Real estate diversification : does it pay-off to be diversified during the financial crisis? : case of REITs in US

(1)

Real Estate diversification: does it pay-off to be diversified during the financial crisis? Case of REITs in US.

By

Katarzyna Ewa Szypowska

Submitted to the Program of Master of Science in Business Economics Real Estate Finance

At the

University of Amsterdam

(2)

ABSTRACT

This paper deals with the recently widely spoken topic of real estate diversification opportunities in real estate. Modern portfolio theory proves successfully that diversification brings benefits, applying that via a scope of geographic and property type diversification we hypothesize that diversified REITS will over perform specialized companies during the period of recent financial downturn. For the purpose of analysis we use data on Real Estate Investment Trusts in U.S.A covering the period from January 2003 to December 2011 (monthly). We use CAPM and Carhart four factor model to check for abnormal returns. This study finds however no evidence of diversified REITs performing better during the financial crisis, contrary specialized REITs outperformed the diversified REITs during the financial crisis period of 2007-2009, as well as during the whole period considered 2003-2011. Further analysis points to legal US REIT regime constraints as well as property market polarization, high barriers to entry and liquidity as the reasons for limitations in diversification scope achievement by non-specialized REITs. Moreover, in the eyes of investors REITs should provide simple and stabile investment vehicles, diversification not bringing operational synergies is redundant.

(3)

List of exhibits and tables

Exhibit 1. Size of the U.S. Equity REIT Sector, 1971-2005. Table 1. Top - Commercial Property Markets

Table 2. Top – 15 Listed Property Markets Table 3. Selected REIT Regimes

Exhibit 2. Systematic vs. unsystematic risk

Exhibit 3. The DiPasquale-Wheaton 4 Quadrant Model

Exhibit 4. Source - National Council of Real Estate Fiduciaries. The Russel-NCREIF Property Index Regional and Divisional Subindex

Exhibit 5. Salomon Brothers Eight-Region Segmentation Exhibit 6. UK house prices and lending (2000-2008) Table 4. Projecting a housing supply-demand equilibrium Exhibit 7. House market recovery predictions

Table 5.US housing market, recovery as a regional event Exhibit 8.Foreclosure discounts

Exhibit 9. Dow Jones Equity All REIT index

Table 6. Full period regressions of total, specialized and diversified portfolios for years 2003-2011 using CAPM and Fama-French three-factor model with momentum fixed and random effect, Hausman test, t-test (parenthesis).

Table 7. Subperiod regressions of total, specialized and diversified portfolios for years 2003-2007 using CAPM and Fama-French three-factor model with momentum fixed and random effect, Hausman test, t-test (parenthesis).

Table 8. Subperiod regressions of total, specialized and diversified portfolios for years 2007-2009 using CAPM and Fama-French three-factor model with momentum fixed and random effect, Hausman test, t-test (parenthesis).

(5)

Table 9. Subperiod regressions of total specialized and diversified portfolios for years 2009-2011 using CAPM and Fama-French three-factor model with momentum fixed and random effect, Hausman test, t-test (parenthesis).

Exibit 10. Geographic diversification (Herfindahl Index)

Exhibit 11. U.S. REITs vs. NCREIF percentage of portfolio exposure Exhibit 12. Unlevered REITs return measure those of private real estate

Exhibit 13. Correlations of total returns rolling quarterly over increasing hold periods (Dec

1977-Dec 2011)

Exhibit 14. Comparative Total Return Investment Performance: Equity Real Estate (Public

(6)

Chapter I: INTRODUCTION

1.1 Introduction

This thesis shall attempt to provide a comprehensive discussion on the benefits of real estate diversification. Modern portfolio theory proves successfully that diversification brings benefits, applying that via a scope of geographic and property type diversification we hypothesize that diversified REITS will over perform specialized companies.

The cyclicality in real estate is a big phenomenon observed in real estate with fluctuations bringing great gains or great loses to those who mastered the knowledge on how to use these instruments. For several decades real estate has been highly valued by institutional investors and held as a integral component of their portfolios due to its steady and predictable appreciation over time, inflation-hedging capabilities and its low correlation with other asset classes and across national boundaries, its strong performance in comparison to stocks and bonds (Hudson-Wilson, Fabozzi and Gordon 2003, Chin et al. 2007).

The choice of REITs has been made due to their specificity and behavior similarities to stocks and bonds but still following the trend of real estate, with a time lag – REITs react faster to market changes than direct real estate measured by the transaction. The choice of USA as the market has been due to the size of both the commercial and private real estate market as the US represents 41.68% of the world property portfolio. The six largest property markets, the US, the UK, France, Australia, Hong Kong, and Japan, represent over 89% of the world property portfolio. Due to such a big market and relative uniformity combined with no tax and policy changes between countries the choice is highly plausible (GPR 250 Property Securities Index).

(7)

Exhibit 1. Size of the U.S. Equity REIT Sector, 1971-2005.

The main advantages of investing in REITs vs. in real estate that will be further built on are:

• Avoidance of double taxation, high cash dividends

• Less liquidity risk as compared to direct real estate investments

• Diversification possibility, investors can easily diversify their exposure to real estate both geographically and by property type

• Attractively low correlation levels with equities compared to other asset classes

• Higher possibility of raising debt/equity when opportunities arise on the market than when direct property ownership

Unlike the equity market the property market is not really liquid. Very frequently during the economic growth period the house prices appreciate in value above the fundamentals (bubble) rents boost up. Many homebuyers find themselves in negative equity position and are unable to exit the market without a significant loss. Interestingly Krainer (1999) suggests that house prices are rather sticky they do not

(8)

adjust much downwards to provide the sellers with liquidity, during the periods of high housing service flow sellers also do not raise their prices to take the full advantage of the market increase. Listed real estate companies to a great extend reduce this illiquidity and market inefficiency issues, as one can easily get an exposure to real estate market by buying company shares rather than the property itself.

Having agreed upon the superiority of REITs to direct property investments, the main focus of our paper will be directed towards the choice of underlying portfolios and its performance. Among many categories we can split the REITs sector between diversified and specialized companies. In the light of diversification benefits and possible abnormal returns achieved we would like to investigate whether during recent financial crisis and such adverse economic conditions the diversification strategy actually occurred to be more favorable. Considering both the variations of any particular property category e.g. might happen that when retail property is performing poorly, logistics sector will pick up due to the growth of on-line shopping. Different sectors will present different volatilities and returns widely depending on the country and correlations between countries it is invested in. This considering especially the fact that there is a variation in response to the fundamental economic conditions between the different property sectors and the geographic location (Eicholtz and Hoesli, 1995), Englund and Ioannides (1997) shows that even though there is interdependence on descriptive grounds across different countries, only weak evidence exists to support an international house price cycle. Moreover, different characteristics of the markets will play a role, taking the example of Hong Kong property market where the government has monopoly on land supply and power to impose large restrictions of residential sector.

