Regional house prices cycles in the UK : a Markov switching Var. Rosen Azad Chowdhury Duncan Maclennan

Transcription

1 Regional house prices cycles in the UK : a Markov switching Var Rosen Azad Chowdhury Duncan Maclennan

2 Ripple effects, house price convergence and house price cycles The Ripple effect states that house prices first increase in the South east, South then spread to other parts of UK (Meen, 1999, Wood, 2003) The convergence theory states that whilst the ratio of house prices in different regions may diverge from historic norms in the short run, a long run relative equilibrium price exists ( Cook, 2003, 2006) Empirical work has been done on nation house price cycles (Bracke, 2011). Very few work actually investigate regional house price cycles ( amplitude, duration and timing)

3 What does this paper look at Employing Markov switching vector auto regression model (MSVAR) for each region we are able to observe the asymmetric growth patterns of regional house prices at different point on the cycle We analyse the duration of the regimes across region and over time. The paper analyse the smoothed probabilities of the regimes obtained from the MSVAR and compare central events/episodes in our sample period across regions and focus on the disparities caused by regional and nation factors

4 True statistical process of house price growth rate Most of the work on house prices assumes growth rate of house price are I(0) or stationary variable (constant mean); hp t = αhp t 1 + Z t + v t Actually growth rate of house prices are stationary variables with shifting means, which can be modelled as; hp t = αhp t 1 + hp m + Z t + v t where, Z is a set of exogenous variables, hp m is the mean inflation at regime m.

5 Fig1: House price growth rates of Scotland and Outer East

6 Data and Methodology Data Seasonally adjusted house price indices obtained from Nationwide. Duration: 1978q2 to 2012q3 Methodology Step one House price growth rate of the13 regions are grouped into two groups (based on principal component factor analysis results) Group one consists of Outer metropolitan, London, Outer east, East Anglia, East Midlands, West Midlands. Group two consists of Scotland, Northern Ireland, North, North West, Yorkshire and Humber,Wales.

7 Table 1: Rotated Factor loadings Variable Factor 1 Factor 2 Uniqueness Scotland Northern Ireland North West North Yorkshire and Humber London Outer East East Anglia South West East midland West Midland Wales Outer metropoliton

8 Step two ( Markov switching vector auto regression) We use Hamilton's 1989 Markov Switching Var model to estimate regional house price cycles We have three states in each of the equations, S=3 1)Negative or low growth rate 2)Medium growth rate 3)High growth rate We estimate the following two model; where i=1,2,3 represents the three states in each region, represents the average growth rates in each state in each region.

9 Results Results of Group one London Outer East S. West East Anglia E. Midland W. Midland Outer Metropolitan Average growth rates Low * *** *** -3.82** * Mediu 2.770** 1.900*** 1.956*** 1.84*** 1.01** 1.121** 0.751*** m High 5.027*** 6.025*** 6.191*** 6.042*** 3.67** 4.441** 3.751*** Transitory probabilities P(1,1) 0.774** 0.842** 0.890** 0.874** 0.750** 0.871** 0.880** P(2,2) 0.903** 0.906** 0.877**.937** 0.916** 0.843** 0.914** P(1,3) * P(2,3) Average duration Low Mediu m High

10 Results of Group two Scotland Wales North Y & H Northern Ireland Growth rate Low * * * * Medium 1.895** 1.212** 1.240*** 1.289*** 0.900** High 5.349** 4.995** 5.400*** 3.348*** 3.022** Transitory probability P(1,1) 0.842** 0.784** 0.812** 0.840** 0.895** P(2,2) 0.868** 0.908*** 0.861** 0.863** 0.935** P((1,3).086* P(2,3) * Average duration Low Medium High

11 Factors affecting heterogeneous growth rates and durations Regional heterogeneities in the housing market structure Heterogeneities in the regional labour market Heterogeneities in the regional financial sector Migration Uniform government housing policies and monetary policy

12 Fig 2a: smoothed probabilities of high growth regime (group one) East Anglia Outer East East Midlands South West

13 Fig 2b: smoothed probabilities of high growth regime ( group two) Yorkshire and Humber North Scotland Wales

14 Fig 3: Smoothed probabilities of the low growth regime Outer East East Anglia Scotland North

15 Fig 4a: House price recovery after 2007 crisis (Smoothed probability of medium growth regime, group one) Outer East East Anglia South West East Midlands South West

16 Fig 4b: House price recovery after 2007 crisis (Smoothed probability of medium growth regime, group two) North Scotland Wales Yorkshire and Humber

17 House price growth rate Fig 5: Northern Ireland Low growth regime Medium growth regime High growth regime

18 Conclusion 1) Magnitude of house price growth rate varies across regions and across time During downswing degree of fall in house prices are higher in group one compared to group two ( supports the idea of Beta-convergence during downswings, Cook 2006, 2012). 2) Duration of cycles are different in different regions (South east, London, East Anglia, East midland have the shortest low growth regimes) 3) Different regions act differently to changes in macro variables 4) Ripple effect might be true