The Impact of Noise Trading on the - PowerPoint Presentation

The Impact of Noise Trading on the NAV Spread in REIT PricingEvidence from the New Pan EU REIT Market ERES – European Real Estate Society18th Annual Conference Eindhoven, Netherlands, June 15 - 18, 2011Michael G. Mueller

AgendaJune 17th 2011 | Mueller | TU Darmstadt | Research Center for Real Estate Business Administration 1. Introduction2. Data3. Empirical Evidence4. Conclusions

1. Introduction Problem DefinitionJune 17th 2011 | Mueller | TU Darmstadt | Research Center for Real Estate Business Administration 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 -40% -20% 0% 40% 20% 30% 10% -30% -50% -10% NAV spread REITs/REOCs Europe ( based on EPRA (2010)) NAV spread REITs USA ( based on Green Street Advisors (2011 )) The predominant explanation to NAV spreads (i . e. premia / dicounts ) implicates an effect of company-specific factors (e. g. size, leverage, focus) on the valuation of REITs Although this assertion is able to reason NAV spreads, it can only hardly reason the following characteristics: radical changes (like 2007/2008) continuous alternation h omgeneous trend The Noise Trader Model (NTM) of De Long et al. (1990 ) has the potential to reason these char- acteristics but has not gained too much attention in research regarding REITs NAV spreads and its empirical evidence "is not overwhelming" (Barkham/Ward (1999)).

1. Introduction Purpose of the StudyJune 17th 2011 | Mueller | TU Darmstadt | Research Center for Real Estate Business Administration 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 -40% -20% 0% 40% 20% 30% 10% -30% -50% -10% NAV spread REITs/REOCs Europe ( based on EPRA (2010)) NAV spread REITs USA ( based on Green Street Advisors (2011 )) This study examines to what extent empirical evidence actually supports noise trading as an alternative assertion for NAV spreads To attain more unambiguous results than previous studies, the investigation period covers the financial crisis, which is regarded as a period of high (negative) noise trader sentiment The study concentrates on the new pan EU REIT market, which is barely investigated in contrast to the large body of research regarding NAV spreads of US REITs

1. Introduction The Noise Trader Model (NTM) in BriefJune 17th 2011 | Mueller | TU Darmstadt | Research Center for Real Estate Business AdministrationThe noise trader hypothesis assumes two different categories of market participants: rational investors and irrational investors (noise traders).De Long et al. (1990) base the accordant NTM on four assumptions. Rational investors are:risk averse and have finite investment horizons. While noise trader sentiment is:stochastic and the consequential noise trader risk is systematicA fifth assumption is amended by Lee/‌Shlei­fer/‌Thaler (1991): Noise traders are predominantly invested indirectly, while rational investors prefer the (underlying) direct investments . The NTM reasons long-term mispricing as well as the continued existence of noise traders in asset markets. NTM – empirically revisable implications Taken together, the NTM results to five revisable implications:equity issues in premium-periods,mean reversion , negative long -term average , homogeneity of sentiment, correlation with other indicators of sentiment . The implications are the foundation of the following analysis. Based on a literature review, minor irregularities regarding the systematic effect of noise trader sentiment have to be expected (i. e. implication 3, 4 and 5).

2. Data Overview – the Pan EU REIT MarketJune 17th 2011 | Mueller | TU Darmstadt | Research Center for Real Estate Business AdministrationYear1)State Abbr. REIT Name Abbr. No. 2) 1969 Netherlands NL Fiscale Bel eggingsinstelling FBI 6 1995 Belgium BE Société d'Investissement à Capital Fixe en Immobilière SICAFI 15 1999 Greece GR Real Estate Investement Company 3) REIC 3) 4 2003 France FR Sociétés d'Investissements Immobiliers Cotées 4) SIIC 4) 422003LithuaniaLTCollective Investment Undertakings3)CIU3)-2004BulgariaBGSpecial Purpose Investment Companies3)SPIC3)192007Germany GEGerman Real Estate Investment TrustG-REIT32007United KingdomUK UK Real Estate Investment Trust5)UK-REIT5)212007ItalyITSocietà di Investimento Immobiliare QuotateSIIQ12009FinlandFIReal Estate Investment TrustREIT-2009SpainESSociedad Cotizada de Inversión en el Mercado InmobiliarioSOCIMI-1)2)3) 4)5) Year of national REIT legislation coming into effect. Number of companies at the end of period Q4/2010. Due to better intelligibility, the original REIT nomenclature of Greece, Lithuania and Bulgaria in Cyrillic, Hellenic and Lithuanian have been replaced by the internationally established nomenclature in English.Companies suspended from SIIC status not taken into account. Channel Islands not taken into account. At the end of period Q4/2010: Eleven of the 27 EU member states have established the legal conditions for a national REIT vehicle Starting 2003, a continuous increase of national REIT laws has been noted The total amount of REITs in EU member states accumulates to 111 companiesIn Lithuania, Finland and Spain there have been no conversions to the according national REIT status, i. e. these countries have no national REIT marketThe sample is constrained to the western EU member states with an established REIT market (i. e. the Netherlands, Belgium, France, Germany, United Kingdom and Italy)