Hence, our research questions is:

Did the diversified REITs perform better during the financial crisis?

Although we expect the returns on assets to be correlated during the crisis period, we still assume that there are time lags between different locations, and then different asset classes will have different betas and market cycles would differ. Each of the property types is driven by different market forces, retail would be very much dependant on the consumption, offices on employment, logistics driven by supply

(9)

chain reconfiguration, consumption and global trade (drivers more structural than cyclical), thus the economic impact should be also differentiated among them.

Furthermore as concluded by Maier (2009) and Gatzlaff (1995) that real estate market is an inefficient market – characterized by illiquidity, limited market activity, asymmetric information, incorrectly priced properties - which can bring other arbitrage strategies apart from diversification.

The remainder of this paper is organized as follows in chapter two we give the key understanding issues behind the structure and relevance of REITs in our thesis, further the literature review will critically discuss the previous research conducted and apply it in the context of recent financial crisis. Chapter three will focus on the methodology applied that will be followed by presentation and evaluation of the results. As a concluding note the future research direction will be suggested.

CHAPTER II: OVERVIEW OF RELEVANT ISSUES

2.1. Understanding REITs

‘REITs smell like real estate look like bonds and walk like equity. ‘ Greg Whyte, Analyst, Morgan Stanley.

This chapter aims to provide the reader with though understanding of REITs, focusing especially on the characteristics that differentiate them from direct real estate investments. Furthermore some crucial details of the regime will be given and the size of REITs market and its significance will be highlighted.

Investment in real estate through REIT ownership does not require the large and long - term financial commitment typical of other real estate investment alternatives. As majority of REITs stocks are publically traded, the transfer of ownership is very easy. Therefore, REITs provide a mechanism to pool resources that enables investors, especially small investors, to gain the economic and other benefits of commercial real estate investments. In today’s illiquid real estate market, ‘REITs have attracted more and more attention in liquid real estate investment vehicles, even from institutional investors such as pension funds’. (Han and Liang, 1995, p. 235)

(10)

Traditionally, the only option available to investors needing to gain exposure to domestic or international commercial real estate was through direct (physical) real estate investing - that is, by means of purchasing the actual physical property. The high costs of acquisition and of setting up a team to manage the property, coupled with illiquidity, made it difficult if not impossible for smaller investors and pension funds to gain exposure to commercial real estate.

Since the advent of securitization in the early 1960s and especially over the last decade, REITs (real estate investment trusts), REOCs (real estate operating companies) and private indirect vehicles have emerged as viable alternatives to domestic commercial real estate ownership, making real estate available, albeit indirectly, to a wide audience. With institutional investors' increased appetite for these securities, foreign countries have also started to introduce tax-transparent REITs or REIT-like structures throughout the world, thereby fostering the growth and indirectly promoting the transparency of the global real estate securities market. This growth is evidenced by the fact that, according to NAREIT, the global market capitalization of publicly traded property securities has grown 170% from approximately $350 billion to $945 billion over the 7-year period beginning January 2000 and ending March 2007.

Due to its tremendous growth and strong risk-adjusted performance over time, the securitized property sector has gained widespread recognition as a distinct asset class that deserves permanent allocation in a real estate only or multiclass portfolio (Idzorek, Barad and Meier 2006). As a result, institutional investors and pension funds have become gradually more interested in increasing international real estate allocations in their portfolios (Dhar and Goetzmann 2006, Idzorek, Barad and Meier 2006).

(11)

Table 1. Top - Commercial Property Markets

(12)

Table 3. Selected REIT Regimes

2.2. REAL ESTATE DIVERSIFICATION IN LITERATURE

It has been generally accepted that real estate as an asset class with specific risk/return characteristics when added to a portfolio of stocks and bonds should improve the risk/return profile. The reasoning behind this is its low correlation with other assets together with the potential of providing a hedge against both expected and unexpected inflation, coupled with relatively stable returns Seiler, Webb and Myer (1999).

The literature reaches no clear consensus as to the benefits of real estate diversification, however there is more supportive than non-supportive evidence in favor of it (Wilson and Zurbruegg, 2003). The distinction can be marked between direct and indirect property diversification. The primary weakness for direct real estate investment being the absence of good quality information on price caused by the lack of transactions and confidentiality agreements. The result of this is the usage of appraisal-based indices for performance measurement; these are however biased as appraisers trying to extract the most efficient information from the market usually use the combination of past transactions and current market information; causing strong autocorrelation within the series. Although the ‘de-smoothing’ procedures have been developed to restore the volatility in the data (Geltner et al, 2003), it still involves large subjectivity in the selection of the smoothing coefficient and does not leave the

(13)

issue completely unresolved. Further, direct property ownership limitations include: large upfront investment, low liquidity, high transaction costs, and maintenance expenditure coupled with the need of local market knowledge and management requirements.

It is not rare to see some vacant premises usually owned by private parties or passive investment funds remaining is such condition due to the lack of active management and market knowledge that is possessed by the active players (e.g. exclusivity leasing arrangements). The indirect real estate investment is supposed to eliminate such limitations to a large extent by resolving the two issues market size (markets are large enough to absorb substantial amount of capital) and liquidity (assets can be sold quickly if there is a need to do so, e.g. when markets are heading down) Eichholtz et al. (1999). Yet we can argue some limitations faced by direct real estate investments are mitigated but not completely resolved by indirect investments. International diversification although practically out of scope for smaller investors trying to get exposure by direct property acquisition might be of a high risk and leave limited possibilities for expansion also for the listed real estate companies. In general real estate is considered to carry a higher risk as compared to other asset classes. One of the most essential risks to be accounted for is the political risk. Geurts and Jaffe (1996) define it as the ‘probability of economic losses due to government actions that could hamper, curtail or preclude investment projects’ that further includes the possibility of ‘unfair administration laws, the lack of law enforcement, corruption levels, nationalization and expropriation threats’. It is plausible that economies of scale play here a significant role and for those listed companies who are global and widely present international expansion and exploitation of market inefficiencies to achieve higher gains and better market exposure is much more attainable and appealing. However for a smaller size company with lower market knowledge will remain of a too high cost and risk that might not be compensated by returns. Yet due to such risks we might argue that many market inefficiencies are not as well exploited as in the case of other asset classes, as even these giant real estate players would have their risk/return requirements that would prevent them from more risky market investments. This might to an extent imply that the real estate market integration is higher as some of the markets still remain largely unexploited while the real estate players are widely present among the largest securitized markets (USA, Canada, UK, Australia) eliminating the inefficiencies and increasing correlations.

