A. Introduction

1 The United Kingdom has experienced three major episodes of house price booms since 1970 and large recent increases in house prices have raised concerns about a new housing market bubble. U.K. house prices have increased by close to 50 percent over the last three years, with the annual increase in December 2002 amounting to about 25 percent. This chapter examines the factors determining house prices in the United Kingdom and, based on econometric evidence, assesses whether recent house price increases can be explained by fundamentals or whether they represent a temporary overshooting of house prices, characteristic of a bubble, Building on observations on the U.K. housing market in the recent literature, we estimate a simple error-correction model that describes the dynamics of real house prices since the 1970s. Estimation results suggest that (i) earnings and interest rates are the key determinants of house prices; (ii) changes in real house prices exhibit a large degree of persistence, which can contribute to price overshooting, and, (iii) actual house price increases in the first half of 2002 have significantly deviated from those implied by long-run fundamentals, even taking short-run adjustments into account.

B. Characteristics of the U.K. Housing Market

2. Theoretically, equilibrium house prices are determined by the interaction of supply and demand. Empirically, this would translate into a simultaneous model of demand and supply equations, where demand for houses would be affected by house prices, income and mortgage interest rates, and supply would respond to house prices and construction costs. Often, U.K. house price models in the literature contain no explicit supply side on the assumption that supply is rigid due to restrictive planning policies. This section briefly reviews the literature to assess the validity of this assumption and goes on to discuss factors characterizing the demand for housing. It provides a context and a guide to the specification of the empirical model estimated in the next section.

Demand Conditions

5. The U.K. housing market is characterized by a high income elasticity of housing prices. Estimated long-run income elasticities of housing prices range from 1.7 to 3.0 (see evidence summarized in Meen, 1996). A recent study by PricewaterhouseCoopers finds a long-run income elasticity of 0.9. It is difficult to assess whether the high sensitivity of house prices to income places the United Kingdom in a unique position, as studies differ in their estimates of income elasticities across other European housing markets.⁴ However, crosscountry studies have linked high income sensitivity of house prices to high Loan to Value (LTV) ratios Almeida, and others (2002) show that in countries with high LTV ratios such as the United Kingdom, house prices increase by 1.2 percent for a 1 percent increase in per capita GDP in contrast to countries with low LTV ratios like Italy, where a 1 percent increase in per capita GDP produces a 0.8 percent increase in house prices.

6. The high LTV ratios in the United Kingdom may also be behind the speculative elements of house price dynamics. Muellbauer (1994), contrasting the evolution of German and U.K. house prices, suggests that the LTV ratios observed in the United Kingdom during the late 1980s (in excess of 85 percent), combined with persistent house price inflation, amplified the rates of return on equity, producing returns significantly greater than the returns on saving in liquid form.⁵ These consistently large rates of return could fuel what the author calls the “explicitly speculative” characteristic of U.K. homeownership. Indeed, a survey across European mortgage markets provides a sound basis for that hypothesis: average LTV ratios for new conventional mortgage loans over 1981–90 in the United Kingdom, at 87 percent, were significantly higher than LTV ratios in Germany (65 percent), Italy (56 percent), and the Netherlands (75 percent).^{6, 7} Current LTV ratios in the United Kingdom, while on a pronounced declining trend, remain high: the majority of new mortgages have LTV ratios between 75 and 90 percent, although some 4–7 percent of new mortgages have LTV ratios of 100 percent and higher.⁸

7. While the high LTV ratios observed in the U.K. mortgage market may provide relatively more scope for speculation in house prices than in some other European countries, regional dynamics may also play a role. A more in-depth analysis of regional housing markets, which is outside the scope of this paper, may reveal further insights into the dynamics of house prices in the United Kingdom. Some studies show a rippling effect from the South East (Meen, 1996), suggesting a role for speculative forces in subsequent price increases elsewhere. However, a recent assessment of the evidence by analysts at Goldman Sachs⁹ does not substantiate the claim that the London market leads prices elsewhere.

8. Short-term interest rates are likely to be more important determinants of housing demand in the United Kingdom than long term rates. Most mortgage contracts in the United Kingdom are variable rate mortgages. In contrast to the U.S. market where the majority of mortgages are fixed-term at long time horizons, fixed-term mortgage contracts in the United Kingdom represent only about 35 percent of total mortgage loans,¹⁰ and usually imply a fixed term of 5 years or less. Hence, mortgages and house prices are relatively more sensitive to short-term interest rates.