2. Data Sample CoverageJune 17th 2011 | Mueller | TU Darmstadt | Research Center for Real Estate Business AdministrationState/REIT  Number of REITs   Market capitalization 1)       market sample perc .   market 2) sample 2) perc . NL FBI   6 6 100%   9,201 9,201 100% BE SICAFI   15 15 100%   5,255 5,255 100% GR REIC   400% 1,49800%FRSIIC 42 3993% 45,00244,62599%LTCIU --- ---BGSPIC 1900% 72700%GEG-REIT 3 3 100%   957 957 100% UK UK-REIT   21 18 86%   28,082 27,211 97%ITSIIQ 11100% 452452100%FIREIT --- --- ESSOCIMI --- ---Total 1118274% 91,17487,70196%1)2) End of period Q4/2010 in 1,000,000 EUR (mEUR), based on Thomson Financials and company reports.Due to international inter-REIT interests (e. g. REIT subsidiaries in foreign countries), the total market capitalization is less than the aggregated figures and amounts to 90,571 mEUR (87,097 mEUR in sample).The total market capitalization of all 111 REITs in EU member states amounts to 90,571 mEUR The sample covers 82 of 111 REITs (74 percent) and96 percent of the total market capitalization.France and the United Kingdom represent the biggest marketsGermany and Italy are of minor size (together only four REITs)The investigation period 2005 to 2010 covers the entire financial crisisThe total number of observations is about 800

3. Empirical Evidence 1st Implication – Equity Issues in Premium-PeriodsJune 17th 2011 | Mueller | TU Darmstadt | Research Center for Real Estate Business Administration-30%-20%-10% 0% 10% -40% NAV spread 20% 1,800 mEUR 1,600 mEUR 1,400 mEUR 1,200 mEUR 200 mEUR 2,000 mEUR 1,000 mEUR 400 mEUR 600 mEUR 800 mEUR Quarterly IPO/SEO volume ( mEUR ). Trendline . In the period from Q4/2005 to Q4/2010, over 400 single transactions affecting share capital are reported for the selected REIT sample. The data is constrained to the quarterly mEUR volume of IPOs and SEOs The trend line suggests an additional emission of ten mEUR for every percent of quarterly increase in the NAV spread. Outliers at discounts of about 20 to 30 percent result from SEOs in the first half of 2009 (in this period of exceptionally difficult credit availability and decreasing real estate values, REIT's valuation related banking covenants and tax exempt status have been threatened)Results: in line with the NTM, apart from the reasonable outliers

3. Empirical Evidence 2nd Implication – Mean ReversionJune 17th 2011 | Mueller | TU Darmstadt | Research Center for Real Estate Business Administration 0.75 - 0.75 0.50 - 0.50 - 1 - - 2 - - 3 - - 4 - - 5 - - 6 - Lags - 7 - - 8 - - 9 - 0.00 95% confidence level. Results: in line with the NTM Test for serially autocorrelation with nine time lags The sample shows a typical pattern of mean reversion: positive autocorrelation regarding short time lags and negative autocorrelation regarding longer time lags The first, fourth and fifth lag show significance at the 95 percent confidence level

3. Empirical Evidence 3rd Implication – Negative Long-Term AverageJune 17th 2011 | Mueller | TU Darmstadt | Research Center for Real Estate Business Administration 2005Q4 2006Q4 2007Q4 2008Q4 2009Q4 2010Q4 -20% 0% 30% 20% 10% -30% -40% -10% -4.62 % NAV spread . Average 2005 – 2010. The average NAV spread for the entire sample amounts to -4.6 percent. The single national markets show a broad range of average NAV spreads (from -43.03% regarding the IT SIIQ to 2.54% regarding the BE SICAFI) The irregular results regarding the single national markets can be reasoned since company specific factors affect the valuation of REITs, i. e. noise trader risk is only one factor among others that affects the valuation level average values depend significantly on the selected period of investigation Results: - entire sample: in line with the NTM - separate national markets: mixed results, but reasonable

3. Empirical Evidence 4th Implication – Homogeneity (1) June 17th 2011 | Mueller | TU Darmstadt | Research Center for Real Estate Business Administration-40%-30%-20% 10% 0% -50% Average NAV spread 30% 0 5% 25% 10% 15% 20% GE UK FR NL BE IT Volatility Established REIT market . Young REIT market . The systematic effect of noise trader sentiment suggests a homogenous impact on national NAV spreads, i. e. a: homogenous valuation level (measured by the average NAV spread), homogenous fluctuation (measured by volatility ) and a homogenous trend of the national NAV spreads (measured by the pairwise correlations of the national markets) The evidence reveals divergent valuation levels (reasonable according to the 3rd implication) anddivergent fluctuation levelsbut a homogenous trend among NAV spreads since...

