Capital structure and the business risk-return relationship in the listed REIT market

Capital structure and the business

risk-return relationship in the listed REIT

market

MSc Thesis

written by

Daphne Esmée Penning

11031662

Degree: Master Finance and Real Estate Finance

Faculty: Economics and Business

Supervisor: Prof. Gianluca Marcato

Universiteit van Amsterdam

July 1, 2019

Abstract

This thesis investigates whether the capital structure of a firm has a significant influence on the risk and return of real estate investment trusts. The study is based on a balanced panel dataset of North American equity REITs, from 2011 to 2018. Capital structure is assumed to be two-dimensional, existing of both leverage and debt maturity. First the relationship between risk and return is investigated with a double-sorted portfolio analysis, where a significant positive relationship is detected. Double-sorting on leverage and maturity does not seem to change the positive risk effect. A panel regression is conducted to investigate this relation in more detail, by including multiple factors simultaneously. The results demonstrate the positive relationship between risk and return, whereas leverage is shown to have a negative effect on REIT returns but to have no effect on the relationship between risk and return.

Statement of Originality

This document is written by Student Daphne Esmée Penning who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

C

ONTENTS

Introduction 4

1 Theoretical Framework 7

1.1 Real Estate Investment Trusts . . . 7

1.2 Asset allocation and Portfolio risk . . . 9

1.3 Capital Structure and Return . . . 11

1.3.1 Leverage . . . 11 1.3.2 Debt Maturity . . . 14 1.4 Hypothesis construction . . . 16 2 Empirical Methodology 17 2.1 Research Model . . . 17 2.1.1 Measuring Risk . . . 17 2.1.2 Multivariate Regression . . . 19 2.2 Research Sample . . . 20

2.3 Variables and Measurement . . . 20

3 Data Description 23 3.1 Data Gathering . . . 23

3.2 Descriptive Statistics . . . 24

4 Empirical Results 28 4.1 Double-sorted Portfolio Analysis . . . 28

4.2 Multivariate Regression Results . . . 30

5 Robustness Checks 36

6 Limitations and future research 39

7 Conclusion 40

Bibliography 41

I

NTRODUCTION

Real estate investors have the opportunity to invest directly through private real estate invest-ments or indirectly through public real estate investinvest-ments, which are both contemplated as different investment vehicles. Public real estate investments are often proxied as real estate investment trusts (REITs). These funds give investors the opportunity to invest in real estate, without being obligated to buy the whole underlying asset. Several advantages are accompa-nied with REITs, as these vehicles provide high liquidity, transparency and a standard market place to its investors (Niskanen and Falckenbach, 2012).The market capitalization of REITs has drastically increased up to 2002, which has made them to one of the most prevailing real estate vehicles to invest in. This high rise has come together with high returns and has led to the outperformance of REITs relative to private real estate investments (Brady & Conlin, 2004).

General financial principles imply that high risk investments are rewarded with high returns. As of such, it is common for treasury bills to earn less high returns compared to the relatively more risky bonds or stocks. The fund managers decide upon the investment opportunities and risk taking of the REIT, however it is well know that stock volatility varies over time. Different studies have been conducted on investigating the relationship between risk and stock returns, and this positive relationship does not always seem to uphold. As REITs are traded on the public exchange market, these public real estate products are liable to capital flows of equity markets. Therefore, return characteristics of public real estate investments can be interpreted as more aligned with common stocks relative to private real estate investments (Yunus, Andrew, Kennedy, 2012). However, it can also be suggested that public and private real estate are co-integrated and act as substitutes in the long-run. Therefore, risk and return characteristics of REITs are not always aligned with the stock market.

REITs as individual investment vehicles have singular product characteristics and restric-tions. Due to their characteristics, REITs are often very dependent on external financing. ’The accessibility and usage of debt has substantially grown through the years. Factors of influence are economic growth and development, expanded financial integration and inno-vations, lower borrowing costs and more secure institutional structures. Despite the fact that many firms have de-levered their accounts as a result of the financial crisis of 2008, these firms, including public REITs, still persist in the cultivation of significant levels of leverage’ (Giacomini, Ling & Naranjo, 2017, p1). Different debt levels are dependent on the capital

structure decisions made by the REITs management. Niskanen and Falkenbach (2012) claim that these decisions are very complex but of great importance.

A wide number of research has been dedicated to leverage and its effect in the financial market (Modigliani & Miller, 1958; Fama & French, 1992; George & Hwang, 2010). As the market for REITs is somewhat new, only a small amount of existing research has investigated the effect of financial leverage on REITs. Given the significant use of leverage in real estate investments, recent studies of Giacomini, Ling & Naranjo (2015), Giacomini et al. (2017) and Cheng & Roulac (2007) have thoroughly studied the influence of leverage on returns in the public real estate market. Giacomini et al. (2015) refer to the importance of leverage in real estate and address the importance of further research on this specific subject.

As leverage and risk are theoretically associated with one another, this paper makes a contribution by investigating the relationship between leverage and both risk and return and questions whether it influences the risk return relationship of REITs. This thesis extends on the existing literature, by combining different researches and theories on capital structure and risk-return relationships. The main research question is formulated in the following way: how does capital structure influence the risk and return of listed equity REITs. Historical research has not reached a consensus to what factors can consistently explain the influence of leverage on this relationship of investment vehicles which are highly dependent on debt usage. Research has been mainly applied during downturns of the recent financial crisis. However, no research has been conducted on this relationship in the last couple of years which period is characterized by a recovering economy and significant low interest rates. This paper concentrates on the years 2011 till 2018. Therefore this thesis presents an interesting issue to the existing literature and could contribute to investment allocations.

The research question is investigated with the use of two analyses. First a double-sorted portfolio analysis is conducted in order to investigate the risk and return relationship of REITs. Capital structure effects are then controlled for, to investigate whether these impose an effect on the existing risk-return relationship. Secondly, multivariate panel regressions are conducted to analyze the effect of different factors on REIT return simultaneously.

The first chapter covers the theoretical framework, where different theories applying to capital structure and risk and return are discussed. Chapter three is the empirical method-ology, where the research method is formed and its variables are addressed. The following chapter gives an analysis of the data sample, providing more insights in the variables and their movements. Chapter five provides the results of the double-sorted portfolio analysis

and the multivariate regressions. Results will be reviewed ant interpreted in this section. Robustness checks will be discussed in chapter 6, where the robustness checks already applied in the empirical research will be considered and additional checks to investigate robustness will be provided. Following is a chapter regarding limitations of this paper and future research on the investigated research topic. The last chapter provides the conclusion on the complete paper.

1 T

HEORETICAL

F

RAMEWORK

The first chapter consists of the theoretical background of this paper. Relevant theories and former studies related to the relationship between a REITs risk decisions, capital structure and performance are discussed. Multiple frameworks are used to explain this relationship. First, the real estate investment trust as a specific investment vehicle is outlined. Thereafter, the allocation of fund portfolios and the corresponding risks are discussed. The next section is assigned to capital structure decisions. Here, the use of leverage and the maturity of external financing are considered with its relation to risk and return. Finally, the hypotheses are formed on the basis of the theoretical background.