9. However, whether housing demand responds to real or nominal interest rates is an empirical question. Decreases in the real cost of mortgage financing should result in an higher equilibrium housing price. However, nominal rates may also matter. High nominal rates (stemming from high inflation) shift the burden of mortgage prices to the early years of the mortgage, while lower nominal rates (driven by lower inflation with unchanged real rates) reduce the initial burden of mortgage payments. The decline in nominal rates may have resulted in increased demand by bringing more first-time buyers into the market—buyers who might otherwise have been liquidity constrained.

10. In summary, the U.K. experience with house price inflation is quite unique. While there has been a dramatic increase in the number of households, the supply of new houses has been severely constrained (Figure 1)—mostly due to restrictive planning policies—and notably unresponsive to house price changes. Conventional determinants of housing supply, such as construction costs and house prices themselves, are unlikely to play a significant role in the United Kingdom, and house prices appear to be determined predominantly by demand factors. Thus, our model focuses on demand-driven dynamics, incorporating factors such as real earnings per household and short-term interest rates.

View Full Size

U.K. House Prices and Structural Factors

Citation: IMF Staff Country Reports 2003, 047; 10.5089/9781451814170.002.A001

• Download Figure
• Download figure as PowerPoint slide

C. The Empirical Model

11. We employ an error-correction model, where real house prices adjust to their long-run equilibrium while responding to short-run movements in house prices in previous quarters, interest rates, and real income per household. Real house prices are calculated by deflating the ODPM house price index by the Retail Price Index (RPI). Nominal interest rates are short-term, inter-bank 3 month interest rates, and are converted to real interest rates using a 8-quarter moving average of RPI. Real income per household is calculated by deflating total nominal household resources by the consumer expenditure deflator and dividing by the number of households.¹¹ The model is estimated over 1972 Q4 through 2001 Q3. We then forecast real house prices three quarters out, through 2002 Q2, and compare the actual house price increases to those predicted by the model.

12. All series are I(1), i.e., contain a unit root¹², and real house prices, real income, and interest rates co-integrate. The Augmented Dickey-Fuller unit root tests for the prices, income, and interest rates over the estimation period are shown in Appendix Table A1. The selected lag length for the model is 3 lags (Table A2). Both the model with nominal interest rates and the model with real interest rates co-integrate, although the evidence of one co-integrating vector is weaker for the model with nominal interest rates (Table A3).

13. While both models show a long-run income elasticity of house prices consistent with results from the literature, only the real interest rate, and not the nominal interest rate, enters the long-run relationship at standard significance levels. The model with real interest rates is presented below¹³, while the two models are compared in Table 1:

Noteworthy in both models is the high degree of persistence in real house price changes, as measured by the coefficients on lagged house prices, which are large and significant as far as three quarters back. This type of persistence is consistent with a tendency of prices to overshoot. Interestingly, while higher real interest rates reduce equilibrium real house prices, nominal interest rates do not appear to have a long-run effect on real house prices. In the short-run, however, real house prices do respond to changes in nominal interest rates (see Table 1).

14. The dynamics of house price inflation, as captured by the empirical model shown above, predict house price increases significantly below actual house price increases in the first half of 2002. We fit the model¹⁴ through the last three quarters of available data (2001 Q4-2002 Q2). Figure 2 shows the deviation of real house prices from long-run equilibrium.¹⁵ While actual house prices continue their strong upward trend, equilibrium house prices level off in the first half of 2002, consistent with an observed slowdown in earnings growth. Figure 3a compares the actual changes in real house prices with the estimated changes given by the Vector Error Correction (VEC) model which incorporates both short-run movements and adjustments to long-run equilibrium. Figure 3b confirms that the actual price change lies more than one standard error above the price predicted by the model. Evidence from both panels of Figure 3 suggests that the magnitude of recent house price increases over their equilibrium value cannot be explained by short-term developments in real income and interest rates.

View Full Size

Equilibrium and Actual Real House Prices in the United Kingdom

Citation: IMF Staff Country Reports 2003, 047; 10.5089/9781451814170.002.A001

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Actual vs Predicted Changes in U.K. Real House Prices

Citation: IMF Staff Country Reports 2003, 047; 10.5089/9781451814170.002.A001

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Residuals from Estimateded VEC model of U.K. Real House Prices

Citation: IMF Staff Country Reports 2003, 047; 10.5089/9781451814170.002.A001

• Download Figure
• Download figure as PowerPoint slide

15. Despite the apparent gap between actual and estimated equilibrium prices as early as 2000, it is only in 2002 that increases in real house prices appear significantly out of line with movements in their determinants. The residuals of the estimated ECM lie within the 1 standard deviation band up until Q1 2002 (suggesting that the predicted values are reasonably close to the actual house price changes). However, the estimated prices themselves could reflect a possible bubble because they are computed using the lagged actual house prices. Thus, the existence of a bubble could go undetected. To address this concern, we estimate the predicted values for Δp using lagged predicted—rather than actual—values for the 2001–02 period. The results are shown in Figure 4, which compares the actual price increases to the predicted price increases, including those predicted values generated using the method outlined above. This exercise clearly confirms that it is indeed only in the first two quarters of 2002 that the real price increases have been significantly larger than can be explained by adjustments to equilibrium or responses to short-run movements in interest rates nd income.