(14)

Literature also distinguishes between real estate only and mixed asset portfolio investments (Worzala and Sirmans, 2003). In general inclusion of real estate in mixed-asset portfolio is considered due to its risk reduction (inflation hedging) qualities that it is enhancing capabilities, with highest value to those investors with longer holding periods (Lee and Stevenson, 2006). Gordon et al. (1998) quantifies the added value of international property securities to the portfolio of mixed-asset stocks bonds and domestic real estate, based on GPR quarterly return series for 14 countries between 1984 and 1997. They concluded that the minimum variance portfolio standard deviation decreased by 46 basis points when international real estate stocks were included in the investment opportunity set. And it further fell by 245 basis points at higher return levels. These come in line with research published by Eichholtz (1996) who proved that correlation coefficients between countries for property investments were significantly lower than for stock and bond investments, by examining property, bonds and stocks performance during the period 1985-1994. He further compared efficient portfolios and found that an internationally diversified property portfolio outperformed a domestic portfolio in the UK, Japan, US and France. He also stated that the internationally diversified property outperformed both an international stock and bond portfolio. It is plausible that the benefits of the negative correlations between asset classes exist and can be exploited, the big issue to be considered is their stability - ‘correlation and its relevance to the diversification issue is clear – since correlation coefficients are temporally unstable, a well-diversified portfolio initially selected through correlation analysis in one period may not hold in subsequent periods, possibly leading to far less diversification benefits than originally anticipated’ (Wilson and Zurbruegg, 2003, p.25). The great majority of research is rather skeptical and tends to show that asymmetry in the correlations between better times and worse times exist ‘(…) when you need diversification, you don’t have it, and you get it when you don’t need it’ (Patel and Sarkar, 1998). Both research done by Lu and Mei (1999) and Conover et al. (2002) supported the above conclusion. The former explored the return distributions of property shares in emerging markets during the period 1973 to 1998, tracking them via US NAREIT and S&P 500 index and comparing them between better and worse quarters (Asian Crisis). The latter studied the period between January 1986 and June 1995, covering the period of 1987 crash; S&P Global Vantage and NAREIT. The correlation between US

(15)

stocks and foreign real estate for the full period was 0.59. These however picked up highly to about 0.70 at the time of the 1987 crash and further dropped to as low as about -0.10 in 1993.

These studies carry some constraints; the major one being the rather limited statistical power of correlation coefficients, which as argued by researchers, should be replaced by cointegration analysis. Further, none of these concentrates on better exploiting the diversification possibilities within the real estate class. Gyourko and Nelling (1994, pp.3) ‘systematic risk tends to vary across the firms depending on the types of properties they own. In particular, a REIT owning only retail properties tends to have a beta almost 50% larger than that of REIT specializing in industrial properties’, this being compensated by higher returns. We can thus argue that due to the different risk/return profiles within real estate asset classes as well as regional economic differences, real estate diversification might improve the investment performance during the downturn markets.

Our research will concentrate only on real estate as asset portfolio. The below table lists three articles on the performance of specialized vs. diversified REITs that have been identified as the most adequate for the purpose of this research.

(16)

Moreover Capozza and Seguin (1999) provide some more analytical insight into diversification by examining the relationships among focus, cash flows and firm value. They conclude that diversification even within a single industry, reduces value. They show that less focused trusts actually earn higher gross yield from their properties, ‘where yields are calculated relative to real-estate-market values for the properties’. However, the higher gross-cash-flow yields are offset by higher corporate – level expenses (interest costs and especially general and administrative expenses) for more diversified trusts. They also point to the effect of focus on the informational asymmetries that cause the equity of diversified firms to be less liquid. The latter are in line with Danielsen and Harrison (2007) research outcome that REITs focusing their investment activities within a single property type sector enjoy enhanced liquidity and ease of valuation. An interesting paper is also written by Kim et al. (2002) who investigated the performance of hotel real estate investment trusts (REITs) over the 1993-1999 period in comparison with the overall market and six other REIT sectors. His results indicate that hotel REITs carried the highest market risk as compared with other REIT sectors and it underperformed office, industrial, and diversified REIT sectors. Benefield et. al. (2008, p. 77) in his concluding remarks and consideration for further research, points to the fact that the differential performance of property - type diversified and property - type specialized REITs in overall rising markets versus overall flat or declining markets should be investigated. This establishes a further prove of validity and importance of our research and underlines its importance not only for the sole purpose of the academic research but also for the interest of the asset/portfolio managers.

(17)

Bers and Springer (1997) claim that sources of cost efficiencies, which can result in an economy-of-scale, can be categorized as either internal or external to the firm. A firm can achieve internal scale economies by specializing. For example, REITs that focus on a single property type or that geographically concentrate their assets would expect cost efficiencies. External economies-of-scale arise when the prices of a firm's inputs are reduced. In the case of REITs, larger REITs may achieve a level of market power, perhaps in financing or management contracts, that results in cost economies unavailable to less capitalized REITs.’ A REIT can diversify by mode of investment; that is, investment in equity ownership, mortgages, or combinations of both (hybrid REITs). An investment in mortgages entails a different cost structure than an investment in equity real estate. REITs can also diversify geographically and across property types. Geographic dispersion of properties results in an increased likelihood of contracting for property management and in increases of costs associated with monitoring the dispersed ownerships. Diversification across property types reduces specialization and requires managerial knowledge of diverse asset types within real estate. Because diversification alters the cost structure of the REIT, it is important to test for operational efficiency of REITs in light of these imposed costs1.

Glascock and Kelly (2007) show that ‘property type effects are smaller than the country effects. Property type specialization explains only 6% of the variance of national real estate securities index returns. Hence, country diversification is a more effective tool for achieving risk reduction than property type diversification.

On the other hand, Miles and McCue (1982, 1984) find that property type diversification is more beneficial than geographic diversification within the US. More recently, Fisher and Liang (2000) and Lee (2001) support this finding. They create pure property type and geographic indexes for the US and the UK, respectively and find that property type diversification is more effective than geographic diversification.

Davis (2001) shows that diversification across property types within a single country reduces portfolio variance to 20% of the average real estate security variance, diversification across countries within a single property type reduces portfolio

(18)

variance to 11% of the average real estate security variance, and diversification across countries and property types reduces portfolio variance to 8% of the average real estate security variance.

2.3. DIVERSIFICATION AND PORTFOLIO RISK REDUCTION

Diversification of asset portfolio achieved by investment in various asset classes lowers the overall risk of portfolio. It principally relies on the correlations between the assets held within the portfolio, the more negative the correlations the better the risk elimination achieved, thus improved mean-variance efficiency as compared to holding a single asset. As shown on the below graph total risk is split between two types: systematic and unsystematic risk, here a clear distinction is needed.