3. Empirical Evidence 4th Implication – Homogeneity (2)June 17th 2011 | Mueller | TU Darmstadt | Research Center for Real Estate Business AdministrationState/REIT1)NLFBI BE SICAFI FR SIIC UK UK-REIT EU Sample NL FBI Correlation (Pearson) 0 1 0 0.947 ** 0 0.974 ** 0.652 * 0 0.927 ** Significance (one sd.)   0 0.000 0 0.000 0.028 0 0.000 N 11 11 11 9 11 BE SICAFI Correlation (Pearson)00.947**0100.973**0.58100.933**Significance (one sd.)00.000 00.0000.05000.000N111111911FRCorrelation (Pearson)00.974**00.973**010.722*00.974**SIICSignificance (one sd.)00.00000.000 0.01400.000 N111111911UKUK-REITCorrelation (Pearson)00.652*00.58100.722*100.859** Significance (one sd.)00.02800.05000.014 00.002N090909909EUSampleCorrelation (Pearson)00.927**00.933**00.974**0.859**01Significance (one sd.)00.00000.00000.0000.002 N111111911***1)Significant at the 0.05 level (one sided).Significant at the 0.01 level (one sided).Germany and Italy not taken into account (due to minor market size)…all four markets show a close correlation (at least 86 percent) with the EU sample at the highest levels of significanceThe established markets indicate a very close pairwise correlation (about 95 percent or higher) and are highly significantThe results regarding the UK (young market) show a lesser but still unambiguous extent of positive correlation (between 58 and 72 percent)Altogether: NAV spreads follow the same trend (i. e. increase/decrease homogenously) with different intensity (i. e. a heterogeneous gradient)This is in line with results from aligned fields of research (CEFs/DLCs) that reason these irregular-ities by the particular shareholder structureResults: in line with recent adjustments regard- ing the systematic character of noise trader sentiment

3. Empirical Evidence 5th Implication – Correlation with Sentiment-IndicatorsJune 17th 2011 | Mueller | TU Darmstadt | Research Center for Real Estate Business AdministrationState/REITIFOEC EA1) STX EI EA 2) EC ESI EA 3) EU Sample Correlation (Pearson) 0 0.704 ** 0 0.730 ** 0 0.706 ** Significance (one sd.) 0 0.008 0 0.005 0 0.008 N 11 11 11 State / REIT EC ESI NL 4) EC ESI BE 4) EC ESI FR 4)EC ESI UK4)NLFBICorrelation (Pearson)00.830**00.846**00.839**00.860**Significance (one sd.)00.00100.00100.00100.001N11111111BESICAFICorrelation (Pearson)00.787**00.778**00.757**00.788**Significance (one sd.)00.00200.00200.00400.002 N11111111FRSIIC Correlation (Pearson)00.766**00.767**00.748**00.787**Significance (one sd.)00.00300.00300.00400.002N11111111UKUK-REITCorrelation (Pearson)00.19600.29200.29500.352Significance (one sd.)00.30600.22300.22100.177 N 09 09 0 9 09 ** Significant at the 0.01 level (one sided). 1)Ifo Economic Climate (Euro area). 2) Sentix Economic Index (Euro area). 3) European Commission's Economic Sentiment Indicator (Euro area).4)European Commission's Economic Sentiment Indicator (national sentiment).Three different sentiment indices regarding the general economic development are selected.The EU sample is highly correlated with all three indices (at least 70 percent) at the highest level of significance, i. e. NAV spreads are highly reflective of the general economic sentimentThe national NAV spreads (established REIT markets) reveal a high correlation with the national sentiment indices (between about 75 and 83 percent) and are highly significant.The results regarding UK (young REIT market) are less distinct, but anyhow do not contradict the NTM (potentially an effect of the exceptional circumstances accompanying the market launch of the UK-REIT)Results: in line with the NTM

4. Conclusions June 17th 2011 | Mueller | TU Darmstadt | Research Center for Real Estate Business Administration The study supports noise trader sentiment as an alternative assertion for NAV spreads The results of the empirical analysis are compliant with the models implications and feature a high statistical significanceIdentified irregularities can be reasoned sufficiently, however minor refinements regarding the systematic effect of noise trader sentiment need to be consideredSince the sample period relates to the financial crisis, the regular impact of sentiment on NAV spreads may be lesser Both the predominant "rational" explanation and noise trader sentiment contribute to explain NAV spreadsImplication 1. Equity issues in premium-periods compliant 2. Mean reversion compliant 3. Negative long -term average compliant , with adjustments 4. Homogeneity of sentiment compliant , with adjustments 5. Correlation with other indicators of sentiment compliant

Contact June 17th 2011 | Mueller | TU Darmstadt | Research Center for Real Estate Business AdministrationMichael G. Mueller Hochschulstraße 1 64289 Darmstadt Phone: + 49 (0) 221 / 99 37 87 24 Web: www.immobilien-forschung.de Darmstadt University of Technology Research Center for Real Estate Business Administration michael.mueller@bwl.tu-darmstadt.de Email: Thanks for your attention !