper then concludes with a brief summary. Key analyses are located in the Appendices.

Data Characteristics

To establish the key determinants of prices in Guernsey’s two-tier housing market, 78 transaction records covering January 2001 to June 2007 form the basic data for analysis. The data represents a sample of the average local and open market monthly house price (£) in the two-tier housing market in Guernsey. Along with the average house prices, the dataset also contains detailed characteristics of domestic and external factors of house prices. From these 78 observations, 10% of the data has been removed randomly to calculate out of sample mean square error (MSE), which would help in predicting the best model fit as compared with other models (Tabachnick and Fidell, 2001). For the remaining 70 observations, Figure 1 shows the trend in average house prices in the local and open markets in Guernsey over the period studied. In both markets, price has been on the increase over the period, however, as noted, prices on the open market (averaging £827,294 in our sample) are consistently higher than local market prices (averaging £313,023).
In the open market, prices are more dispersed, ranging from £406,700 to £1,232,216, compared to a range of £214,632 to £384,801 in the local market. This is borne out by the standard deviation, which is £38,520 for the local market compared to £196,237 for the open market. Mean and standard deviation figures must be interpreted with caution, however, as they are strongly influenced by extreme values that can have a disproportionate effect on its size. This may seem to be the case in terms of house prices on the open market, as some of the houses in this market tend to be very large and expensive, as detailed in Cooper Brouard’s Guernsey Property and Island Life real estate brochure (2007). This, combined with the fact that fewer houses are sold on the open market each month, means that the type of house sold in any particular month is likely to have a disproportionate effect on the average house prices and the apparent dispersion of these prices.
Still, while houses sold for an average of over £770,000 in 10 of the 36 months between 2001 and 2003, the average house sold in 40 of the 42 months between January 2004 and June 2007 was over this price point. This indicates that there has been a substantial increase in the price of houses in this market, independent of the size or type of house being offered. Additionally, it is clear that there was a substantial fall in average prices in late 2001, coinciding with the downturn in the world economy and then the 9-11 terrorist bombings in the USA. House prices averaged £825,905 in the first five months of 2001, nose-diving to £481,316 over the following eight months. The recovery seemed relatively swift though, with house prices going above £700,000 and into seven figures, before slowing down again in 2002. Since the end of 2003, however, prices in this market have stayed consistently above £700,000 and the monthly average since 2006 has been over £800,000.
With regard to the local housing market, prices remained consistent throughout 2001, reaching £300,000 on average for the first time in September 2002 and remaining in that range until 2004. While the month-on-month increase in house prices has been largely steady between 2001 and 2007 in the local market, as evidenced by Figure 1, prices gained momentum at the end of 2005, surpassing £350,000 for the first time and peaking at £380,000 less than one year later. Prices then fell significantly for the first time in 2007, about a year after the sub-prime housing market in the USA collapsed, putting pressure on the global financial markets (likely the affects would have trickled down to Guernsey at about this time). This is in contrast to the open market, which seemed to have felt the impact of the sub-prime crisis more quickly, as house prices fell back into the £800,000 range in mid-2006 after three consecutive months of averages over £1m. Overall though, recovery from external shocks in both markets seems to happen in under one year.
Nine potential determinants of house prices in Guernsey have been identified (see Appendix 2). The price has been defined by domestic factors: GDP, INFLATION, the Bank of England Base Rate (INTEREST), the total number of people in the workforce (EMPLOY), the percentage of finance workers in the total workforce (FINANCE), the number of planning approvals needed (PLANNING), and the number of property transactions taking place (TRANS), as well as external factors: the impact of the 9/11 terrorism attack in 2001 (EVENT) and UK average house price (UK).
Figure 2 shows that the distribution of house price in the local market is skewed to the right, suggesting that the weights would be tilted towards high priced properties. It is clear that the transformation of the data using the logarithmic scale did not bring a significant improvement and therefore the variable was used in its original form.
Text Box:
A similar result was found for house prices (see Appendix 3) and this data were also used in their original form. GDP, EMPLOY, FINANCE, TRANS, UK, and PLANNING conformed to the normal distribution and did not need to be transformed. To further check whether the variables were normally distributed, Kolmogorov-Smirnov tests were undertaken. These showed that all but one of the variable (INTEREST) have Asymp.Sig.(2-tailed) values > 0.05, therefore it can be assumed that these variables are normally distributed (Tabachnick and Fidell, 2001).