Exhibit 2. Systematic vs. unsystematic risk

Unsystematic risk also referred to as idiosyncratic or specific risk is the risk specific to an individual asset held, examples when relating to real estate include location, lease type, tenant profiles of particular property. Unsystematic risk can be largely mitigated while holding within the portfolio at least approximately 30 assets that are not perfectly correlated. The degree in which the specific risk can be reduced varies based on the number of assets and the co-movement between the assets returns. On the contrary, systematic or market risk cannot be diversified as it is caused by common/economic factors such as inflation, change in interest rates, GDP growth. All

(19)

assets in the portfolio are exposed to market risk.

Markowitz (1952,1959) as the pioneer has laid the basis for the modern portfolio theory introducing the mathematical concept of investment diversification. He has presented the concept of mean-return variance stating that one can minimize the risk for a given level of expected return by selectively choosing the proportions of other assets based to on their correlation level.

Clearly one key reason why investors seek to diversify their portfolios is to gain protection against synchronized poor performance in a bear market However, as become apparent in the recent international financial crisis, the contribution of the systematic risk may increase during the market depression, consequently dragging all investments down together and curtailing the benefits of holding a diversified portfolio of assets.

While realistically the possibility of such widespread financial crisis does not entirely defeat the purpose of the diversification, the recent experiences understandable have caused investors to question the effectiveness of international diversification in providing risk reduction benefits in a down cycle. With this in mind, understanding the risk reduction achievable in a down market is of great interest these days, and the 2008 to 2009 financial crisis presents a significant historical case to examine.

2.3.1. PROPERTY – TYPE DIVERSIFICATION

The aim of this section is to have a closer look at the behavior of different property sectors over time, examine the cyclicality, response to the economic shock and thus show the rationale behind the importance of property - type diversification for successful real estate portfolio creation. We begin by examining the literature on property cycles and then study each sector separately.

Cyclicality in real estate market has been widely proven. Wheaton (1999) claims that different types of real estate have quite different cyclic behavior. For some types of property, movements are closely related to the economy. Thus the ‘cycle‘ in these property types is largely due to varying economic demand shocks. However, some types of real estate experience much longer fluctuations, which show almost no relation to broader economic cyclicality. He further claims that they depend on the elasticity of supply, development lags and the durability of the real estate assets.

(20)

Mainly three types of time lags are distinguishable: ‘the price-mechanism lag, the decision lag and the construction lag’ (Rottke, 2001).

Wechsler and Grupe (2010) provide some further reasoning for the differences between the market cycles in commercial and private real estate. Interestingly commercial real estate market cycles span from 17-18 years in terms of their duration. However, the public and private cycles are not coincident. Public equity markets show in general better efficiency and information ratio, with investors responding quickly to publicly available information and hugely anticipating future economic and market developments. Consequently, downturns in the cycle may occur more quickly in public markets and may be more prolonged in the less transparent private markets. In the last market cycle, REITs experienced a downturn of four quarters, from the third quarter of 1989 through the third quarter of 1990. Private equity funds, by comparison, suffered a downturn of nearly 3 years, from the third quarter of 1990 through the second quarter of 1993 for core and value-added funds, and through the end of 1993 for opportunistic funds.2

DiPasquale and Wheaton (1992) four-quadrant model perfectly reveal the real estate market mechanisms. It displays the interaction between the space market, the asset market and the development industry which causes ‘boom and burst cycles’. Non - economic factors: ‘Periodic over - and under building, real estate is particularly prone to such instabilities or oscillations because of its durability and because of long lag between capital demand and delivery‘ (Wheaton, 1999) - office sector is pointed as most vulnerable to these factors.

2

http://www.reit.com/Publications/~/media/Portals/0/PDF/REITsRealEstateWithAReturnPrem uim.ashx

(21)

Exhibit 3. The DiPasquale-Wheaton 4 Quadrant Model

Acknowledging that real estate fluctuations are of a great measure, value and spread, one should also acknowledge that different mechanisms stand behind and such differences should be exploited.

Apart from the nation wide factors also the property specific factors bring an influence. Dokko et al (2002) by examining property movements between 1980’s to beginning 1990’s reaches the conclusion that commercial real estate markets across various cities were not uniformly depressed, suggesting that cyclical behavior in different geographic real estate markets is not synchronous. For example, in 1987, data from Coldwell Banker show downtown office buildings in Denver and Houston had vacancy rates greater than 30% and 20%, respectively. Simultaneously, the vacancy rates in Philadelphia and Boston were less than 10%, while those in Los Angeles and San Francisco were approximately 15%. While the same data suggest that, by early 1995, Denver vacancies had declined to nearly 10%, Houston 's vacancy rates had stabilized and hovered at around 20%, and, Philadelphia and Boston vacancy rates had cycled up and then down to 15% and 10%, respectively.

(22)

Concurrently, the office markets vacancy rate in Los Angeles was increasing and peaked at nearly 20%, while San Francisco's vacancy rate dropped to a low of 11%.’ (Dokko et al) Pritchett's (1984) analysis indicates that the magnitudes of the construction cycles for office, industrial and retail are different, with office being the most volatile, industrial the least volatile and with retail placing itself in the rank somewhere in between the two. The residential construction cycles tended to be counter-cyclical, while the commercial construction cycles tended to be co-incidental with the macroeconomic cycle. Guttentag (1960) attributes the observed counter-cyclical residential construction activity as a function of credit and other resource availability to the residential building sector.

2.3.2. GEOGRAPHICAL DIVERSIFICATION

Having discussed the property-type as a powerful tool for real estate portfolio diversification, we now move to the second category being geographical diversification. Worth recalling might be here the old mantra claiming that ‘location, location, location‘ is what primarily matters for property returns following the logic of immobile assets. However, let it be examined to what extend the geographical location plays a role as a successful portfolio diversifier and whether it compensates for additional costs associated with such actions.

In general literature agrees on the fact that heterogeneous performance is to be observed between different geographical locations, showing low or negative correlation for the property returns. Hence, we might hypothesize that the growing tendency of REITs to invest in one property type might be compensated by the benefits of geographical exposure. We take it for granted the bigger the company the greater the possibilities for a wider scope of presence, both national and international. For instance we can recall the case of European real estate giant Unibail-Rodamco that after the merger has switched its strategy, now the core portfolio consists of prime location shopping centers, mainly in European capitals, that they actively manage. The diversification is purely by a high exposure to the international market. We thus put through the assumption that if the tendency for REITs is to remain specialized, then their diversification strategy should move onto the geographic expansion. An interesting question though should probably be to examine the

(23)

trade-off between geographic and sector type diversification by following the pure economic approach that costs related to each diversification should equal to the benefits obtained. Fisher and Liang (2000) state that indeed there are costs associated with diversification strategy, pointing to the ‘hard costs of developing, implementing and monitoring diversification schemes and opportunity costs resulting from changing market conditions and reduced flexibility of capital deployment‘. They conclude that sector diversification is more effective or has higher importance than regional diversification. Great majority of research conducted on that topic though uses Markowitz Modern Portfolio Theory, what makes it extremely sensitive to the period taken. They use correlation matrices and min-variance, efficient frontier, which are very difficult to compare between each other and to measure the improvement in performance between them.

