Home Away from Home? Safe Haven Effects and London House Prices

Transcription

1 Home Away from Home? Safe Haven Effects and London House Prices Cristian Badarinza Tarun Ramadorai Oxford, 25 October 2013

2 Overview Global political and economic uncertainty are currently extremely high both fluctuate substantially over time. One frequently mentioned reaction to uncertainty flight to safe haven assets (with associated asset price impact). Gold, U.S. treasuries, cash, real estate. Empirical evidence for safe-haven demand sparse time-series power limited by short samples and infrequent global crises. Our paper: Provides empirical evidence that London real estate acts as a global safe haven asset. Innovative strategy utilizes cross-sectional variation in population of London wards, identified with specific global regions. Sheds light on mechanisms through which safe haven demand shocks affect London house prices. 2 of 28

3 Public media "Worsening financial and political turmoil in Southern Europe caused a surge of interest in London property [...] with buyers from Greece and Spain showing strongly among investors seeking a safe haven for their money." Reuters (2012) "There is a real lack of data to underpin a systematic study of the market impact of overseas investment in London. Such information as exists is usually either anecdotal or is irrelevant [...]. Even where data is collected spatial definitions do not always correspond to those used in other housing related research. [...] Anyone doubting the scale and potential impact of overseas investment in London should note that at 5.2 billion in 2011 it was a larger sum than the whole Government investment in the Affordable Housing Programme for England for four years. The Programme stands at 4.5 billion for the entire period." Andrew Heywood (2012), International Union for Housing Finance 3 of 28

4 2 Figure 1 Evolution of the London housing market House price spread between London and the rest of the UK Nationwide Lloyds Group Land Registry National Statistics Affordability spread between London and the rest of the UK Nationwide Lloyds Group Note: All variables are normalized by subtracting the in-sample mean and dividing through the standard deviation. 4 of 28

5 Figure 2 External environment Exchange rate (in s per local currency) Stock market (in local currency) 10 year bond yield spread vs. the UK ICRG indicator of political risk London HPI spread Note: The figure reports 1-year moving averages of all variables, normalized by subtracting the insample mean and dividing through the standard deviation. 5 of 28

6 Methodology - An Illustration Time-series evidence difficult usual "out of the money put option return" problem. Utilize the power of the cross-section. Consider the following example: In late 2009 and early 2010, shock to economic and political uncertainty in Greece. Assume Greek residents wish to purchase safe haven residences in London. Identifying assumption: Preferred locations for Greek safe-haven demand are London wards with relatively high pre-existing share of Greek population. Also use other identifying variables: language share, high-income locations, etc. Expect to see relatively higher price increases in these wards following rise in Greek uncertainty. Note: cross-sectional identification does not use aggregate rise in London prices during the crisis. Note: need to control for variation in hedonic characteristics of properties. 6 of 28

7 Methodology Consider the following hedonic house price model: ln P l,w,t = α + βx l + Π w,t +u l,w,t. Here, P l,w,t is the price of property l, which is physically located in ward w, in time period t X l are hedonic characteristics for the property. Π w,t are ward-time fixed effects. Let {z m,k,t } m M,k K be a set M of indicators of economic and political conditions, for a country group K, in period t. Approach is to estimate the relationship between Π w,t and {z m,k,t } m M,k K. 7 of 28

8 Consider the following specification for Π w,t : Π w,t = θc w + δ t + γ k,0 f w,k + (γ m,k,1 f w,k + γ m,k,2 y w,k )z m,k,t. k K Here f w,k is the fraction of people in ward w who were born in country k, and y w,k is a ward-level indicator of desirability, e.g. pre-existing net average income. Implied restrictions: k K The variation across wards is driven by: C w and {f w,k, y w,k } k K. The variation across time is driven by: z m,k,t 8 of 28

9 Data Land Registry Residential property transactions in London Above 2 mil. observations January 1996 to December 2011 Property details: Postcode and full address Date of transfer Property type: house or flat New build indicator Freehold or leasehold ownership 9 of 28

10 Nationwide Building Society Residential mortgages with London property as collateral Around 150,000 observations January 1996 to December 2011 Property details: Postcode Mortgage approval date Property type: house or flat Building construction date Floor area, number of bathrooms, bedrooms and garages Freehold or leasehold ownership 10 of 28

11 Figure 3 Overview of sample coverage 0,000 Land Registry ,000 Nationwide of 28

12 Office for National Statistics Neighbourhood Statistics Ward- and MSOA-level Census data: Housing arrangements Socio-economic structure Income Crime and environmental characteristics Country of birth and primary language spoken 624 wards and 983 MSOAs 12 of 28