View Full Size

Actual House Price Changes and Simulation of Model Dynamics, 2001–02

Citation: IMF Staff Country Reports 2003, 047; 10.5089/9781451814170.002.A001

• Download Figure
• Download figure as PowerPoint slide

D. Concluding Remarks

16. The empirical model of U.K. house prices presented in this chapter focuses on demand-side factors (income and interest rates) in explaining the type of house price dynamics that we observe in the United Kingdom, as supply is notably unresponsive to price changes. Yet, clearly the rigidity of supply amplifies the price effects of shifts in demand. The empirical model of house prices presented in this chapter shows a 26 percent positive deviation of actual real house prices from their estimated long-run equilibrium in the second quarter of 2002, a relative overvaluation that is nonetheless smaller than that at the peak of the last housing boom in the late 1980s.¹⁶ An error correction specification of changes in real house prices shows that real house prices adjust to their long-run equilibrium while responding to short-run movements in house prices in previous quarters, interest rates, and real income per household. Results show that only in the first half of 2002 did actual house price increases rise significantly above than the price increases predicted by the model, raising the likelihood that recent price increases are unsustainable and can lead to a sharp correction.

Table 1.

United Kingdom: House Price Model, Estimation Results, 1972:4 - 2001:3

Constant omited from table t-statistics in [] All variables except interest rates are in logs. Δ denotes first differences. p represents real house prices, r represents short-term interest rates, and y represents real income per household.

Table 1.

United Kingdom: House Price Model, Estimation Results, 1972:4 - 2001:3

	Model with Real Interest Rates	Model with Nominal Interest Rates
Cointegrating Vector
Interest Rate	-0.017	-0.003
	[-1.805]	[-0.150]
Real Income per Household	1.329	0.979
	[5.555]	[2.836]
Error Correction:
Lagged Deviation from Equilibrium	-0.043	-0.020
	[-3.604]	[-2.194]
Δp(t-1)	0.434	0.399
	[4.461]	[4.204]
Δp(t-2)	0.296	0.267
	[2.837]	[2.654]
Δp(t-3)	0.173	0.231
	[1.698]	[2.334]
Δr(t-1)	-0.004	-0.006
	[-2.562]	[-3.723]
Δr(t-2)	-0.002	-0.003
	[-0.915]	[-1.917]
Δr(t-3)	0.001	-0.002
	[0.431]	[-1.217]
Δy(t-1)	0.235	0.152
	[1.932]	[1.221]
Δy(t-2)	0.033	0.023
	[0.260]	[0.178]
Δy(t-3)	0.092	0.113
	[0.761]	[0.942]
Number of Observations	116	116
Adj. R-squared	0.54	0.56

Constant omited from table t-statistics in [] All variables except interest rates are in logs. Δ denotes first differences. p represents real house prices, r represents short-term interest rates, and y represents real income per household.

View Table

Table 1.

United Kingdom: House Price Model, Estimation Results, 1972:4 - 2001:3

	Model with Real Interest Rates	Model with Nominal Interest Rates
Cointegrating Vector
Interest Rate	-0.017	-0.003
	[-1.805]	[-0.150]
Real Income per Household	1.329	0.979
	[5.555]	[2.836]
Error Correction:
Lagged Deviation from Equilibrium	-0.043	-0.020
	[-3.604]	[-2.194]
Δp(t-1)	0.434	0.399
	[4.461]	[4.204]
Δp(t-2)	0.296	0.267
	[2.837]	[2.654]
Δp(t-3)	0.173	0.231
	[1.698]	[2.334]
Δr(t-1)	-0.004	-0.006
	[-2.562]	[-3.723]
Δr(t-2)	-0.002	-0.003
	[-0.915]	[-1.917]
Δr(t-3)	0.001	-0.002
	[0.431]	[-1.217]
Δy(t-1)	0.235	0.152
	[1.932]	[1.221]
Δy(t-2)	0.033	0.023
	[0.260]	[0.178]
Δy(t-3)	0.092	0.113
	[0.761]	[0.942]
Number of Observations	116	116
Adj. R-squared	0.54	0.56

Constant omited from table t-statistics in [] All variables except interest rates are in logs. Δ denotes first differences. p represents real house prices, r represents short-term interest rates, and y represents real income per household.