1.1 R

E

I

T

Public real estate investments, as proxied by REITs, are firms that invest in, manage, and fund income-generating real estate properties (NAREIT, 2018). The REIT investment vehicle was established by the U.S. Congress in 1960. It was created in order to provide small individual investors with the opportunity to invest in a diversified real estate portfolio. The public real estate market has grown into an asset class which presents the possibility to expand the revelation to commercial real estate, without directly acquiring the property (Ling and Naranjo, 2002).

Public real estate investments are commonly systemized as either (listed) REITs, or as non-listed REITs. (Glascock, Prombutr, Zhang & Zhou, 2018). Non-non-listed REITs are publicly offered and registered, but they are not listed on a public exchange or secondary market. Comparable to listed REITs, non-listed REITs are assortments of different real estate assets upon which dispersed claims are issued. They are sold in the public market and are administered by the regulatory environment for all publicly registered securities, and therefore SEC registered. An additional existing REIT form is the private REIT. This is a privately issued entity or fund that fulfills all qualifications to be classified as a REIT, but is not publicly registered or listed on a public exchange (Seguin, 2016). As data availability of non-listed and private REITs is minimal with the database subscriptions of the University of Amsterdam, systematic investigation of these two REIT forms are unfeasible, and therefore not object of this thesis. Publicly listed REITs can also be further divided into three main forms; equity, mortgage and hybrid REITS. Equity REITs invest directly in real estate properties, gaining both income and appreciation return on their assets. Mortgage REITs issue loans to third parties for the

purpose of real estate investments and generate income through the interest on these loans. Hybrid REITs are a combination of both REIT forms, where direct investments as well as loan issues are performed. Equity and debt funds are complete other investment vehicles, as several different aspects and risks are associated with both funds. This thesis concentrates its analyzes on the equity REIT market, as the risk of the underlying is taken into consideration rather than the loan risk. Furthermore, REITs are to be referred to as equity REITs in the remaining of this paper. REITs are often classified by the real estate sector they substantially focus their investments on, as REITs invest most of their assets in one real estate sector. They are however obligated to make investments in more than one sector or property type (Giambona et al., 2008).

In general REITs exhibit risk and return characteristics which are comparable to those of levered investments in the underlying tangible asset. However, there are some differences that appear. REITs are traded in the public market and accordingly they present investors with higher liquidity compared to direct investments in the private market. These liquid real estate investments are liable to flows from the capital markets and therefore may also share return features that are consistent with common stocks (Yunus et al., 2012). REITs are intensively managed, as besides holding and operating a portfolio of properties by buying and selling assets, they employ in the development of real estate. Accordingly, investors will experience little management burden, because the property management is performed by the REIT management itself. This is analogous to the stock market, as similarly, investors do not have to be actively involved in the management of the firm. Therefore, REIT’s risk and returns are reflected by the risk and return characteristics applied by its management (Geltner, Miller, Clayton & Eichholtz, 2014).

Compared to most other types of stocks, REITs are contingent on a different tax regime. A remarkable characteristic of REITs is that they are exempt from corporate income tax. Internationally there are differences among the regulations and restrictions, but in general, U.S. REITs determine the standard. To be qualified as a REIT and maintain its tax exempt status, it is subject to several requirements. The five-or-fewer rule has to be fulfilled, which means that a REIT cannot be a closely held corporation and therefore should have a minimum of 100 shareholders, of which five or fewer are not allowed to hold more than 50% of the REIT’s stock. Additionally, 75% or more of a REIT’s total assets must be invested in real estate, and at least 75% of the REIT’s yearly gross income should be acquired from real estate assets (Geltner et al., 2014). Many REITs retain substantially more than this obligated threshold,

this partly results from an insufficient amount of tangible asset investment opportunities in the capital markets and has led to higher REIT demand from investors. Therefore, these commercial real estate properties are applicable and often preferred as guarantee on debt, presenting REITs with considerable leverage capacities (Riddiough & Steiner, 2015). A third requirement, which is called the ’income test’, reassures that a minimum of 75% of a REIT’s income originates from primarily passive sources such as rents, opposed to short-term buying and selling of real estate. In this way, the tax-exempt status of REITs is not violated in order to protect ’prohibited transactions’, producing non-real-estate income, from corporate taxation. (Geltner et al., 2014). Finally, every year 90% of the annual taxable income should be paid out as dividends to the shareholders of the REIT. As dividend payments are tax deductible, the vast majority of REITs distribute 100% of the annual taxable income as dividends, leading to enlargement of the shareholder value (Geltner et al., 2014). This way, it is generally not possible for REITs to hold on to much of its income and consequently REITS are essentially dependent on external debt and equity (Riddiough & Steiner, 2015).

1.2 A

P

REIT managers can chose their investment strategies and applicable risk of their real estate investments by allocating their portfolios. Adequate investment management is crucial in order to achieve returns, therefore asset allocation is a very valuable determinant of portfolio return. Business risk refers to both the systematic and idiosyncratic risks these REIT portfolios are exposed to. Where systematic risk is the risk generated by general economic conditions and idiosyncratic risk is the risk to a specific security.

Several traditional portfolio allocation theories propose that idiosyncratic risk can be considered as irrelevant as firm specific risk factors can be diversified away and therefore systematic or market risk is the singular risk component to be considered in determining expected returns. The modern portfolio theory (MPT) or equally the mean-variance portfolio theory assumes that diversification lowers return volatility. Total portfolio wealth should not be assigned to only a single investment or asset class as this will cause excessive vulnerability. Generally, diversification can be obtained by combining investment products with inverse tendencies. Portfolios combining investment products that are less than perfectly correlated will always provide a more enhanced risk-return profile than holding each product individu-ally (Geltner et al., 2014). Based on the MPT developed by Markowitz (1952) other portfolio frameworks such as the Capital Asset Pricing Model (CAPM) were developed, assuming that

no relationship exists among portfolio returns and idiosyncratic risk factors.

However, there are also studies challenging these statements. According to Fama and French (1992) who developed a 3-factor model extending on the CAPM, distinctions in cross-sectional returns are determined by both systematic risk and idiosyncratic factors such as firm size, market-to-book ratio, and prior return. Merton (1987) states that investors are not able to access all information within the market, due to incomplete information. Therefore, investors are not able to invest in the market portfolio. Not being able to hold a fully diversified portfolio will require a higher return for the corresponding idiosyncratic risk. This implies that total business risk including firm specific risk factors will have an effect on return volatility. Several studies imply that although investor’s objectives and concerns to diversify their investments, they tend to incompletely diversify their portfolios. This is partly caused by the high transaction costs (Barher Odeon, 2000; Barberis Thaler, 2003). Goetzmann and Kumar (2004) found evidence of under diversification by retail investors, as they show that around 15% of retail investors hold on average only one investment product in their portfolios and additionally the average amount of products per portfolio are only equal to three. Concentrating on the two separate risk factors within business risk, systematic risk components are of greater importance to diversified portfolios, whereas idiosyncratic risk factors are also applicable to undiversified portfolios. Empirical research analyzing the relationship between idiosyncratic risk and return volatility shows opposing results that appear to be subject to the way risk is assessed. Using the Fama and French 3-factor model, Ang et al. (2006) find a negative relation in their research based on U.S. stocks. Stocks have been separated into five divisions classified by the idiosyncratic risk. They show that portfolios with the lowest idiosyncratic risk classifications gain greater average returns relative to the portfolios with the highest idiosyncratic risk classifications.