Of these nine variables, GDP and real interest rates are hypothesised to be the key determinants of UK house prices as outlined in the literature (cf. Chesire, 2004, Farlow, 2004b, IMF, 2003). GDP is the value of island output (wages plus profits and other local income) and is close to the value of total expenditure taking place within the island. Over the months sampled, GDP per capital averaged £26,384, with a steady increase in GDP over the period. Interest rates are an important consideration because mortgage repayments are the biggest part of a homeowner’s monthly spending (Meen and Andrew, 1998). The Bank of England Base Rate is used in this report since Guernsey does not set its own interest rates. Over the period, interest rates fell from a high of 6% at the beginning of 2001 to a low of 3.5% in mid-2003, before being increased by 0.25 percentage points every few months to end at 5.5% in June 2007. A simple correlation analysis indicated that both GDP and interest rates were highly correlated with house prices in both the local and open housing markets. The rate of inflation fluctuated over the seven-year period, with the lowest rate being 1.9% and the highest being 5.2%. The average over the period was 3.8%. The rate of inflation is important, as it is used to calculate the real GDP per capita and interest rate.
The employment level can also act as a proxy for the buoyancy of the economy since an increase will have an impact on purchasing power and general confidence in the market. An increase in the employment is likely to put pressure on housing prices because, as noted above, earning is positively related to house prices. The number of people in the Guernsey workforce has remained relatively stable and has stayed close to the mean of 31,636 persons. A further increase in Guernsey’s workforce may take place, but it unlikely to come from the local population as the number of people moving into retirement is likely to outweigh the number of people moving entering the workforce. The financial services sector has remained the largest employing sector; however this percentage has fluctuated steady between 22% and 23% over the entire period with no definite signs of increase or decrease. Examining the financial sector specifically is important since many finance workers are brought from outside the island and thus they will pay higher prices for housing. Additionally, finance sector wages are much higher than wages in other sectors and financial institutions buy up properties to house their staff.
The number of property transactions is used as a proxy for demand. Over the period, the number of transactions varied each month, as expected. The lowest number of transaction was 59 in February 2005, while the highest number of transactions was 205 in October 2002. The demand for housing seemed to be increasing, although it fluctuated from month to month, as no fewer than 100 property transactions took place each month between March 2005 and June 2007 and few months over this period fell below the average of 124 transactions. The number of planning approvals needed is hypothesised to have a dampening effect on demand. The number of planning approvals needed fluctuated widely over the period, from a low of 9.67 to a high of 97.3.
External variables were also included in the analysis since it is imperative to determine the degree to which the two housing markets can withstand external shocks. The EVENT dummy variable was included to capture the 9/11 terrorism attack in 2001. Additionally, the UK variable was included to capture the impact of the average house prices in the UK on house prices in Guernsey. The UK house market has been booming for several years, and the data set used here shows a dramatic 130% increase from the lowest average price of £85,879 at the start of the period to a high of £197,450 at the end of the period. Unlike the Guernsey house markets, there has been no significant decline in the average price of houses when measured on the scale of the UK, only steady month-on-month increases in most cases. The analysis showed a very high correlation between UK house prices and GDP (0.988). A rule of thumb is that |r|>0.8 indicates a potentially harmful collinear relationship (Norušis, 2008), as a result UK average house prices was not used in the model estimation.
Overall, based on the literature showing that regional house prices are impacted by earning, real interest rates, housing stock, demographic changes, credit availability, and tax structure (Farlow, 2004a), it is expected that GDP, inflation rate, total employment and financial sector employment are positively related to the house prices in Guernsey, while interest rate and number of planning approvals are expected to have a negative impact on house prices. Inflation and the number of housing transactions are expected to impact house prices, but this can be in either direction.

Model Selection and Interpretation

In this section, the significant contribution of each economic characteristic as defined in the previous section and directly affecting the housing prices have been encapsulated using regression models for both markets. The models use enter, stepwise, and log rhythmic stepwise techniques to enter variables in the model. Following the statement of the model and its interpretation, this section describes the features of the data that lead to the selection of these models.