13 External developments International Country Risk Guide Composite index: 12 risk categories Assessment of government stability, the risk of armed conflict, religious tensions, general law and order, ethnic tensions and democratic accountability. Exchange rates Spot rates from Reuters (Datastream) Stock market Benchmark domestic stock indexes Bond yield spreads 10-year domestic sovereign bond yields relative to the UK 13 of 28

14 Country coverage 1. China 2. East Asia Japan Malaysia Singapore 3. South Asia India Pakistan Sri Lanka 4. Russia 5. Middle East Egypt Libya Tunisia Qatar Saudi Arabia United Arab Emirates 6. Southern Europe Italy Ireland Greece Portugal Spain 14 of 28

15 Figure 4 Shares of foreign-born people - correlation across wards - China Japan Malaysia Singapore India Pakistan Sri Lanka Greece Italy Ireland Portugal Spain Russia Middle East China Japan Malaysia Singapore India Pakistan Sri Lanka Greece Italy Ireland Portugal Spain Russia Middle East of 28

16 Figure 5 International political risk and London house prices 0.02 People of foreign origin Southern EU China South Asia East Asia Middle East Russia Net average income Southern EU China South Asia East Asia Middle East Russia Land Registry Nationwide 16 of 28

17 Table 1 Estimated ward-level effects Land Registry Nationwide Domestic population share *** *** Population density 0.047*** 0.048*** Percent of detached houses *** *** Percent of flats 0.038*** 0.022*** Cars per household 0.054*** 0.087*** Median age 0.097*** 0.032*** Higher occupations 0.125*** 0.117*** No education *** *** Net income 0.159*** 0.149*** Net income (squared term) *** Assault *** Benzene score ** *** Mortgage ownership *** *** 17 of 28

18 Results Evidence for house price appreciation in areas with higher shares of foreignborn people, in response to higher political risk in the respective country regions. Parallel mechanism: in response to risk shocks in China, the Middle East and Russia, capital is channeled towards areas with higher income levels. Effects are consistent across the mortgage and transactional data sets. A foreign risk shock of one standard deviation is found to lead to up to 1 percent higher house prices in wards which lie one standard deviation above the median in terms of their share of foreign-born people. The same shock would lead to up to 2 percent higher house prices in richer wards, depending on the country of origin. 18 of 28

19 Robustness alternative foreign macroeconomic variables: exchange rates, evolution of stock market wealth and the 10-year bond yield spread versus the UK assessment of the differential impact of foreign political risk across property price categories 19 of 28

20 Figure 6 International political risk and London house prices - People of foreign origin Southern Europe 0.02 China 0.01 South Asia Risk Exch. Stock Yield 0.00 Risk Exch. Stock Yield 0.01 Risk Exch. Stock Yield 0.01 East Asia 0.01 Middle East 0.01 Russia Risk Exch. Stock Yield 0.01 Risk Exch. Stock Yield 0.01 Risk Exch. Stock Yield Land Registry Nationwide 20 of 28

21 Figure 7 International political risk and London house prices - Net average income Southern Europe 0.04 China 0.03 South Asia Risk Exch. Stock Yield 0.01 Risk Exch. Stock Yield 0.08 Risk Exch. Stock Yield 0.02 East Asia 0.08 Middle East 0.03 Russia Risk Exch. Stock Yield 0.00 Risk Exch. Stock Yield 0.08 Risk Exch. Stock Yield Land Registry Nationwide 21 of 28

22 Figure 8 Safe haven effects across price categories - Nationwide sample People of foreign origin Southern EU China South Asia East Asia Middle East Russia Net average income Southern EU China South Asia East Asia Middle East Russia Bottom third Middle third Top third 22 of 28

23 Immigration and house price appreciation Our identification assumption is that wards with high immigration shares attract international investment, in times of financial and political distress. One possibility is that these investments are carried out by wealthy, well-informed people and are driven by a purely speculative motive. However, when the external risks actually materialise for a broader part of the domestic population, we expect the capital flows towards London to be accompanied by subsequent increases in immigration. Research question: are price increases in wards with higher shares of foreignborn people a signal of increased future immigration? The relocation of people is expected to occur at a much slower pace and to have long-lasting effects on the demographic structure of the UK population. We use Census information between 2001 and 2011 to test this hypothesis only available for a restricted set of countries Note: this is currently work in progress 23 of 28