Given the comprehensive researches on the effect of idiosyncratic risk on stock volatility, several studies have also been conducted on the REIT sector specifically. Sung and Yung (2009) conclude there exists a positive relation. However, after excluding specific observations such as small and low priced REIT stocks from the model, insignificant results are found. Investigating equity REITs, Gerlach et al. (2015) also take into consideration the effect of leverage. Using a portfolio-level analysis they initially sort the portfolios into different idiosyncratic risk profiles. Subsequently these are divided into components sorted by their debt-to-equity ratios. They show there is a significant nonlinear relationship between debt levels and the idiosyncratic risk-return relationship before and during the crisis period.

However, they do find evidence for a positive significant relationship in the post-crisis period.

Strategic asset allocation is a process within which a REIT management team determines the acquisition opportunities of real estate assets. Within the real estate markets, many diversification possibilities occur. Portfolios can be diversified by investing among different real estate sectors and by geographical spreading investments. Other important aspects, are legal structures, tenancy and building criteria. Different asset allocation strategies can lead to different investment styles among which business risks can range from conservative to more aggressive, where both systematic as well as idiosyncratic risk factors have been proven to play important roles (Parker, 2011).

1.3 C

S

R

1.3.1 LEVERAGE

The capital structure of a firm describes the amount of equity relative to the level of debt used by a firm. The fund management of each REIT can make decisions about the amount of debt to use for their investments. Several studies examining the effect of capital structure decisions on the performance of each fund are discussed. A wide number of research has been dedicated to leverage and its effect in the financial market. A classic theoretical research investigating this matter is the work of Modigliani and Miller (1958). The first proposition of Modigliani and Miller states that the addition of leverage results in tax benefits and eventually in higher risk of bankruptcy, therefore it would be assumed that no linear relationship between leverage and return exists. In the second proposition a direct connection between the capital structure of a firm and its expected returns on equity is presented. According to MM Proposition II, financial leverage has a direct negative effect on the risk of cash flows to equity holders and accordingly the required rate of return should increase. Moreover, higher leverage is attributed to higher rates of required return on equity and thus the theory indicates a positive relation between leverage and expected returns. MM Proposition II is contemplated as the primary acknowledged theory examining the relationship between leverage and returns.

Empirical research corresponds to the significance of this relationship, but the tendency of the effect of leverage on equity returns is open to question. Fama and French (1992) investigate the expected excess returns contingent on market return, size ,and book to market

factors. They find a positive relationship on the effect of financial leverage on expected returns. More recently, George and Hwang (2010) found an inverse relationship. They show that firms with high levels of leverage experience lower returns resulting from their asset base levels of risk. These results are divergent and opposing the theoretical evidence provided in MM Proposition II. In order to explain this negative effect, George and Hwang (2010) incorporate an additional determinant of financial distress, as a firm’s capital structure is subject to its distress cost. They argue that firms experiencing higher levels of distress are more likely to take on reduced levels of debt and consequently there is a lower possibility of default. But as financial distress cost seem to add to market risk, firms with low leverage levels experience a small probability of distress and are more vulnerable to business risk compared to firms with high leverage levels. Therefore they conclude financial distress costs play an important role and affect this relationship. Here, the trade-off theory comes in. The trade-off theory implies that when firms maximize the trade-off among tax advantages obtained from deduction of interest on debt funding and the cost associated with financial distress from rising leverage, the optimal leverage levels are attained (Modigliani and Miller, 1958). Firms tend to make decisions regarding their capital structures by a trade-off between the advantages of the use of leverage against the costs affiliated with financial distress.

As shown above many financial studies have been applied to common stocks, but these exclude REITs from their samples, and therefore the outcomes may not be completely appli-cable to the real estate sector. Riddiough and Steiner (2015) question whether ’one size fits all’ traditional corporate capital structure theories are applicable among different industries and therefore separate examination of the REIT industry is of importance. Additionally, an important anomaly of REITs is that they approximately exhibit twice the leverage of industrial firms and on top of that do not pay taxes at the corporate level. Nevertheless, as stated by several researchers the implications within the REIT sector are of importance and knowledge on this subject can also influence investors’ decisions on asset allocation and portfolio strate-gies, but it might also help to better understand market efficiency in this sector (Cheng and Roulac, 2007).

There is a limited amount of studies conducted to examine the effects financial leverage have on REIT returns and thereby show conflicting results. A weak, but negative relation is found between financial leverage and returns by Cheng and Roulac (2007). They conducted their research using five firm specific factors between 1994 and 2003, and found correlation in two out of the three time periods included in the study. Their analysis assembles some

appealing evidence on REITs that opposes and disputes the Efficient Market Hypothesis used in the security market. This theory asserts that in a perfectly-efficient market, public information and past performance data cannot provide a forecast on returns. Accordingly, variations among returns are attributed to risk premiums compulsory to compensate the risk taking, and therefore market risk is assumed to be the only driving factor of discrepancies among returns in this theory. While as discussed in the last section, firm specific risk factors also have a significant effect on return. In a study conducted by Pavlov, Steiner and Wachter (2015) it is investigated whether capital structure changes ahead of the financial crisis have impact on the pricing changes endured during and after the crisis. No significant results to prove an existing relationship between financial leverage and returns were detected.

Ling and Naranjo (2015) and Giacomini et al. (2015) show that additional returns of financial leverage turn out to be unproportionate to the corresponding additional risks persuaded by financial leverage. Giacomini et al. (2015) used data of in total 400 international REITs. They state that levered public real estate returns are substantially higher and more volatile compared to unlevered returns, concluding on a positive unconditional relationship between financial leverage and returns. When additionally taking into consideration default risk as indicated by Kaplan and Zingales’ (1997) financial distress measure, they show a less consistent relationship. To investigate this effect during the crisis period of 2007-2008, a dummy is included. A highly significant, but negative result is found on the relationship between the crisis dummy and the leverage factor, illustrating that higher levels of leverage in a downturn market cause larger REIT share price declines.

In another paper, Giacomini et al. (2017) investigate leverage decisions and their effect on risk and return. Using a sample of U.S. REITs they find that highly levered REITs tend to underperform REITs with lower debt levels. Taking into account the target leverage ratios as a measure of the risk taking of the firm, they state that highly levered REITs show better returns than REITs with lower debt levels relative to their target levels. In their research, Giacomini et al. (2017) conclude that higher levels of debt are used by larger REITs, while REITs with higher returns and more restrictions use less. These findings are in accordance with the trade-off theory. The conventional trade-off theory balances the advantages of the corporate tax shield against the financial distress costs of leverage, but only predict for a very low or even non-use of leverage for the tax-exempt REITs. A tax disadvantage of leverage is found when analyzing tax effects at the level of the individual investor, as returns to debt are shown to have a more efficient individual tax rate than returns to equity. Therefore, finding that the

performance of REITs worsens with the use of leverage is in accordance with the trade-off theory. But as REITs also have a very high us of leverage compared to industrial firms, the trade-off theory is not fully able to describe the capital structure of REITs (Riddiough and Steiner, 2015).

Green Street Advisors (2015), a distinguished independent research and advisory firm concentrating on REIT analysis in North America, state that delivering would cause REIT shareholder values to appreciate. As REITs have a tax-exempt status, and dividend payouts are tax deductible, cash flows might be used more to distribute dividends instead of paying down debt, imposing additional leverage risk on the investment vehicle (Geltner et al., 2014). As the existing literature is contradicting, there is no definitive evidence available on this relationship.