The key aspect of running such an analysis though is to define the scope of the geographic diversification. The early literature indeed started with a pure geographical split, as illustrated below by the pioneering Russell – NCREIF four regions index (East, West, Midwest, South).

(24)

Exhibit 4: Source - National Council of Real Estate Fiduciaries.

The simplicity of this criterion has also let to its criticism. Economic diversification has been brought into life, with literature showing consensus on its superiority. The fundamental behind the use of economic diversification is that the local economic conditions are a major factor affecting real property performance, while purely geographic grouping only vaguely indentifies economic characteristics‘. It is further claimed that the ‘move from geographic to economic diversification can reduce the nonsystematic component of risk in real estate portfolio returns.‘

The Salomon Brothers Eight Economic Regions (Harzell, Shulman and Wurtzebach, 1987) is the most known economic diversification index, it divides U.S. between the following eight regions: Northern California, Southern California, Mineral Extraction, Farm Belt, Industrial Midwest, Old South, Mid-Atlantic Corridor, New England.

(25)

Exhibit 5 - Source: Salomon Brothers.

The third major theory is the Metropolitan Statistical Standard industrial classification codes, soon considered by the literature as superior. The trigger for its creation was the response to the high diversification costs. It has been pointed that market segmentation is present on the submarket level, thus possibilities exist to achieve to achieve diversification within a city, state or region.

A more microeconomic research approach is also taken by Mueller (1993), he removes the arbitrary geographic restriction and looks at the local economic drivers of individual metropolitan areas as the key determinant for more efficient diversification. The result shows that economic diversification can be more effective diversification strategy than previously used strategies that have geographic constraints‘. He further groups U.S. urban metropolitan areas into five dominant Economic Employment Categories (DECs) and five Employment Performance Zones (EPZs), interestingly he finds that e.g. for the finance and service sector none of the MSAs falls in the same economic region as specified by NCREIF or Hartzell, Shulman and Wurtzebach method. The basic evidence proves that it is the economic conditions that have a much bigger impact on real estate returns, not geographic groupings which show only a rough indication as to the economic character. Mueller (1993) hence argues that the categorization should be performed on the basis of dominant industry and patterns of employment growth. He proves both the benefit of diversification on the national and city level as successful.

Mueller (2001) in his more recent paper further confirms the superiority of economic over geographic diversification by showing lower overall and fewer statistically significant correlations between the categories. The key remark for the purpose of our research is that his economic-factor model does not work in case the real estate market itself goes out of balance: noneconomic factors like overbuilding, the economic effects of employment become less important than the constraints on new supply in given markets. He further defends the theory that fastest recovery among the markets, is observed where the employment sector growth is the strongest.

(26)

No ‘one size fits all‘ strategy for diversified portfolio exists eventually all the property diversification is very much dependent on the geographic diversification; interdependent between each other. A trade-off occurs between the benefits of wide geographic diversification and the information efficiency gained by concentrating investments in fewer geographic areas - intercity diversification may provide sufficient portfolio efficiency improvement, thus limit the number of cities to obtain maximal benefit from geographic diversification (Wolverton et al, 1995).

2.3.3. REAL ESTATE BEHAVIOR AND THE FINANCIAL CRISIS

This subsection aims at giving some background information and understanding of the nature of the economic and real estate developments during the financial crisis – time period of our analysis.

Each property type and location will observe a point in time when the market suffers, usually because of the lag in the supply demand or more from the perspective of the growth of cities, urban economics. However, this crisis was caused by the property bubble, due to the market innovations, transfer of the credit risk, availability of housing for everyone, that further spurred construction – the phenomenon of property ladder, where people do believe housing is a great future investment but find themselves heavily leveraged; non-recourse mortgage. Hence, it clearly hit the residential market, with a huge drop in prices but also it is a liquidity crisis as affecting the overall economy.

The timing is an interesting aspect as the residential market was already heavily impacted long time before the crisis hit and impacted the other sectors and brought the depression to the economy. Malpezzi and Wachter (2005) claims this feature is quite common, he points to a certain phenomenon that can be observed. Mainly that when evaluating the financial crises which are accompanying business cycle downturns, the economies suffering by largest extend experience first a sudden drop in property prices. Hence, the initial collapse of the property market can be interpreted as an indicator of the approaching business cycle trough, which usually consequently weakens banking system and further moving on to exchange rate crisis and financial

(27)

crisis, finally resulting in business cycle burst. (Renaud, Zhang and Koeberle, 1998; and Herring and Wachter. 1999). Even though this sequence does not necessarily imply a causal link, the plunge in land prices is clearly of central importance to recent Asian financial crises, particularly in Japan, Indonesia and Thailand as noticed by Mera and Renaud (2000). It has been due to the speculative boom in land markets and consequent collapse of land prices, that damaged the banking system. One can presume if the banking system has been healthy, the foreign exchange crises would have had a much of a lesser impact, and could have even resulted in Japan encountering a much quicker economic recovery process.

Liberalization of housing markets should be pointed to as the primary reason of the financial crisis 2007. It has unleashed a massive demand for credit by households and firms, however no respective increase in the saving rate followed to compensate for it. Hence, liberalization has led to expansion of borrowing opportunities. The result can be easily predicted - substantial expansion in the supply of mortgage loans in many countries (ECB, 2009; Ellis, 2006). In the longer term, the increased demand of households for more credit to finance purchases or consumer durables, increased the loan rates, leaving the bank happy to oblige (Arestis, et al., 2010).

Exhibit 6. UK house prices and lending (2000-2008) (Bone, J. and O’Reilly, K., 2010)

Two main factors that play a role in peoples’ decision to buy a house are the respective price as compared to historical movement and the availability of mortgage. When looking at the interest rates trend, they have been historically low for most of the last decade. According to Wilcoxx (2005) that above mentioned conditions and

(28)

their effects on mortgage affordability, result in the residential market following an upward trend in terms of prices. Further amplification of this behavior is attributed towards the significant reduction in the deposits requirement, thus permitting households to rely primarily on debt to finance the house purchases. The group benefiting from such liberalization the most are young households together with lower income consumers, who have had less time and capacity to collect the necessary amount of capital for the down payment. Hence, these households will be given the chance to get on the property ladder much easier and faster, increasing the homeownership rate.

The market forces reacted to the increased housing demand due to the expansion of credit availability with an increase of house prices in many countries. If we quantify this impact attributed to financial deregulation, results are appalling – house prices are estimated to have increases on average by 30 percent in the period of 1980 to 2005 in the OECD countries (OECD, 2011). Easier and higher access to capital achieved via debt financing, does not go without an echo on the real house price volatility when looking at the large sample of OECD countries. This increased house price volatility consequently destroys the macroeconomic stability and decreases the income level for some households.

