1 / 12

JUNE 2011

JUNE 2011. Factor Modelling of UK Unlisted Funds: Panel Data Analysis of Performance Drivers Kieran Farrelly CBRE Investors & Henley Business School, University of Reading & George Matysiak Henley Business School, University of Reading. Table of Contents. Research questions and objectives

amie
Download Presentation

JUNE 2011

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. JUNE 2011 Factor Modelling of UK Unlisted Funds: Panel Data Analysis of Performance Drivers Kieran FarrellyCBRE Investors & Henley Business School, University of Reading& George MatysiakHenley Business School, University of Reading

  2. Table of Contents Research questions and objectives Sources of risk and return in unlisted funds Prior literature Data Panel unit root testing Panel regression analysis Conclusions and next steps

  3. Research Questions & Objectives • CAPM (market model) is based on the assumption that there are no additional factors present which are correlated with the market return • Inclusion of other factors has been found to better explain the cross section of asset returns • Ross (76): macroeconomic factors – Arbitrage Pricing Theory • Fama & French (92), Jegadeesh & Titman (1993), Carhart (97) : fundamental factors – value/growth/momentum • Multifactor models employed extensively in equities for risk management and performance attribution purposes • Generally the property investment industry has been unable to quantify well the key sources of risk in property portfolios • Unlisted property funds have become a significant conduit in the real estate investment landscape • Purpose of this study is to identify which direct property portfolio and unlisted fund ‘structure’ characteristics/factors explain the cross section performance of unlisted property funds • End goal is to develop a multifactor model and subsequent portfolio management tool for understanding portfolio risk of both property funds/ and funds-of-funds

  4. Property Fund Risk & Return Portfolio Structure / Market Risk Stock Risk Fund Structure Sources of Risk and Return in Property Funds • Structure (market risk): • Allocations to more volatile sectors • Macro risks • Stock risk: • Asset level (operating) leverage • Risk continuum from ground rents to speculative developments • Age, structure • Fund Structure: • Financial leverage risk where used • Vehicle characteristics: age, structure, fees/costs

  5. Prior Studies: Multifactor Modelling of Property Market/Portfolio/Fund Returns • Market Risk • Macroeconomic factors (APT): • Ling & Naranjano (90,97) – per cap consumption, real govt bond yields, term structure, unexpected inflation • Liow (94) – industrial production, unexpected inflation significant predictors of expected risk premia • Marcato & Tira (10) – GDP, stock market • Property markets • Pai & Geltner (07) location (Tier I & III location performance differential), Fuerst & Matysiak (08 ) - weighted direct market return, IPF (11) – UK region exposure, property type tracking error/concentration • Stock risk – direct portfolio assets’ characteristics • Yield – Fuerst & Marcato (09) high/low yield return differential, Bond & Mitchell (09) equivalent yield, IPF (11) relative equivalent yield • Size – Zieiring & McIntosh (99) – size positively related to risk and return, Pai & Geltner (07) + Fuerst & Marcato (09) - performance differential between asset sizes, IPF (11) – average lot size, asset concentration • Income: Pai & Geltner (07) - performance differential between assets with short/long ease lengths, IPF (11) - void rate, covenant strength, % income from top 10 tenants • Development/Vacancy: IPF (11) • Fund structure • Financial leverage: Fuerst & Matysiak (08), Marcato & Tira (10), IPF (11) all found financial leverage to be significant • Liquidity: Lee (00) found no evidence, Marcato & Tira (10) found evidence • Cash exposure : Marcato & Tira (10) • Style: Fuerst & Matysiak (08) – core/value added/opportunisitc styles impacted performance • Performance Persistence: Fuerst & Matysiak (08), Marcato & Tira (10), IPF (11)

  6. Dataset • Unique sample of UK unlisted funds • Quarterly returns from 2003:Q4-2010:Q4 • Good depth in terms of fund/portfolio characteristics (x variables) • Data runs over what we’d consider to be a full cycle • Sources: • CBRE Investors database 2003:Q4 – 2004 Q3 – collated by HSBC/IPD • IPD UK Property Funds Vision data 2004:Q4 to 2010:Q4 • Consistently collected data via quarterly questionnaire • Unbalanced panel with sample of funds with sufficient data points growing through time • Commences with data on 28 funds • Maximum of 75 funds in any given period • Large proportion of the sample are open-ended funds and would be considered as having a core risk profile • Both balanced/diversified and sector specialist vehicles UK Pooled Property Fund Indices Performance 2003:Q4 = 100 Source: IPD

  7. Sample Statistics Histogram – 3 Month Excess Total Returns Histogram – Initial Yield Histogram – Loan to Value Ratio

  8. Identifying Factors: Panel Approach • First stage of multifactor modelling is the identification of statistically significant factors • We have employed a panel data approach to do this • This approach allows us to identify and test parameters without restrictive assumptions • e.g. do investment styles have differential impacts ? • Firstly we used a number of panel unit root tests to assess whether the variables are trend stationary • We then tested for the presence of fixed and/or random effects • Fixed effects: used when we want to control from omitted /unobserved variables whose impact will differ between cases • Random effects: used when we want to control from omitted /unobserved variables whose impact will have the same constant impact but vary randomly between cases. Hausman test used to assess whether random effects are present

  9. Panel Unit Root Tests Summary • Panel unit root tests are statistically more powerful than individual unit root tests • Panel unit root tests show both yield variables and ‘number of assets’ are I(1) • Otherwise other variables can be deemed I(0)

  10. Fixed Effects Regression – 2004:Q1 to 2010:Q4 • Fixed effects regression was found to be the appropriate – model has good explanatory power • Thus there are significant differences between funds and over time periods • Not surprising given there are a range of fund structures and styles Distribution of Cross Section Fixed Effects

  11. Panel GMM Regression • As there is a lagged dependent variable (momentum) in the preferred specification we have used the Panel GMM estimator • Coefficients magnitude have changed though signs and significance remain for 4 of the variables - but %top ten tenants variable is no longer significant (but note that Arellano-Bond standard errors can be very unreliable!) • Second GMM discards this variable and significant variables remain similar

  12. Provisional Conclusions • Identified the key fundamental factors which best determine the cross section of unlisted property funds over time • Factors found amongst what we consider to be the three key sources of risk-returns in funds • Presence of fixed effects points to differences across funds and over time Next steps: • Continue to test additional factors • Creation of ‘factor returns’ via cross section regressions • Use these as a basis for estimating a factor covariance matrix which can then be used to create portfolio construction/optimisation tools • Risk budgeting via factors • These will also be used for performance attribution purposes • Estimate asymmetric impacts of factors upon performance

More Related