1.3.2 DEBTMATURITY

The level of financial leverage is often affiliated as the capital structure. However, the capital structure of firms can also be observed as multidimensional, where both the level and maturity of external financing affect the firm’s financial operation.

An alternative of the trade-off theory is the pecking order theory, presented by Myers (1984), which indeed considers the capital structure to be multidimensional. This theory assigns differences among capital structures to various asymmetric information levels that are related to several capital sources. It is suggested that distortions can be mitigated through debt maturity decisions, and accordingly the capital structure should be considered by the mutual choice of leverage, maturity and the corresponding gains and disadvantages. Capital structure theories are often not completely applicable to predict leverage and maturity decisions within the real estate sector (Riddiough and Steiner, 2015). However, REITs could be argued as rather transparent because these vehicles concentrate on stable income generation from both income from operations and appreciation yield. Therefore, the existence of asymmetric information within REITs could be debated. On the contrary, valuation of real estate investment opportunities demands sophisticated comprehension of the industry, which could respectively confirm the existence of information asymmetry. It is hence that Boundary et al (2010) find no evidence for pecking order in the REIT industry, while others do substantiate the pecking order theory in REIT capital structure decisions (How and Shilling, 1988; Brounen and Eichholtz, 2001).

of debt and maturity, as leverage with long maturities can be substituted to higher levels of leverage and the other way around. They state that long maturity and high debt levels limit the flexibility of the firm’s management to overinvest. But high components of leverage and maturity also seem to support to the problem of underinvestment, where more risky debt could pass up positive net present values of investment opportunities (Giambona et al., 2008). Alcock, Steiner and Tan (2014) investigate the interdependence of debt and its maturity. They conclude that leverage and maturity decisions of listed REITs are interdependent. When comparing these findings of REITs to industrial firms, they find that the decisions regarding leverage and maturity are not made simultaneously as within industrial firms. They come to the conclusion that REITs prioritize the maturity choices. Afterwards, the matching level of value-maximizing debt is chosen. They confirm the existence of a positive relationship between leverage and maturity. Giambona et al (2008) examine the joint determination of leverage and maturity within the REIT industry. Contrary to Alcock et al. (2014) they find that the two components are indeed determined simultaneously. In their research they also illustrate the difference regarding capital structure decisions of REIT firms and industrial firms. They state that leverage and maturity are employed as substitutes, in order to minimalize underinvestment due to lack of funding. While on the other hand industrial firms seem to utilize these components as complements in order to decrease financing risk.

Evidence suggests that refinancing risk is enhanced when taking on more leverage, and lower maturity concentration could avoid additional risks. Therefore, long maturity debt could be associated with lower debt levels (Alcock et al., 2014). Pavlov et al. (2015) define firms with risky capital structure characteristics as firms with high debt levels and a large portion of short-term debt. Sun, Titman and Twite (2015) show that U.S. REITs with high capital structure risk, as indicated by high debt levels and short maturities experience a greater decrease in their price levels during economic downturns.

1.4 H

Based on the above discussed theories and papers several hypotheses are formed. Historical theories predict a positive relationship between leverage and return within the capital mar-ket. However, contradicting results have been found in the real estate sector. Furthermore, many of these theories are not fully suitable to the REIT investment vehicle with its singular characteristics. Therefore, the first hypothesis tests the risk return relationship of REITs.

Hypothesis 1: Risk has a positive effect on the return of REITs.

Discussed theories mainly predict leverage to impose a positive effect on returns, on the other hand opposing evidence has also been found. Theories could partly predict for REITs to not take on any external financing at all, while empirical evidence has shown REITs to make substantial use of debt as these investment entities are highly dependent on external funding. In order to investigate the effect of leverage on the risk return relationship, business risk of individual funds is included. Taking into account the risk portfolios are exposed to by the investment allocations, this relationship might seem to change. As a second hypothesis, the leverage effect on the risk return relationship of listed REITs will be tested. Another very important component of the capital structure of firms is the maturity on the leverage that is acquired. Therefore, an additional regression is conducted taking into account the effect of debt maturity alongside the effect of the level of debt.

Hypothesis 2: Leverage has a positive effect on the risk return relationship of REITs.

Hypothesis 3: Short-term leverage has a negative effect on the risk return relationship of REITs.

Financial distress has been proven to be an important determinant of firm performance, bearing the possibility of default. Firms that are considered to experience financial distress on its financial obligations are more likely to be condemned by its investors. Due to the tax exempt status of REITs, it would be expected that financial distress plays an important role in capital structure decision making. Therefore, financial distress is taken into account.

Hypothesis 4: Financial distress enhances the negative effect of capital structure on the listed risk return relationship.

2 E

MPIRICAL

M

ETHODOLOGY

The second chapter specifies the methodology that is used to analyze the discussed hypothe-ses. First, the research model will be explained. This involves a two-part examination of both a double-sorted portfolio analysis and a multivariate regression. Thereafter the sample which is used and its details will be addressed. The last section of the methodology chapter describes the variables used in the research models and how these measurements have been determined.

2.1 R

M

The research model used in order to test the hypotheses of this paper is divided into two components. First the risk of each REIT will be estimated in order to implement a double-sorted portfolio analysis controlling for risk and capital structure of the firm. Thereafter multivariate regressions will be conducted in order to test multiple factor simultaneously. Details of both steps are discussed in the following sections.

2.1.1 MEASURINGRISK

As a first step the business risk will be measured using the Carhart 4-factor model. This model is often used in order to correctly capture the risk firms are exposed to, as it is able to capture both systematic and idiosyncratic risk. Carhart (1997) based his model on the Fama and French 3-factor model, which is a generalization of the Capital Asset Pricing Model, and added an additional momentum factor. According to Fama and French (1992), the Capital Asset Pricing Model fails to correctly capture the risk of stocks as beta is not statistically significant in measuring systematic risk. By including several factors such as firm size and book-to-market value, the model is more capable of measuring risk which the CAPM fails to evaluate. So in addition to the excess return on the market over the risk free rate, systematic risk is also acquired by the additional factors in the first regression model:

Ri ,t− RFt = αi+ βM K T(M K Tt− RFt) + BSM BSM Bi ,t+ βH M LH M Li ,t

+ βMOMMOMi ,t+ ²i ,t (2.1)

Here the value weighted returns of the REITs over the risk free rates are dependent on the four variables of the Carhart 4-factor model (M K T − RF ), SM B , H M L , MOM . From

the monthly error term²i ,t the idiosyncratic volatility of the REITs is then measured on a

quarterly basis by deriving its standard deviation:

σi= s Pn t =1 ¡ ²i ,t− E(²i ,t)¢2 n (2.2)

Hereσi is the standard deviation of the residuals.²i ,t is the residual found in regression 1

and n denotes the number of REIT return observations each quarter. Idiosyncratic risk is measured individually for every REIT for each period. The standard deviation of the error term will be used as an explanatory variable in the following multivariate regression models.