Another financial development associated with the above described financial liberalization era, that need to be elaborated upon, as it played a valid role in the creation of the immense liquidity and debt of 2000s, is the banking regulation. The liberalized regulations have allowed banks to follow a much fiercer industry competition, the reserve requirements have been reduced, which spurred an increase in competition and lending in the mortgage market. Banks or any other institutions would hence be able to offer more ‘attractive’ loans, leading in the longer term to an increase in property prices (Hamnett, 2009). A new derivative of regulation called securitization has been the following step post the 1990s crisis to hugely influence the housing market. The collateralized debt obligation (CDO) can be used as the key example of the product derived post the process of securitization. Collateralized debt obligation is a wide derivative instrument which packages various kinds of debt, from corporate bonds to securities underlying mortgage to debt backed by money owed on credit cards. They are hence structured and split into tranches based on the risk-return profile, so as to satisfy the need of different investors. The biggest implication and rationale for the lenders has been to ‘clean-up’ their balance sheet,

(29)

thus allowing them to lend more, whereas the risk of lending would be spread amongst a large number of lenders. As for the investors purchasing the CDOs, they would buy a ‘slice’ of debt in a form of bond. CDOs became even more popular, as low interest rates caused investors to allocate their capital to products, which could offer the potential outlook for higher yields. The result of the securitization process can be seen as twofold. As already mentioned lenders could pass on most of the risk of the loans to the investors, what dramatically lowered the lending standards and eventually the reliability of credit rating agencies. Secondly, as the loans could be re-used providing capital to subsequent loans, more lending could take place (Mints, 2007). This leads to a rise in the demand, and consequently a rise in the price of houses.

It is plausible that house prices could not have increased so much without the latest period of excessive risky credit expansion in the mortgage market (Hamnett, 2009). BlackRock largest, multinational Investment Management Corporation published a report assessing the economic state of the housing sector in America. Based on supply-demand equilibrium they produce three scenarios (Bull/Base and Bear Case) as to the time it will take for the housing market to recover, assuming 3.8% as equilibrium vacancy rate. They consider factors such as: household growth (household formations), homeownership rates, inventory, and vacancy rate. The following graph illustrates this model. Blackrock analysts argue that market recovery will most probably follow the form of a long flat ‘U’ rather than ‘V’.

(30)

Table 4. Projecting a housing supply-demand equilibrium. Source: BlackRock, Berau of Economic Analysis, Federal Reserve, CoreLogic, Mortgage Bankers Associaton, Bureau of Labor Statistics, US Census Bureau. Note: Figures in thousands are annual3

3_{Is the Home Stretch? The US Housing Market Recovery, BlackRock Investment Institute, June 2012}

https://www2.blackrock.com/webcore/litService/search/getDocument.seam?venue=PUB_IND&source=GL OBAL&contentId=1111166111

(31)

Exhibit 7. House market recovery predictions. Soures: BlackRock, Federal Reserve, Bureau of Economic Analysis, US Census Bureau, CoreLogic and Mortgage Association4

Supporting our argument on the regional differences, we see quite drastic differences between national housing markets. Certain markets are recovering rapidly, BlackRock believes the reason for it might be that those areas that experience the steepest peak-to-though price falls appear to be in recovery mode. Among other factors are the determinants of regional house prices (employment, location, transportation, crime rate, availability of schools). On top of that the foreclosure laws seem to play a significant role. Nationwide foreclosures are selling at huge discounts. The difference between a foreclosure sale to the market price was $67,000 in February 2014, as reported by Morgan Stanley. This is equal to a 28% differential, when placed in comparison with the historic average of 5%. Numbers differ per state, more details on the below map.

Table 5. US housing market, Recovery as a regional event; Case-Shiller Regional Price Indices, May 2012; Metro Areas by Peak-to-Trough Price Changes, December 2011 Sources: S&P Indices and Fiserv, National Association of Realtors and Moody’s Analytics4

(32)

Exhibit 8 Foreclosure discounts - Source: Renwood RealtyTrac4

CHAPTER III: RESEARCH METHODOLOGY

This chapter aims at explaining the data collection process and the methodology applied for the purpose of this research.

3.1. DATA

Data on Real Estate Investment Trusts in U.S.A. is easily accessible and reliable, thus used for the purpose of this analysis. Data span used covers the period from January 2003 to December 2011 (monthly). Time - series total returns have been downloaded from Wharton database Center for Research in Securities Prices (CRSP)/ Ziman US Real Estate Data Series. It provides both return series for individual REITs trading on the NASDAQ, New York Stock Exchange and American Stock Exchange. The returns are calculated as follows4:

4_{CRSP/Ziman real estate data guide}

http://wrdsweb.wharton.upenn.edu/wrds/support/document_show.cfm?key=/CRSP%20Ziman%20Real% 20Estate%20Data%20Series%20Guide.pdf

(33)

Where, for trading day t, t-1 is the previous trading day. r(t) = security’s return for day t

p(t) = last sale or closing bid/ask for day t f(t) = price adjustment factor for day t

d(t) = cash adjustment for day t

We use monthly total returns and market capitalization, available for each company and per property type index. Data series are available both for indices (REITs/property type/Herfindahl – Hirschman Index & Concentration Ratios) and individual securities (REIT type, property type information, covering 450+ REITs). It differentiates between property type (diversified, health care, industrial/office, Lodging/Resorts, Mortgage, Mortgage backed securities, residential, retail, self – storage, unclassified) and REIT type (equity, mortgage, hybrid). From SNL the composition and geographical spread (location by state, number of properties and number of total owned sqft) of REITs has been collected to allow for further analysis. The Fama French factors to run the CAPM model have been downloaded from the website http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html. The data has been split into subperiods accordingly to the REIT market movements, as shown on the graph below:

• Full period years: 2003-2011 • Subperiod years: 2003-2007 • Subperiod years: 2007-2009 • Subperiod years: 2009-2011

(34)

Exhibit 9. Dow Jones Equity All REIT index5

3.2. Methodology

We hypothesize that the diversified REITs should not earn significantly higher abnormal returns compared to specialized REITs (the null) during the market downturn of 2007-2009. Rejection of the null hypothesis would provide evidence that diversification pay-off during the crisis. We split the portfolio into diversified and specialized REITs as classified by CRSP/Ziman database and apply the capital asset pricing model and Fama French three factor model with momentum.