As an outcome of the analysis of the local market, the following model has been selected:
LM PRICE= β0+ β GDP – β INTEREST – β3EMPLOY – β4FINANCE – β5EVENT + β6TRANS
…………………………………………………………………………………….……….EQ (1)
Here LM PRICE is the local market price. The variable GDP denotes is the value of island output whereas INTEREST denotes the Bank of England Base Rate. EMPLOY denotes the total employed people whereas FINANCE denotes finance workers as % of total workforce. EVENT is the dummy variable included to capture the impact of the 9/11 terrorism attack in 2001 and TRANS denotes the number of property transactions. The β (Betas) are the unknown parameters, which can be determined using the sample data.
This model can be used to estimate the price of local houses whether they are a part of this sample or not. Additionally, EQ2 is helpful in understanding the relationships between the dependent and independent variables by analysing the signs and magnitude of the estimated regression coefficients. The model shows that GDP has the biggest impact on prices in the local housing market. As expected from the literature, an increase in earnings of £61,727 leads to an increase in the price of houses in this market by 1%, as reflected in the positive sign associated with GDP. Similarly, the negative sign associated with INTEREST means that an increase in the interest rate will lead to a fall in the price of houses in the local market.
Surprisingly, the total number of people in the workforce and the percentage of financial sector workers in the workforce have a negative impact on the local property price in Guernsey, which is contrary to the hypothesis that they would positively impact prices. The inverse relationship between the number of financial sector workers in the workforce and house prices in the local market can be explained by the fact that these workers do not have access to the local market, and thus an increase in the percentage of these workers will mean a reduction in the number of people that can invest in the local market. The large percentage of overseas workers in the Guernsey population is therefore likely to be the reason there is also an inverse relationship between population growth and prices in the local market, since population growth of this magnitude (19,122 persons) is most likely to come from overseas workers.
External shocks, specifically the 9/11 terrorism attack in 2001, was shown to have a negative impact on the local property price in Guernsey, as hypothesised. Similarly, the number of transactions undertaken – a proxy for demand – proved to be positively related to house prices. This is in line with evidence in the literature that local real estate markets tend to be demand-led because of supply constraints (Barker, 2003).
The model is a very good fit to the available data, as measured by the adjusted R2. By definition, R2 measures the proportion of the variation in the dependent variable explained by the explanatory variables. Adjusted R2 is a modified measure of R2 that takes into account the number of independent variables in the regression equation and the sample size. It is good practice to use adjusted R2 because R2 tends to give an overly optimistic picture of the fit of the regression (Tabachnick and Fidell, 2001). In the local market, the model selected had an adjusted R2 of 0.859, indicating that the model explains 85.9% of the variation in house prices. The standard error of estimates for the final regression was £238.492 and all F-tests and t-test were exceeded. The t-test shows that all of the 6 variables included in the model are significant at the 0.05 level. The following equation shows the regression fit:
Other detailed information about the model is available in Appendices 4 and 5. Auxiliary regression analysis between various explanatory continuous confirmed the potentially harmful collinear relationship between UK house prices and GDP (0.979), supporting the decision to remove UK house price from estimated model. However, the Durbin Watson Test value was 0.858, which was less than dL (1.369), showing that a positive autocorrelation exists (which is also apparent by examining the plot of least squares residual, Figure 3).
This is not totally unexpected in time series data and some standard statistical tests that depend on the assumption of random samples, such as regression analysis, can still be applied to a time series despite the autocorrelation in the series (Chatfield, 2003).
Variance Inflation Factor of all the variables lie between 1 and 3), demonstrating that there is no important collinearity between the six variables used in the final estimated model. Additionally, the model shows no patterns in its residuals form the final model and its explanatory variables, which confirms that there is no heteroscadisticity (Figure 4).
A separate model had to be estimated for the open market, and as an outcome of the analysis of the data relating to the open market, the following model has been selected:
OM PRICE= β0+ β1GDP…………………….…………………………………………….……….EQ (3)
Here OM PRICE is the open market price units and the variable GDP again denotes is the value of island output.