24 Obtain the residuals and fitted values from hedonic regressions, estimated for each country and variable separately: ln P l,w,t = α + βx l + θc w + δ t + γ 0 f w + (γ 1 f w + γ 2 y w)z t + u l,w,t. Calculate yearly ward-level averages u τ w and lnp τ w for each year τ. Consider the following ward-level quantities: change in the share of foreign-born people: f w fw 2011 fw 2001 actual price change between 1996 and 2001, net of control variables: u w u 2001 w u 1996 w predicted change in average prices between 1996 and 2001: ln P w P 2001 w P 1996 w 24 of 28

25 Figure 9 Persistence in immigration shares The figure reports the estimated coefficient ρ from the regression: f w = α + ρf 2001 w +e w, for each of the respective countries. Error bars indicate 1% confidence intervals Spain Italy China India Pakistan 25 of 28

26 Figure 10 Forecasting immigration shares The figure reports the estimated coefficient φ from the two alternative regressions: f w = µ + φ u w +ɛ w and f w = µ + φ ln P w +ɛ w, for each of the respective countries. Error bars indicate 1% confidence intervals Risk index Stock market Spain Italy China India Pakistan Exchange rate 0.40 Spain Italy China India Pakistan Spain Italy China India Pakistan Actual price change Predicted price change 26 of 28

27 Figure 11 Forecasting immigration shares The figure reports the estimated coefficient φ from the two alternative regressions: f w = µ + ρf 2001 w + φ u w +ɛ w and f w = µ + ρf 2001 w + φ ln P w +ɛ w, for each of the respective countries. Error bars indicate 1% confidence intervals Risk index Stock market Spain Italy China India Pakistan Exchange rate 0.40 Spain Italy China India Pakistan Spain Italy China India Pakistan Actual price change Predicted price change 27 of 28

28 Conclusions The paper proposes a novel empirical identification methodogy to document the influence of international macroeconomic and political developments on the London housing market. The estimation results indicate economically large, statistically significant and robust spillover effects along two parallel transmission mechanisms, while also highlighting the differential impact of foreign shocks across property categories. The analysis incorporates two micro-level datasets on mortgage loans and house purchase transactions, includes hedonic property characteristics and controls for regional heterogeneity through a set of social, economic and demographic characteristics. We show that pre-2001 price developments predict post-2001 migration patterns at ward level. This provides additional evidence on the mechanisms likely to play a role in the international transmission of financial and political shocks. Work in progress: further understanding of the relationship between immigration and house price appreciation analysis of regional spillovers and diffusion patterns of house prices 28 of 28

29 Appendix

30 Table 2 Estimated hedonic effects - Nationwide sample - Bedrooms Construction date Two bedrooms 0.170*** 1900 to * Three bedrooms 0.229*** 1920 to *** Four or five bedrooms 0.277*** 1940 to *** More than five bedrooms 0.380*** 1960 to *** 1980 to *** Bathrooms after Two bathrooms 0.062*** Floor area Three bathrooms 0.022*** 50 to 70 m *** More bathrooms to 90 m *** 90 to 110 m *** Garage 110 to 130 m *** Parking space 0.021*** 130 to 150 m *** Single garage 0.052*** 150 to 170 m *** Double garage 0.083*** above 170 m ***

31 Table 3 Estimated effects of property and contract types Nationwide sample Land Registry sample Semi detached house *** Semi detached house *** Terraced house *** Terraced house *** Detached bungalow Flat *** Semi detached bungalow *** New property 0.243*** Purpose built flat 0.019* Leasehold tenure *** Purpose built maisonette *** Flat conversion *** Maisonette conversion *** New property 0.136*** First purchase *** Leasehold tenure ***

32 Figure 12 Summary statistics - mean values in respective top quartiles - Population density Market share of flats Sth. Europe Middle East Russia China East Asia Average Sth. Asia UK no./ha Russia Sth. Europe China Middle East East Asia Average Sth. Asia UK percent Note: The figures report mean values for selected variables, calculated for the wards in the top quartile of the respective distributions, according to the share of people born in our set of country regions.

33 Figure 13 Summary statistics - mean values in respective top quartiles - Net income Cars per household East Asia Sth. Europe Russia Average China UK Middle East Sth. Asia /week UK Sth. Asia Average East Asia Middle East China Russia Sth. Europe no./hh

34 Figure 14 Summary statistics - mean values in respective top quartiles - People in higher professional occupations Fraction of mortgage holders East Asia Sth. Europe Russia China Average Middle East Sth. Asia UK percent UK Average Sth. Asia East Asia China Russia Middle East Sth. Europe percent

35 Figure 15 Safe haven effects across price categories - Land Registry sample People of foreign origin Southern EU China South Asia East Asia Middle East Russia Net average income Southern EU China South Asia East Asia Middle East Russia Bottom third Middle third Top third