A double-sorted portfolio analysis is applied in order to classify both risk and leverage, based on the methodology by Gerlach (2015) and Sun and Yung (2009). This double-sorted method only controls for risk and leverage but provides the possibility to observe the perfor-mance differential among portfolios containing low idiosyncratic risk REITs contrary to high idiosyncratic risk REITs with the effect of different levels of leverage. First the idiosyncratic risk levels of all REITs are sorted into three portfolios from low to high expected risk each quarter. Then the performance of each of the three portfolios, low, medium and high idiosyn-cratic risk, are evaluated. This method is repeated for each quarter. Hereafter the effect of leverage is taken into account. Again three portfolios are formed, based on the debt levels of each REIT. The leverage of all REITs are sorted into portfolios from low to high levels of debt. Within each leverage-sorted portfolio the REITs are again sorted into low, medium and high idiosyncratic risk portfolios. This way, a total of nine portfolios are constructed. Then finally, for all three leverage sorted portfolios the average expected returns are estimated for a given level of idiosyncratic risk. In case of conducting a study with a somewhat small sample, it is important to supplement portfolio level analysis (Gerlach et al., 2015). As debt maturity is considered as an additional important component of capital structure next to leverage, a double-sorted portfolio analysis is also conducted between idiosyncratic risk and short term maturity. Here, REITs are sorted into three portfolios on short debt maturity concentration. Debt due in two to three years is considered as short maturity, where a low concentration of short term debt is sorted in the low portfolio and a higher ratios are sorted in the medium and high portfolios. portfolio. In this case, a total of nine portfolios are created as well. The portfolios are estimated and rearranged on a quarterly basis.

2.1.2 MULTIVARIATEREGRESSION

Additional variables of importance related to both risk and capital structure are also taken into consideration. Therefore, another model should be used because the double-sorted portfolio analysis cannot control for multiple factors simultaneously. A balanced panel data regression is conducted. This regression method is applied as it controls for heterogeneity of independent variables, which would otherwise not be observed by time-series and cross-sectional examination singularly. The model is based on the Carhart 4-factor model, but additional variables are included as well, based on methodology from discussed studies. The regression model is constructed as follows:

Ri ,t−RFt = β0+β1(M Kt−RFt)+β2SM Bt+β3H M Lt+β4MOMt+β5σi ,t+β6Lever ag ei ,t+ β7ST M at ur i t yi ,t+ β8K Zt− I ndext+ ²i ,t (2.3)

The main variables of interest areσi ,t(idiosyncratic risk), Lever ag ei ,tand ST M at ur i t yi ,t.

Idiosyncratic risk is measured on a quarterly basis, according to the above described esti-mations. Lever ag e_{i ,t} indicates the level of external debt financing taken on by the REITs, whereas ST M at ur i t y_{i ,t} stands for debt due in two to three years which implies the use of short term leverage. An additional variable of interest is the K Zt− I ndext, which is used to

measure financial distress. The four factors from Carhart’s model will be used again, but here these variables are estimated based on quarterly data. They are applied as control variables, indicating systematic risk sources. The explanatory variables are gradually added to the control variables, in order to be able to clearly investigate the impact of each factor.²i ,t is the

2.2 R

S

In order to test whether the capital structure influences the business risk and return rela-tionship within the REIT industry, a sample of U.S. Equity REITs is used. A balanced sample of 95 REITs is constructed with a time span from the start of 2011 till the end of 2018. The final sample consists of 3,040 quarterly REIT observations. The focus of the investigation is on North American REIT market, as it has a very active public real estate marketplace. The specialization of different real estate sectors are investigated. Therefore, REITs that specialize in one of these eight different real estate sectors are included. The different sectors covered are: Diversified, Health Care, Industrial/Office, Lodging/Resorts, Residential, Retail and Self-storage. Only listed equity REITs are incorporated in the data set. The exclusion of both non-listed REITs, private REITs, and mortgage and hybrid REITs, a very homogeneous sample remains.

2.3 V

M

The dependent variable Ri ,t− RFt in regression models represents the excess return of the

REITs at time t and is measured as the value weighted return of each REIT over the risk free rate. The risk free rate is the one month US treasury bond yield.

The independent variables are on the right hand side of the equation. al phai is a constant.

The Carhart 4-factors are used as control variables and are estimated in the following ways. M Kt− RFt is the excess return on the market over the risk free rate. The market risk is

represented by the FTSE/NAREIT All Equity REIT Index, as this benchmark is the closest to the sample used. SM Bt represents the size premium or otherwise called the “Small minus

Big” factor. Size is represented by the market capitalization relative to the median annual market capitalization, where below median firms are denoted as small and above median firms are denoted as big. It is derived by the average return on the three smallest portfolios minus the average return on the three largest portfolios. A positive small firm effect factor indicates that the smaller REITs have outperformed the larger ones. The H M Lt variable is the

value effect factor or similarly “High minus Low” factor and illustrates the average return on the difference between the value and growth portfolios. Book to market values are classified from low, medium to high. 30% of both the lowest and highest book to market values are sorted in the low and respectively high groups, the middle 40% of the values are classified as medium. A positive factor denotes that the growth REITs have been outperformed by the

value REITs. MOMt is the momentum factor, which is created by measuring and categorizing

firms based on their cumulative returns using similar sorting groups as with the value effect factor. Then the difference between the averaged two high and two low prior portfolios give the momentum factor return.

Lever ag e_{i ,t} is the ratio of total debt divided by the sum of total debt, preferred stock and market capitalization, which sums up to the market value of invested capital. ST M at ur i t y_{i ,t} represents the debt that is due in the short run and is the amount of debt due in two to three years divided by total debt, following the estimation used by Pavlov et al. (2015). The measurements for debt maturity seem to differ in the existing literature, however there is no evidence of different measurements to impact the empirical results (Alcock et al., 2014).²i ,t

is the error term, which is assumed to have a normal distribution:

²i ,t ∼ N

³

0,σ2_{i ,t}´. (2.4)

To measure financial distress, the Kaplan-Zingales index is used as an additional explana-tory variable. The KZ-index estimates the relative dependency a firm has on the use of external financing. A high score indicates that a firm has a higher possibility of suffering financial problems in case of tightening financial conditions as these firms may already have problems with financing their current operations. The KZ-index is measured following the five-factor model reported by Kaplan and Zangali (1997):

K Zt− I ndext = −1.001909 ·C ashF l ow si ,t+ 0.2826389 · Tobi n0s Qi ,t+ 3.139193 · Debti ,t

− 39.3678 · Di vi dent si ,t− 1.314759 ·C ashi ,t. (2.5)

Where C ashF l ow si ,t is measured as the funds from operations divided by total assets. Tobi n0s Qi ,t is total assets to 0,9 times the book value of assets plus 0,1 times the market

value of assets. Total debt divided by total assets are indicated by Debti ,t. Di vi d ent si ,t are

estimated as the amount of cash dividends payed to total assets. The last variable of the five-factor model is measured as the total of cash and short-term investments to total assets.

All variables are investigated on their statistical quality. First, outliers are detected in the data sample. Outliers can lead to the distortion of the variables and therefore of the whole regression. These are recovered for by winsorizing each individual variable at the correct percentiles. As a consequence, the estimators will be robust to outliers. Next, the variables

heterogeneity is detected, robust standard errors have to be used in the regression analyses. S, after testing for distribution and linearity, variables have been transformed when necessary. Fixed effects are used in the regressions to control for both time-invariant unobservable firm characteristics and quarterly time-varying unobservable factors.

3 D

ATA

D

ESCRIPTION

This chapter gives an overview of the data sample that is used in this research. First, the gathering of the data sample is discussed, where the data is obtained from different databases. Next, descriptive statistics are provided in order to give a clear overview of the sample that is used for the examination of the hypotheses.