The capital market theory divides total risk into two components, systematic risk and unsystematic risk. Systematic risk represents the uncertainty of future returns due to the sensitivity of a particular investment to movements in the returns of the market portfolio. Alternatively, unsystematic risk is a function of the particular characteristics of an individual company, a specific industry, or a type of investment interest. The total risk of an investment depends on both systematic and unsystematic risk factors. However, capital market theory makes the assumption that investors can diversify away unsystematic risk by holding stocks in large, well-diversified portfolios. Therefore, in CAPM, the only risk that affects the expected return on a stock (and hence the cost of equity capital) is systematic risk. The CAPM leads to the conclusion

(35)

that the equity risk premium (the required excess for a security above the risk-free rate) is a linear function of the security’s beta coefficient. The function is described by the following equation (eq1):

- monthly portfolio return at month t

- one-month Treasury bill as the risk-free rate at month t - monthly return on the CRSP index at month t

- Jensen alpha (abnormal returns) - regression parameter

- error term

Since its inception, the simple yet powerful linear prediction of the CAPM has been the subject of a large body of empirical research, and a number of studies have been published which provide both theoretical and empirical criticisms of the model (Brennan 1970&1971, Black 1972, Roll 1977). These studies show that stock returns may be related more to firm-specific variables such as size, price-to-earnings ratio, book-to-market equity ration and the leverage ratio.

A fundamental criticism of the CAPM is that the pure-form equation almost always has an intercept above the riskless rate. Therefore, the model systematically understate the true cost of equity capital for any stock having a beta below one, while systematically overstating it for any stock having a beta above one (Elton, 1994). Since real estate as an asset class tends to have a beta less than one, the CAPM is not useful indicator of the true cost of equity for the real estate companies.

Despite criticisms, the CAPM remains as the preferred method used by financial theorists and practitioners and hence will be applied for the purpose of this research. To address the limitations of this model we will apply the Fama French three-factor model with momentum, which is a multivariate extension of the Capital Asset Pricing Model. They add two additional portfolios on which the security’s expected return

(36)

depends: SMB (small minus big) and HML (high minus low). SMB is the difference between the returns on a portfolio of small stocks and a portfolio of big stocks, measured in terms of equity capitalization. The motivation for including SMB is to capture the size premium present in historical common equity returns. Many studies have proven that total returns on smaller companies have been substantially greater than predicted returns over a long period of time (Banz 1981, Berk 1995, Huberman and Kandel 1987).

The other factor, HML, is the difference between the returns on a portfolio of high-book-to-market-equity stocks and low-high-book-to-market-equity stocks. This ‘relative distress factor’ assumes that the earnings prospects of firms are associated with a risk factor in returns. Firms that the market judges to have poor earnings prospects, signaled by low stock prices and high ratios of book-to-market equity, have higher expected stock returns (hence, a higher costs of equity capital) than firms with strong earnings prospects (Fama and French, 1992).

Carhart four factor model (1997) that we employ constitutes a further extension of Fama French model as it contains an additional momentum factor (UMD), which is a long previous 12-month return winners minus short previous 12-month loser stocks. The function is described by the following equation (eq.2)

where:

monthly portfolio return at month t

one-month Treasury bill as the risk-free rate at month t

Carhart’s alpha (abnormal returns)

regression parameters

monthly return on the CRSP index at month t the monthly premium of the size factor

(37)

the monthly premium of book-to-market factor

the monthly premium on winners minus losers

Further we use two techniques to fit our models using panel data: fixed and random effects regressions. The former model assist in controlling for unobserved heterogeneity when this heterogeneity is constant over time and correlated with independent variables. Hence it assumes that the individual specific effect is correlated with the independent variables. ‘The key insight is that if the unobserved variable does not change over time, then any changes in the dependent variable must be due to influences other than these fixed characteristics’ (Stock and Watson, 2003, p.289-290). It is described by the following equation:

Yit=ββββ1Xit+αααα1+µµµµit

where,

• α1 (i=1…n) is the unknown intercept for each entity (n entity-specific intercepts).

• Yit is the dependent variable (DV) where i=entity and t=time,

• Xit represents one independent variable (IV),

• β1 is the coefficient for that IV,

• µit is the error term

On the contrary the random effect model assumes the variation across entities is random and uncorrelated with the predictor or independent variables included in the model.

‘…the crucial distinction between fixed and random effects is whether the unobserved individual effect embodies elements that are correlated with the regressors in the model, not whether these effects are stochastic or not’ (Green, 2008, p. 193).

It is described by the following equation:

Yit=ββββ1Xit+αααα1+µµµµit+

• α1 (i=1…n) is the unknown intercept for each entity (n entity-specific intercepts).

(38)

• Xit represents one independent variable (IV),

• β1 is the coefficient for that IV,

• µit is the between-entity error,

• is the within-entity error.

To decide between the fixed and random effect model we employ the Hausman test, it tests whether the unique errors (µit) are correlated with the regressors, the null hypothesis is if they are not.

Furthermore, using the data obtained from SNL regarding the geographical spread of properties by state and sqft owned, we calculate the Herfindahl index to establish the degree of concentration. Herfindahl index also referred to as Herfindahl-Hirschman index and HHI is widely applied as an economic concept for antitrust and competition law management. The formula used:

where si is the market share of firm i in the market, and N is the number of firms. It can be

interpreted in the following way:

1. H below 0.01 (or 100) indicates a highly competitive index. 2. H below 0.15 (or 1,500) indicates an unconcentrated index.

3. H between 0.15 to 0.25 (or 1,500 to 2,500) indicates moderate concentration. 4. H above 0.25 (above 2,500) indicates high concentration.

CHAPTER IV: QUANTIFYING DIVERSIFICATION

BENEFITS FOR REITs DURING THE MARKET

DOWNTURN

4.1. REGRESSION RESULTS

In the tables below summarized are the results of calculated regressions using the Stata software. They have been split accordingly to the time period considered:

• Full period years: 2003-2011 • Subperiod years: 2003-2007 • Subperiod years: 2007-2009

(39)

• Subperiod years: 2009-2011

For both CAPM (eq1) and CAPM Fama French three factor model with momentum (eq2) fixed and random effect have been calculated. Further the Hausman test has been run to determine the choice between the fixed and random effect.

Table 6. Full period regressions of total, specialized and diversified portfolios for years 2003-2011 using CAPM and Fama-French three-factor model with momentum fixed and random effect, Hausman test, t-test (parenthesis).

