ERES / CUREM SEMINAR, ZURICH
Diversification and Real Estate Allocation
How Effective is Sectoral and Regional Diversification?
Professor Colin Lizieri
University of Reading Business School
Real Estate Portfolio Structure
General Issues and Problems
• We Need to Diversify Away Asset Specific Risk:
- Build a Diversified Portfolio Within Each Asset Class
- Real Estate Characteristics Make This Problematic
- Need a Framework for Making Fund Decisions
• We Need an Operational Structure:
- Sector Specialists, Area Specialists
- Management and Investment Teams
• Need criteria for benchmarking performance:
- How Well Has Our Fund Done?
- Is That Because of Fund Structure or Stock Selection?
Sector and Geography:
The Standard Approach
• We Separate Real Estate into Sectors:
- Office, Retail, Industrial, Residential …
• Or Sub-Sectors?
- Shopping Centres, Retail Warehouses, Standard Shops …
• We Separate by Geography:
- Global Regions: Europe, Asia, North America
- Countries: Switzerland, Germany, UK
- Regions: Fribourg, Bernese Oberland, Schweizer Mittelland
- Cities: Zurich, Genève
ve, , Bern
- Sector / Region Splits for Strategy and Management
- Cannot Segment Too Far …
Is Segmentation Useful?
• Segmentation is Useful IF AND ONLY IF:
- Properties WITHIN a Segment Behave in a SIMILAR Fashion
- Properties in DIFFERENT Segments Behave DIFFERENTLY
- Segments Behave in DIFFERENT Ways
• IF This Holds Then:
- Segments are Useful in Shaping Strategy
- Segments are Useful in Benchmarking Performance
• But ….
- “All Properties are Unique”
• We need to TEST the validity of Sector-Geography Splits
UK Evidence …
The Value of Segmentation:
• UK Commercial Real Estate:
- Three Standard Sectors (Office, Industrial, Retail)
- Regions – 10 Government Regions
- “Super Regions” London, Rest of South East, Rest of UK
- IPD Portfolio Analysis System Segments:
Standard Retail, SE
Standard Retail Rest UK
W End Offices
Rest UK Offices
Industrial – Rest UK
What We Know So Far …
• Byrne & Lee (various)
- Even at town level, sector dominates geography
- Limits to risk reduction due to common structure, size effects
• Cullen (1993)
- Industrial returns homogenous
- Retail property structured by ownership and lease terms
- Offices have geographical sub-structure, structure, City distinctive
• Hamelink et al. (2000), Hoesli et al. (1997)
- Importance of sector clear
- London offices, Scottish retail different
- Economic regions more effective than standard regions
• MacGregor & Schwann (2003)
- Strong common cycle, separate retail cycles, London different
Devaney & Lizieri
UK Segmentation Tests: Research Strategy
• Take Individual Property Returns from IPD
• Collect Characteristics of Those Properties
• Test Whether Return Patterns Fit Sector-Region Groupings
• Cluster Data and Characterise the Groupings
- do they map onto sector-region region groups?
- can we identify other dimensions of structure?
• Do Other Dimensions/Variables Explain Return Patterns?
• Sub-set of IPD databank
• Properties with continuous returns 1995-2004
• Property data covering:
- Region, town, micro-location type
- Size (floorspace, value)
- Yields, reversionary potential, cap-ex
- Tenancy information (unexpired term, multi-let)
• Final dataset 1,219 properties
• Commercial constraints limit scope of analysis
Methods … the Technical Bit …
• Discriminant Analysis:
- Properties are assigned to their Segments
- Returns are analysed statistically
- Do returns predict the Segment well?
• Cluster Analysis:
- Create groupings based on patterns of returns
- Are these groupings consistent with Segments?
Three Sectors ... Three Regions
• 3 Sectors: 69% buildings correctly grouped
• 75% of retail buildings correctly grouped
• But high within group heterogeneity
• 3 Regions: just 49% correctly classified
• London 58% correct
• Geography alone insufficient as structure
Rest of UK
Rest of UK
Sector-Super Super Region and PAS Segments
• Three by Three Sector Super-Region
- Just 35% of buildings correctly classified
- Highest group 54% - Rest of UK Offices
- Worst group 20% Rest of SE Industrial
- 76% of offices, 70% retail correct into sector
- 61% of London classed as London
‣ Questionable Value for Strategy or Performance Measurement
‣ Sector Splits Still Dominate
‣ London is Different
• 10 PAS Segments
- < 35% of Buildings Correctly Classified
- Useful Segments:
o Retail Warehouses (64%), RUK Offices (62%), City Offices (62%)
- Almost No Information Content:
o SE Retail (15%), RUK Industrial (18%), SE Industrial (28%)
‣ Sector Still Dominates: Geography Not Helpful
‣ Validity / Usefulness of Sector-Region Classification Cast
into Doubt for Portfolio and Benchmarking?
What’s s Missing Then?
What Else Might Explain Variation?
• Yield (Cap Rate)?
- Assign Properties to High / Medium / Low Yield
- 61% Correct (75% of Low/High)
- 50% Correct from 3x3 Sector-Yield Grouping
- May imply a Growth-Value distinction?
• Tenancy Characteristics
- Single-Let versus Multi-Let?
- Improves Predictive Power of Other Factors
- Weak Size Effect Emerges
‣ Some Evidence of Segmentation by Income Characteristics?
Thinking Outside the Nation State …
• The UK is a Small Country …
- Weaker regional effects outside London?
• Cross-National Diversification?
- Nation Specific Factors?
- But Convergence Effects?
- Equity Market Shows Sector Dominates …
- But Real Estate IS More Local in Nature
- We Invest in Cities Not in Countries!
• Are There Global / Regional Sector Effects?
- Business Service Demand?
- Consumer Expenditure?
- Industrial Production?
Some Investment Implications
• Do Standard Segmentations Really Reflect Different
Patterns of Return?
• Attribution-Based Performance Measures Based on Faulty
Segmentation May Mislead
• Results Suggest:
- Sector is More Important than Geography …
- Cashflow Characteristics are Significant …
• Strategic Implications:
- What are the Economic Fundamentals Determining Cashflow
- These Will Determine Diversification
• Fund and Vehicle Structures
- Economic Logic is for Specialist not Diversified Vehicles
- But … Specialist in What??