3.1 D

G

In order to be able to estimate the variables used in the regression models, data has to be obtained from several databases. The Wharton Research Data Services (WRDS) consists of multiple databases and is the primary source for the data generation of this paper. Firm specific data on capital structure is obtained from the Compustat database, which is available through WRDS. Variables for the KZ-index are also obtained from Compustat. Compustat data for North America is provided on a quarterly basis. Specific REIT data can be found with the use of the Standard Industrial Classification (SIC) code 6798.

Total monthly return data on each individual REIT are gathered from Factset. Tickr symbols are then used to find the fitting REIT variables from other databases. The REIT portfolio returns are obtained through the Center for Research in Security Prices (CRSP) Ziman REIT Database, which is also provided through WRDS. The total return data gathered also controls for any dividend payouts to the shareholders. Through the Ziman REIT database it is also possible to gain insights in the real estate sector allocations among property types of each portfolio on a monthly basis.

The one month US treasury bond yield is gathered from the Federal Reserve. The FTSE/ NAREIT All Equity REIT Index is a public float adjusted, market capitalization-weighted index of U.S. Equity REITs. Included in the index are all tax-qualified REITs that have at least half of their total assets invested in real estate assets apart from mortgage investments secured by the underlying property (NAREIT). Consisting of 171 constituents, the FTSE NAREIT All Equity Index has a total net market capitalization value of 939,991 million (FTSE factsheet, 2018). Through the Kenneth French Database, SM Bt, H M Lt and MOMt monthly values are

obtained. The specific factors are available for the United States for the testable time frame. The factors are composed from the use of six value weighted stock portfolios from the NYSE, AMEX and the NASDAQ.

All collected archival data is then filtered. Mortgage and Hybrid REITs are removed from the databases, therefore leaving only the Equity REITs. All REITs with missing values are removed from the sample. As a balanced panel regression model is conducted, only REITs existing from January 2011 until December 2018, without data gabs, are used. All monthly data is converted into quarterly data. Data from the multiple databases are merged using STATA. The following multivariate regressions are conducted with quarterly variables, as firm specific data gathered form Compustat is only available with annual or quarterly reports.

3.2 D

S

The graph in Figure 3.1 displays the development of both REIT returns and idiosyncratic risk during the time span form 2011 until 2018. Idiosyncratic risk is based on the standard deviation of the error term from the Carhart four factor model. Returns are excess returns averaged for all REITs in the corresponding quarters.

Figure 3.1: Idiosyncratic risk and REIT excess return plotted with quarterly data.

During the first years risk and return perform within the spread of around -5% till 5%. After 2011, risk and return are shown to be pretty stable within this range over time. However, from 2017 the two lines start to move in opposite directions. Idiosyncratic risk makes a steep upward movement while return decreases strongly. This line provides interesting insights in the movements of both variables. Risk and return do not seem to have a constant opposite or

equal direction. In several periods risk increases seem to be aligned with an upward return rate. While in other periods the two variables move inversely. However, from 2017 a steep decline in REIT return rates can be detected, which is accompanied by a steep increase of the idiosyncratic risk. When determining the correlation coefficients, negative correlations appear in 2011, 2017 and 2018, which can be affirmed by the strong inverse movements in these years. Yearly correlations are shown in Table A1 in the Appendix.

Descriptive statistics of the variables are shown in Table 3.1. The mean, standard deviation and minimum and maximum observations are provided. The mean return is negative for the sample. This negative return is mostly due to a large decline in returns from 2017 till the end of the sample period, which is also visible in Figure 3.1. This strong decline has resulted in an overall negative return for the sample period. Comparing the REIT returns to the market excess returns shows that both have a comparable negative mean. However, the spread between minimum and maximum values of the excess REIT returns is much larger. As the market excess return is based on the historical returns of the FTSE/NAREIT All Equity index, a comparison between the return components could provide clarifying insights. An overview of the return developments is provided in Figure A1 in the Appendix. It is shown that the excess market and REIT return trends show very similar movements. Therefore, the performance of the investigated dataset is very close to that of the benchmark.

Table 3.1: Descriptive Statistics. Descriptive statistics are shown on the entire dataset from 2011-2018 using quarterly data Giving insights in the mean, standard deviation and minimum and maximum values of all variables used. Values should be read as percentages, where 0,01 should be interpreted as 1%.

Variables Mean Standard Deviation Minimum Maximum Excess REIT Return -0.0195 0.0607 -0.1584 0.1329 Excess Market Return -0.0173 0.0471 -0.1123 0.0510 Small minus Big -0.0007 0.0110 -0.0210 0.0193 High minus Low -0.0016 0.0109 -0.0257 0.0213

Momentum 0.0029 0.0136 -0.0262 0.0310

Idiosyncratic Risk 0.0483 0.0411 0.0003 0.1812

Leverage 0.3719 0.1232 0.1279 0.6754

Short Term Maturity 0.2210 0.1463 0.0000 0.5669 Kaplan Zingales Index 0.7716 0.7244 -1.4048 3.1254 Number of Observations = 3,040

Idiosyncratic risk shows a mean value of 4.83%. These risk values are comparable to the idiosyncratic risk components used in the data sample of Gerlach et al. (2018) for the post-crisis period. However, the maximum value is much larger, due to the steep increase in 2017. The mean of the leverage is 37.19% with a spread ranging from 12.03% to 67.54%, which is consistent with the statement that REITs take on a significant level of leverage (Barclay, Heitzman & Smith, 2013). Whereas industrial firms have an average leverage ratio of approximately 20%. This difference can be explained by equity REITs having a rather steady and certain income stream from real estate operations and the tax advantages accompanied with taking on debt. Comparable studies of Alcock and Steiner (2018) and Giacomini et al. (2015) show average leverage levels of around 44%-46%. However, their data samples are measured until 2013. The slightly lower leverage amount of this paper can be explained by Figure 3.2. Both market capitalization and leverage have increased within the tested timeframe. Market capitalization as represented as the number of shares outstanding times the share price, and is external financing as obtained through the issuance of shares. The leverage trend shows clear periodic behavior within time spans of approximately one year. This can be explained by the refinancing of leverage. Both external financing components show limited exponential growth. However, market capitalization has shown a greater increase relative to debt financing. Implying that less debt is attracted and more shares are issued, showing a small sign of delevering.

Short term maturity has a mean of 0.2210, which means only 22.10% of the leverage is due within three years. However, with a standard deviation of 0.1463, a great variety among the different REITs can be detected. Generally REITs tend to attract longer term debt compared to other sectors, as their asset maturity is high and serves as good collateral. According to Giabmona et al. (2008) the use of longer maturity leverage is corresponding to lower leverage levels. This can be confirmed by the discussed delivering and low concentration of short term debt. Accordingly, this results in lower refinancing risks for REITs (Alcock et al., 2014). The Kaplan Zingales index fluctuates between -1.4048 and 3.1254 which means that firms with the lowest or most negative index are the least financially constrained.

