Diversification Gains and Systematic Risk Exposure in International Public Real Estate Markets Marielle Chuangdomrongsomsuk

1
Diversification Gains and Systematic Risk
Exposure in International Public Real Estate
MarketsMarielle Chuangdomrongsomsuk Colin
LizieriDepartment of Land Economy University of
CambridgeERES Vienna 2013
2
Motivation and Agenda

• Context International Real Estate Securities
  Investment
• Cointegration Between Markets Important
• Affects the diversification benefits of asset
  class
• More Independent Markets Better Diversifiers
• Analysis at National Level What If You
  Disaggregate?
• Do results hold at sector level or for types of
  cities?
• If not, what are investment implications?
• Agenda is Boringly Conventional
• Literature, Model, Data, Results, Implications
  yada yada.

3
Prior Research

• International Diversification Literature Shifts
  from Short Run to Long Run Models
• Debate Over Whether Country or Sector Critical
• Heston Rouwenhorst, Bekaert et al., van Dijk
  Keijzer
• In Real Estate Securities
• Evidence of global real estate factor / global
  convergence and importance of regional /
  continental factors
• Growing body of literature using long run methods
  to assess benefits of international investment
• Our Paper from Wilson Zurbruegg (2003b),
  Gerlach et al. (2006) and Gallo Zhang (2010)
• We follow Gallo Zhang but add sector and city
  level analysis

4
Model Set Up

• Test for Unit Root ADF, PP, KPSS, ZA
• Cointegration Tests at Regional and Country Level
• Standard Johansen style tests
• Separate Indices into Two Portfolios
• Cointegrated and Independent
• Test Relative Performance of Portfolios
• Standard measures risk return, Sharpe etc.
• Factor models (market, size, value, momentum)
• Portfolio Risk Analysis
• Fama Macbeth two step process with rolling
  windows
• Test for differences in performance
• Systematic Risk Factors (not reported in paper
  yet)
• Repeated for Sector and City Specifications

5
For the Record

• Cointegration Tests
• Factor Models

6
Data and Transformations

• Base Data
• GPR Monthly Total Return Series 1994-2011 (and
  sub-periods)
• National Level Indices and Company Level Data
• Analysis in Logs / Log Differences and US
• Sector Level Data
• Use SNL to Obtain Company Level Sector Exposure
• Classify as Sector Specialist if gt50 Exposure
• Retail, Office, Residential, Industrial,
  (Diversified)
• Global City / Financial Centre Exposure
• Majority of Portfolio in Leading City / Financial
  Centre
• RFR, Factor Models
• US TBill, F-F factors, calculated market excess
  return, market size, HML, market momentum
  measures (annual rebalance)

7
Aggregate Results
Cointegration Exclusion Tests Aggregate Indices  Cointegration Exclusion Tests Aggregate Indices  Cointegration Exclusion Tests Aggregate Indices  Cointegration Exclusion Tests Aggregate Indices  Cointegration Exclusion Tests Aggregate Indices  Cointegration Exclusion Tests Aggregate Indices  Cointegration Exclusion Tests Aggregate Indices  Cointegration Exclusion Tests Aggregate Indices  Cointegration Exclusion Tests Aggregate Indices  Cointegration Exclusion Tests Aggregate Indices  Cointegration Exclusion Tests Aggregate Indices 
Regional (n5) r North America United Kingdom Oceania Europe Asia
L-R statistic 1 14.000 17.000 14.000 5.800 9.100
p-value 0.000 0.000 0.000 0.016 0.003
L-R statistic 2 18.000 19.000 14.000 6.100 12.000
p-value 0.000 0.000 0.001 0.048 0.002
North America (n2 countries) r United States Canada
L-R statistic 1 10.000 6.100
p-value 0.001 0.014
Oceania (n2) r Australia New Zealand
L-R statistic 1 0.660 17.000
p-value 0.417 0.000
Europe (n9) r Austria Finland France Germany Norway Spain Sweden Switzerland Netherlands
L-R statistic 1 11.000 7.200 0.240 0.670 3.300 3.100 0.024 7.300 0.006
p-value 0.001 0.007 0.626 0.413 0.067 0.078 0.877 0.007 0.940
L-R statistic 2 26.000 16.000 1.300 1.700 17.000 5.200 4.100 7.700 1.200
p-value 0.000 0.000 0.515 0.429 0.000 0.074 0.127 0.022 0.559
L-R statistic 3 30.000 18.000 4.500 1.700 17.000 10.000 7.900 7.700 1.200
p-value 0.000 0.000 0.211 0.638 0.001 0.018 0.048 0.054 0.762
Asia (n5) r Hong Kong Japan Malaysia Philippines Singapore
L-R statistic 1 11.000 1.300 6.800 5.700 1.500
p-value 0.001 0.257 0.009 0.017 0.220
The largest market-cap (n12) r United States Canada Great Britain Australia France Germany Sweden Switzerland Netherlands
L-R statistic 1 2.900 21.000 0.420 0.690 6.300 7.100 0.770 1.900 0.400
p-value 0.088 0.000 0.519 0.406 0.012 0.008 0.380 0.168 0.530
L-R statistic 2 13.000 42.000 22.000 11.000 10.000 11.000 16.000 6.000 24.000
p-value 0.002 0.000 0.000 0.005 0.006 0.004 0.000 0.049 0.000
L-R statistic 3 13.000 46.000 26.000 12.000 17.000 14.000 22.000 11.000 25.000
p-value 0.006 0.000 0.000 0.008 0.001 0.003 0.000 0.010 0.000
L-R statistic 4 21.000 49.000 35.000 18.000 24.000 14.000 30.000 12.000 30.000
p-value   0.000 0.000 0.000 0.001 0.000 0.009 0.000 0.015 0.000
Dont you hate it when people put tiny tables up?
8
Aggregate Results