The estimated version of the model is:
OM Price = -5.377 + 235,170 GDP.………………………..…………………………………….EQ. (4)
As with the previous model, this model can be used to estimate the price of open market houses whether they are a part of this sample or not. As with EQ2, this model shows that GDP is the most significant determinant of prices in the open housing market, as predicted by the literature. As hypothesised in the literature, here is a significant negative relationship between interest rates and housing prices in the open market (at the 0.10 level). However, the incremental contribution to model was so small (discussed below) that it was not included in the model.
This result provides strong evidence that the open market in Guernsey mirrors the UK market, since GDP and real interest rates are found to be the key determinants of UK house prices (Chesire, 2004, Farlow, 2004b, IMF, 2003). This is in contrast to the local market in Guernsey, which acts more like the regional market that it is, being influenced by a larger number of factors, including earnings (GDP), interest rates (INTEREST), demographic changes (EMPLOY) (Farlow, 2004a), as well as demand (FINANCE and TRANS) (Muellbauer and Murphy, 1997) and external factors (EVENT) (Fuerst, 2005). Additionally, it is not surprising that changes in GDP have a greater impact in the open housing market compared to the impact in the local housing market, since the local housing market prices respond to a greater number of factors.
The model is a reasonable fit to the available data, the adjusted R2 = 0.473, indicating that the two variables included in the model explains 47.3% of the variation in house prices on the open market. The standard error of estimates for the final regression was £782.571 and all F-tests and t-test were exceeded. This model is significant at the 0.5 level. As noted above, interest rates were significant (at the 0.10 level), but the increase in adjusted R2 with the inclusion of interest rate is only 0.02 (from 0.473 to 0.493), and this did not seem to justify the inclusion of interest rates in the model, particularly since such a model would only be significant at a 0.10 level. The following equation shows the regression fit:
Other detailed information about the model is available in Appendices 6 and 7.

Summary and Concluding Remarks

Although house prices in both markets fluctuate, this report has shown that GDP and interest rates are key determinants of house prices in Guernsey. In the local market, the recommended regression model concludes that local house prices are positively correlated with GDP and demand, while they are inversely correlated with interest rates, the number of people in the workforce (particularly when they are from overseas), and external shocks. In the open market, house prices respond strongly to changes in GDP, mimicking the UK market.
The data examined have allowed us to assess the risk and return characteristics of the open and local housing markets in Guernsey. While in absolute terms the open market prices have grown into £1 million, the percentage growth in the open market from January 2001 to June 2007 was 54.9%, compared to a 56.2% growth in the local market. The rapid increase in prices in the local market is even more alarming, since the market grew by 79.3% between January 2001 and December 2006 before falling in 2007 by 12.8% between December 2006 and June 2007. This dramatic fall in prices in the local market after several years of steady growth may suggest that houses in this market were overvalued (Cameron, 2005, Muellbauer and Murphy, 1997).
On the other hand, the growth in the local market seems a lot less volatile than the growth in the open market, which fluctuates a lot around the average prices. At the same time, the growth in the open market seems a lot more certain because the small number of houses available in this market and the greater income most people buying in this market have. In comparison, the growth in local prices may now be close to their peak and thus growth in this market may be very uncertain over Linda’s lifetime.
In the final analysis, the critical factor is how strong and well equipped the housing markets are to withstand shocks, i.e. their resilience. The most resilient markets are particularly those in which domestic demand is less dependent on external demand and therefore weaker “shock absorbers” allow external fluctuations to produce larger housing price volatility (Loayza and Raddatz, 2008). In both the local and open markets, favourable cyclical factors have played a role but a global slowdown could set back some of the recent progress and this might have been the case since December 2006 in the local housing market. However, both markets are likely to be open to external shocks from the UK and the rest of the world because of Guernsey’s small size. In the open market, external risks are considerably attenuated by the greater income most people who participated in this market. Based on the data, it appears that the open market, while being more volatile, has better growth potential, less uncertainty, and may be more resistant to external shock, particularly because of the extreme supply constraints and the high demand.
This study was based on a data over a period of six and a half years and also important determinants were identified, future research may uncover other important determinants. This is particularly important in the open housing market, because even though GDP and interest rates were identified as major determinants, these only accounted for about half of the variance in house prices and so the identification of other key determinants is important. Similarly, in the local housing market, several variables performed in ways that were not hypothesised and so this is an area that warrants further examination.