Total portfolio Diversified Undiversified Panel A: CAPM fixed effect

α αα α1 1.2736 (6.30) 1.2530 (1.97) 1.2757 (6.27) Cons 0.00018 (0.22) -0.0005 (-0.19) 0.00026 (0.29) R-sq 0.1959 within 0.1524 between 0.1955 overall 0.1883 within 0.0326 between 0.1886 overall 0.1967 within 0.1739 between 0.1963 overall Panel B: CAPM random effect

α αα α1 1.2775 (6.74) 1.2542 (2.13) 1.27996 (6.68) Cons 0.00016 (0.19) -0.0005 (-0.19) 0.00023 (0.26) R-sq 0.1959 within 0.1524 between 0.1955 overall 0.1883 within 0.0326 between 0.1886 overall 0.1967 within 0.1739 between 0.1963 overall Hausman test Prob>chi2 0.0110 ∗∗∗ 0.8722 0.0055∗∗∗

Panel C: The Carhart 4 factor model fixed effects

α αα α1 0.89199 (9.01) 0.85408 (11.23) 0.89591 (7.37) Cons -0.000892 (-1.07) -0.01724 (-0.62) -0.00081 (-0.93) st 0.4877 (2.23) 0.48298 (3.62) 0.48827 (1.68) ht 0.7172 (2.62) 0.8240 (1.44) 0.70607 (1.68) ut -0.3701 (-1.91) -0.35535 (-1.79) -0.37166 (-1.48) R-sq 0.2387 within 0.1699 between 0.2380 overall 0.2348 within 0.1171 between 0.2349 overall 0.2392 within 0.1764 between 0.2383 overall

(40)

Panel D: The Carhart 4 factor model random effect α αα α1 0.8934 (9.16) 0.855006 (11.31) 0.8974 (7.49) Cons -0.00091 (-1.09) -0.0017126(-0.62) -0.000823 (-0.94) st 0.4944 (2.43) 0.48116 (3.63) 0.49577 (1.89) ht 0.7014 (2.29) 0.81365 (1.42) 0.68978 (1.14) ut -0.3769 (-1.32) -0.3597 (-1.89) -0.37873 (-1.46) R-sq 0.2387within 0.1710 between 0.2380 overall 0.2348 within 0.1176 between 0.2349 overall 0.2392 within 0.1776 between 0.2384 overall Hausman test Prob>chi2 0.0001 ∗∗∗ 0.9434 0.0001 ∗∗∗

Table 7. Subperiod regressions of total, specialized and diversified portfolios for years 2003-2007 using CAPM and Fama-French three-factor model with momentum fixed and random effect, Hausman test, t-test (parenthesis).

α αα α1 0.8487(3.35) 0.90997 (7.09) 1.0137 (3.62) Cons 0.00717 (9.10) 0.00722 (1.98) -0.00100 (-1.22) R-sq 0.0928 within 0.1127 between 0.0917 overall 0.0574 within 0.5110 between 0.0576 overall 0.1063 within 0.0610 between 0.1058 overall Panel B: CAPM random effect

α αα α1 0.8527 (3.58) 0.91359 (7.19) 1.02027 (3.98) Cons 0.00712(9.04) 0.007176 (1.99) -0.00144 (-1.27) R-sq 0.0928 within 0.1127 between 0.0917 overall 0.0574 within 0.5110 between 0.0576 overall 0.1063 within 0.0610 between 0.1058 overall Hausman test Prob>chi2 0.0569 ∗∗∗ 0.8450 0.0116 ∗∗∗

α αα

(41)

Cons 0.006348 (7.35) 0.005571 (1.39) 0.006425 (7.46) st 0.38882 (8.31) 0.17159 (0.79) 0.410733 (8.81) ht 0.1764 (3.47) 0.25221 (1.07) 0.16877 (3.34) ut -0.03816 (-1.37) -0.04735 (-0.37) -0.037187 (-1.34) R-sq 0.1007withn 0.1012 between 0.0996 overall 0.0590 within 0.5329 between 0.0592 overall 0.2392 within 0.1764 between 0.2383 overall Panel D: The Carhart 4 factor model random effect

α αα α1 0.63832 (6.2) 0.82389 (4.57) 0.61947 (5.73) Cons 0.00631 (7.3) 0.005519 (1.39) 0.006387 (7.4) st 0.3917 (8.41) 0.17662 (0.83) 0.41339 (8.89) ht 0.17762 (3.5) 0.25500 (1.09) 0.16992 (3.35) ut -0.03948 (-1.42) -0.05075 (-0.40) -0.03823 (-1.38) R-sq 0.1007within 0.1011 between 0.0996 overall 0.0590 within 0.5329 between 0.0592 overall 0.1099 within 0.0790 between 0.1082 overall Hausman test Prob>chi2 0.7993 0.9993 0.5043

Table 8. Subperiod regressions of total, specialized and diversified portfolios for years 2007-2009 using CAPM and Fama-French three-factor model with momentum fixed and random effect, Hausman test, t-test (parenthesis).

α αα α1 1.51778 (9.19) 1.4467 (4.34) 1.52529 (6.79) ββββ1 -0.00461 (-2.02) -0.01009 (-1.67) -0.00405 (-1.66) R-sq 0.2172 within 0.0510 between 0.2133 overall 0.2853 within 0.6365 between 0.2850 overall 0.2124 within 0.0469 between 0.2083 overall Panel B: CAPM random effect

(42)

α αα α1 1.52271 (9.53) 1.46008 (4.53) 1.52973 (7.10) ββββ1 -0.00607 (-2.04) -0.010032 (-1.66) -0.005647 (-1.78) R-sq 0.217 within 0.0510 between 0.2133 overall 0.2853 within 0.6365 between 0.2850 overall 0.2124 within 0.0469 between 0.2083 overall Hausman test Prob>chi2 0.2169 0.1566 0.2970

α αα α1 0.948446 (2.38) 0.83114 (7.04) 0.960641(9.25) ββββ1 -0.007895 (-3.51) -0.01387 (-2.40) -0.007292 (-3.03) st 0.63728 (6.22) 0.80841 (3.10) 0.61974 (5.64) ht 0.716489 (2.23) 0.87623 (1.02) 0.699512 (2.49) ut -0.418789 (-1.10) -0.38184 (-1.03) -0.42281 (-1.44) R-sq 0.2735 within 0.1088 between 0.2699 overall 0.3812 within 0.6545 between 0.3797 overall 0.2660 within 0.1027 between 0.2624 overall Panel D: The Carhart 4 factor model random effect

α αα α1 0.947818 (2.47) 0.839621 (7.13) 0.95903 (9.32) ββββ1 -0.00878 (-3.31) -0.013879 (-2.41) -0.008235 (-2.92) st 0.65218 (6.38) 0.807179 (3.10) 0.63676 (5.81) ht 0.69796 (2.04) 0.873489 (1.01) 0.67901 (2.28) ut -0.434417 (-1.56) -0.38853 (-1.11) -0.43985 (-1.91) R-sq 0.2734 within 0.1105 between 0.2699 overall 0.3812 within 0.6548 between 0.3797 overall 0.2660 within 0.1046 between 0.2624 overall Hausman test Prob>chi2 0.0001 ∗∗∗ 0.5856 0.0001 ∗∗∗

Table 9. Subperiod regressions of total specialized and diversified portfolios for years 2009-2011 using CAPM and Fama-French three-factor model with momentum fixed and random effect, Hausman test, t-test (parenthesis).