• Unit root testing satisfactory
• Cointegration Tests
• Regional Cointegration Inter-Regional Dependency
• Within Region Cointegration Present (Europe
  complex)
• Exclusion Tests Identify Independent Markets
• Australia, France, Germany, Netherlands,
  Singapore, Japan
• Cointegrated markets are regionally cointegrated
  
• Portfolio Performance
• Indep. better risk-return characteristics and
  Sharpe ratio but
• Greater sensitivity to market factors, momentum
• More nuanced than a simple cointegration story

9
Fama MacBeth Results
Four-factor performance model Four-factor performance model Four-factor performance model
INDE COINT
Intercept -0.056 0.042 aINDEaCOINT 19.772
Rmt 1.519 0.297 ßINDEßCOINT 28.560
SMB -0.244 0.157 ?INDE?COINT 30.379
GMOM 0.467 -0.365 ?INDE?COINT 21.527
HML -0.218 0.390 ?INDE?COINT 74.946
MSE 0.003 0.004 MSEINDEMSECOINT 71.705
SD 0.053 0.060 SDINDESDCOINT 31.609
10
Sector Results Retail

• Reduces Countries from 19 to 13
• Country Betas are Lower than for Aggregate
  Analysis
• Evidence of Inter- and Intra-Regional
  Cointegration
• But Patterns Differ
• Independent Countries Change
• France, Germany, Hong Kong, Philippines
• Cointegrated Group More Global Characteristics
• Higher and significant market betas, momentum
  effects
• Factor models explain more variation, lower MSEs
• Independent group has significantly larger alpha

11
Sector Results Office

• Strong Common Factor High market b and average
  r
• Inter and Intra-Regional Cointegration
• Typically only one cointegrating relationship in
  regions
• Cointegrated group Australia, Germany, Spain,
  US, Canada, Japan, UK, strong common movement
• High market betas in the factor model and F-M
  analysis
• High R2 in factor models, low MSE in F-M
• Portfolio risk analysis suggests strong
  sensitivity to capital market factors risk
  premia, term structure, institutional flows
• France, Sweden, Switzerland More Independence?

12
Sector Results Global City Exposure

• In Part, a Test of Towers of Capital Hypothesis
• Betas, Correlations Lower Japan, Australia Odd
• Switzerland, Hong Kong, Singapore Independent?
• Cointegrated Group Global Not Regional?
• Factor models explain high of variation
• Betas on market index high, persistent and
  significant
• Cointegrated Group Driven By Capital Markets?
• Factor risk model shows high sensitivity to RP,
  TS, Cap Flows

13
Summary and Conclusions - 1

• Aim To Extend Long-Run Analysis of International
  Real Estate Beyond Consideration of National
  Indices
• Aggregate Results Confirm Prior Research
  Cointegration Exists, Regional Location is
  Important, Cointegration Affects Performance,
  Risk and Return
• However, City and Sector Analysis Shows that
  National Level Results Do Not Hold Consistently
• Cointegration varies by sector
• For some sectors (cities) global factors dominate
  regional
• Some markets are more local (but which markets
  varies)
• Systematic risk factors vary across groups

14
Summary and Conclusions - 2

• Results Have Value for Investors
• Greater understanding of what drives risk and
  factor sensitivity
• Need to consider sector and city exposure in
  building portfolios
• Important for fine tuning where there is a
  mandate to invest in a particular country or
  region.
• Further Work and Extensions
• Develop the factor sensitivity analysis
• More work on structural breaks and sub-periods
• Drill into the currency / exchange rate issue?
• Hold-back sample portfolio effects?

15
Diversification Gains and Systematic Risk
Exposure in International Public Real Estate
MarketsMarielle Chuangdomrongsomsuk Colin
LizieriDepartment of Land Economy University of
CambridgeERES Vienna 2013