Descriptive statistics on the investigated real estate property sectors separately are dis-played in Table A2 in the Appendix. The 3,040 observations are fairly spread over the seven sectors. The mean returns of all sectors are rather equal, there is not one sector to be clearly pointed out as the winner or loser. Diversified REITs and Lodging/Resorts show to have the highest amount of leverage. The use of short term debt is overall pretty equally divided among the different sectors. A large difference among the sectors can be detected for the Kaplan Zingales index variable. Self-storage has a relatively low index of 29.76%, while retail is shown to have an index of 95.22% and therefore encounter less financial distress. The level of financial constraints is correlated to the amount of external debt financing. However, Giacomini et al. (2015) state that firms with identical debt levels may still have different distress ratios as short term maturity concentration could also influence the sensibility to these constraints.

4 E

MPIRICAL

R

ESULTS

The fourth chapter displays and discusses the results of this paper. First the double-sorted portfolio analysis is considered. Thereafter, the regression results are provided and discussed. Th results are then interpreted.

4.1 D

-

P

A

The analysis starts by investigating the relationship between idiosyncratic risk and returns. Three idiosyncratic risk sorted portfolios are formed for every quarter in the sample with their corresponding return rates. Panel A of Table 4.1 shows the alphas based on the relationship between excess return and idiosyncratic risk, the Carhart four-factor alpha is included as a robustness check, for the additional effect of size premium, value effect and momentum factors. Alphas are gathered as the constant from the regressions on a quarterly basis.

Increasing idiosyncratic risk levels are associated with growing returns, which is shown by the increasing alpha among the ascending idiosyncratic risk sorted portfolios from low to high. The last column presents the difference between the high and low idiosyncratic risk portfolios. A positive High-Low variable would indicate a positive relationship as the high risk sorted portfolios generate higher returns relative to the low sorted risk portfolios. Also when accounting for the Carhart four-factors, this positive development remains. This contemplates on a significant positive relationship between idiosyncratic risk and return In accordance with Sun and Yung (2009) this positive effect is affirmed, as they also find a positive relationship between idiosyncratic volatility and equity REIT returns. Gerlach et al. (2018) show a negative relationship before and during the financial crisis, but state that this relationship disappears in the post-crisis period up to 2012.

The results of the double-sorted portfolio analysis are shown in panel B, where the re-lationship between risk and return is examined while controlling for leverage. Likewise, idiosyncratic risk sorted portfolios are estimated, sorting the risk levels from low to high. Additionally, within each risk sorted portfolio, terciles are formed based on their levels of leverage. Increasing the levels of leverage in each portfolio, the significant positive relation-ship between risk and return remains. Moving down in panel B from low leverage portfolios to high sorted portfolios, positive alphas are shown which are all significant. The averaged portfolio alphas still tabulate growing returns within the upward movement of risk. Compa-rable results are founded when controlling for the additional Carhart factors, which shows

the results are robust. The effect of leverage will be further investigated in the next section.

Table 4.1: Double-sorted Portfolio Analysis. Alpha estimators are displayed. Panel A shows the alphas based on portfolio sorted idiosyncratic risk. Panel B shows double-sorted alphas for the relationship of excess return and idiosyncratic risk while controlling for leverage. Panel C shows double sorted-alphas while controlling for the concentration of short debt maturity. From the left to the right portfolios are sorted based on their idiosyncratic risk levels, across all panels. In panel B leverage portfolios are shown vertically. Averaged portfolio measures indicate the alpha of the corresponding sorted portfolios. Panel C shows the averaged alpha of the debt maturity sorted portfolio, which are constructed the same way as for Panel B. High-Low in the last column shows the vertical difference between the high and low risk sorted portfolios.

Idiosyncratic Risk Sorted Portfolios Low Medium High High-Low Panel A Excess Return 0.0027*** 0.0426*** 0.0691*** 0.0664*** (2.24) (21.70) (20.44) (12.55) Carhart Alpha 0.0052** 0.0387*** 0.0612* 0.0560** (1.99) (9.10) (1.81) (2.26) Panel B

Leverage Low Medium High High-Low Excess Return Low 0.0691 0.0560*** 0.1343*** 0.0650***

(1.57) (7.39) (9.14) (10.63) Medium 0.0029 -0.0590** 0.0862*** 0.0517*** (0.26) (-2.21) (3.45) (5.08) High 0.0121** 0.0776*** 0.0659*** 0.0352*** (2.05) (8.10) (2.88) (3.56) Averaged Portfolio 0.012*** 0.0690*** 0.1245** 0.0690*** (2.75) (7.50) (7.69) (8.78) Carhart Alpha Averaged Portfolio 0.012*** 0.0524*** 0.015* 0.0432***

(2.95) (6.30) (1.02) (6.06) Panel C

Debt Maturity Low Medium High High-Low Excess Return Averaged Portfolio -0.0066* 0.0151*** 0.0750*** 0.0645***

(-1.66) (2.97) (8.13) (20.45) Carhart Alpha Averaged Portfolio -0.0214*** 0.0063 0.0578*** 0.0581***

(-7.12) (1.49) (6.10) (13.33) t-values in parentheses.

Carhart et al. (2015) and Ang et al. (2006) conduct a similar double-sorted portfolio analysis. They investigate the effect of leverage on the risk-return relationship in three different periods around the financial crisis. In the pre-crisis and within-crisis periods they find a negative relation between idiosyncratic risk and return, but this negative relation does not seem to uphold in the period after the crisis. When controlling for leverage they only find a significant negative effect in the pre-crisis period, while for the other investigated periods there is no evidence of an existing relationship. In accordance with Carhart et al. (2015), Ang et al. (2006) also find no evidence suggesting leverage can be used to explain idiosyncratic volatility and its relation to returns.

An equal tercile investigation on the effect of short term maturity on the risk return rela-tionship is conducted. The levels of short term maturity are also sorted into portfolios from low, medium to high short term debt concentration. The averaged portfolio results of the excess return and Carhart alpha are shown in Panel C. The positive relation between risk and return remains visible and does not disappear with the addition of short term debt portfolios. Similar results are found by Pavlov et al. (2018), who do not find prove for the existence of a significant relation between returns and capital structure, including debt maturity. Further investigation on the impact of different firm characteristics on the relationship with fund returns is provided in the next section.

4.2 M

R

R

Regressions are conducted to investigate the effect of both idiosyncratic risk and capital structure on the excess returns of equity REITs. In order to decide on the statistically correct use of either a random or fixed effects panel regression, a Hausman test is performed. With a chi-squared value of 68.42% and a p-value<0.000, it can be concluded that fixed effects should be applied.

The Carhart four factors all impose a significant effect, where the excess market return captures most of the variability. Market excess returns provide a constantly positive effect on returns over all of the five regressions. The importance of this factor could be explained by the high positive correlation between the data sample and the market. As shown in Figure A1 in the Appendix the excess return trend lines of the REIT sample and the FTSE/NAREIT All Equity Index seem to coincide very well.

The small minus big coefficient as proxied by the return premium of small over large firms, are positively different from zero. This indicates that if a portfolio consists of more small-cap

firms, they would gain a significantly higher return in the long-run. The high minus low factor also imposes a positive effect on excess return. The high component is indicated by firms with high book to market values, whereas the low is proxied by firms with low book to market values, respectively these firms are categorized as either value or growth funds. This significantly positive HML coefficient implies that REITs with higher book to market values should earn higher returns and therefore growth REITs outperform value REITs. A negative momentum coefficient is found to be significant at the 1% level. The momentum factor reflects the trend of firms that show a positive performance in one period to maintain to provide positive results in the future, while firms that perform poorly in one period to also show negative results in the following period. A negative momentum effect means that fund who have performed as ’winners’, are perceived to experience negative results in the following period.

Table 4.2: Multivariate regression results

(1) (2) (3) (4) (5)

Variables Excess Return Excess Return Excess Return Excess Return Excess Return Excess Market Return 1.2122*** 1.2192*** 1.2104*** 1.2172*** 1.2144***

(81.67) (81.80) (79.11) (79.74) (78.19) Small minus Big 0.5286** 0.5135*** 0.5288*** 0.5136*** 0.5061***

(8.35) (8.09) (8.35) (8.10) (7.73) High minus Low 0.1724*** 0.1679*** 0.1713*** 0.1666*** 0.1798***

(3.66) (3.58) (3.67 (3.58) (4.03) Momentum -0.2349*** -0.2279*** -0.2361*** -0.2293*** -0.2263*** (-6.14) (-6.08) (-6.11) (-6.06) (-6.06) Idiosyncratic Risk 0.2144*** 0.21.74** 0.2140*** 0.2168*** 0.2175*** (10.97) (11.16) (11.01) (11.21) (11.23) Leverage -0.0276*** -0.0279*** -0.0314*** (-3.80) (-3.82) (-3.95) Short Term Maturity 0.0038 0.0043 0.0044 (1.06) (1.11) (1.13) Kaplan Zingales Index 0.0014

(1.54) Constant -0.0070*** 0.003 -0.0079*** 0.0022 0.0034*

(-9.84) (1.13) (-6.50) (0.82) (1.83) Observations 3,040 3,040 3,040 3,040 3,040 R-squared 0.8085 0.8098 0.8084 0.8097 0.8099 Fixed Effects YES YES YES YES YES Robust St. Errors YES YES YES YES YES

t-values in parentheses. *** p<0.01, ** p<0.05, * p<0.1

All systematic risk coefficients, except for the momentum factor, are shown to have a positive relation to excess returns. This is in contrast to Sun and Yung (2009), that conclude systematic risk factors impose a negative impact on returns.

Leverage also shows a negatively significant impact on the REIT excess returns. This is opposed to the hypothesized positive effect. Giacomini et al. (2017) also find a negative effect and state that highly levered REITs tend to underperform REITs with lower debt levels. A reason for the negative perception could be explained by investors concentrating more on requiring higher returns with increasing levels of idiosyncratic risk (Gerlach et al., 2018).

Idiosyncratic risk is significant at the 1%-level over all the five different regressions. This means that the incorporation of additional variables does not change the existing risk-return relationship. This implies that the relationship between return and idiosyncratic risk is not distorted by the capital structure of the firm and confirms the results found in the double-sorted portfolio analysis. In accordance with Sun and Yung (2009) controlling for selected firm characteristics that are generally connected to fund risk does not lead to changes. Chaudhry et al (2004) do find evidence for a significant effect of leverage on idiosyncratic REIT returns, but notes that the tendency of this association is sensitive to the specifications of the used model.

The model has high explanatory variables, which is indicated by the high R-squared values. Generally panel data regressions tend to provide high squared values. The overall R-squared is shown, however it is important to note that when comparing leverage and debt maturity, differences are detected. Regression 2 with the inclusion of the leverage coefficient seem to have higher within R-squared values, implying they explain the return differences in time. Higher between R-squared values are found for regression 3, containing the short term maturity coefficient, which means that this variable gives a better explanation for the differences between the REITs.

The inclusion of the Kaplan Zingales index in regression 5 shows an insignificant financial distress coefficient. Therefore, enduring financial distress would not have a significant impact on returns. Giacomini et al. (2015) use the Kaplan Zingales Index as a measure for financial distress in their empirical research, however there is also no evidence of a significant relationship with returns in their U.S. sample. It is also important to notice that the incorporation of the Kaplan Zingales index does not pose an effect on the relationships of capital structure and idiosyncratic risk with REIT returns as the coefficients of risk, leverage and short term maturity remain stable. Moreover, the negative effect of leverage on returns

cannot be explained by the KZ-index as stated by George and Hwang (2010).

The conducted regressions tabulated in Table 4.2 mainly considers the capital structure and financial structure of the REITs. The investment focus of REITs can determine the performance and should therefore also be taken into account (Cheng and Roulac, 2007). As REITs generally have an investment focus in a certain real estate property sector, the underlying assets may also pose an effect on the examined relationship. Because of the use of fixed effects regressions, the time invariant real estate sectors are omitted. Therefore, the fixed effect regressions are repeated on each individual sector. The results are displayed in Table 4.3.

Table 4.3: Multivariate regression results per sector

Variables Diversified Health Care Industrial / Lodging / Residential Retail Self-Storage Office Resorts

Excess Market Return 1.074*** 1.3082*** 1.2007*** 1.0640*** 1.1664*** 1.2301*** 1.1487*** (22.26) (28.47) (48.95) (23.02) (39.00) (49.38) (88.31) Small minus Big 0.4076*** -0.6662*** 0.4481*** 1.2110*** 0.1271 -0.3665*** 0.1025*

(2.53) (2.73) (5.18) (6.88) (0.86) (2.49) (1.69) High minus Low 0.4247*** -0.1946** 0.0251 0.7031*** -0.0949 0.1083* 0.2546**

(3.13) (-2.68) (0.36) (4.81) (-0.75) (1.65) (2.06) Momentum -0.3954*** -0.3597*** -0.1492*** -0.3844** -0.1570** -0.1106** 0.5367*** (-4.98) (-3.32) (-2.72) (-2.29) (-2.05) (-2.07) (5.97) Idiosyncratic Risk 0.0383 0.2925** 0.0973** 0.1675*** 0.1001*** 0.1350*** 0.1167*** (1.10) (2.48) (2.55) (3.33) (3.30) (3.44) (7.22) Leverage -0.0185 -0.0250** -0.0363*** -0.0173 0.0054 -0.0180 -0.0022 (-01.02) (-2.48) (-5.38) (-0.52) (0.55) (-0.74) (-0.59) Short Term Maturity -0.0002 0.0074 -0.0116** 0.0076 -0.0180 -0.0001 0.0174 (-0.02) (0.76) (-2.03) (1.08) (-1.22) (-0.03) (1.34) Kaplan Zingales Index 0.0108** 0.0035 0.0037*** 0.0011 0.0029 -0.0020 0.0042**

(2.43) (1.49) (2.81) (0.28) (1.58) (-1.49) (1.67) Constant -0.0050 -0.0033 0.0123*** -0.0036 -0.0029 0.0024 -0.0003**

(-1.11) (-1.25) (4.32) (-0.48) (-0.58) (1.41) (-2.38) Observations 288 352 768 488 320 736 128 R-squared 0.8470 0.8493 0.8624 0.7026 0.8242 0.8325 0.8426 Fixed Effects YES YES YES YES YES YES YES Robust St. Errors YES YES YES YES YES YES YES

t-values in parentheses. *** p<0.01, ** p<0.05, * p<0.1

In these regressions the market excess return continues to be a highly significant risk factor among the different real estate sectors, as well as the largest risk factor among the four Carhart factors. Idiosyncratic risk has a positively significant relation to excess returns. Notable is the very low and non-significant idiosyncratic risk factor endured by diversified portfolios. This confirms to the traditional portfolio theories discussed in section 2.2 that propose