The effect of financial leverage on REIT returns : pre- and post 2008 financial crisis

(1)

The effect of financial leverage on REIT returns: pre-and post

2008 financial crisis

MSC Business Economics: dual track Finance & Real Estate Finance Master Thesis

Bilars, R. (Rik) 6094368 July 2015

(2)

Statement of originality

This document is written by Student Rik Bilars who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is 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.

(3)

Abstract

This paper examines the relationship between the leverage ratio and returns of 52 US REITs over the period of 2004-2014. The research has special attention for the 2008 financial crisis and the addition of financial distress costs as an important factor influencing the described relationship. The REIT asset class has relatively little existing research in this field. Two models are estimated, one self-constructed model and a model that reproduces the results of a paper by Giacomini, Ling and Naranjo (2015). The results show only a weak relationship between leverage and returns over the whole time-period. The result for leverage during the financial crisis, however, is strongly significant and negative. This implies that firms with a high leverage ratio, are more vulnerable in economic downturns. The financial distress measures do not have a proven influence on the investigated relationship and are questioned to be fit for these kind of data.

(4)

Because the REIT market is relatively young, some areas have been under researched. While several studies have investigated the impact of fundamental and behavioral factors on listed REIT returns, not many studies have looked at the effect of financial leverage. The study of Glascock, Liu and So (2000), states that REITs have become more and more like regular stocks since the beginning of the nineteen-nineties. Due to economic growth, credit became more easily available. Combining that with the fact that leverage has been proven to be an important factor for regular stocks, it can be assumed that leverage also plays a

substantial role for REITs. Only some very recent studies, like Cheng and Roulac (2007) and Giacomini et al. (2015) have investigated this subject. Therefore, the goal of this thesis is to add to these researches and examine the relationship between financial leverage and REIT returns. This leads to the following main research question: To what extent does financial leverage influence REIT returns?

Before the financial crisis of 2008, there was a trend to increase leverage in the REITs sector. During and after the crisis however, REITs have deleveraged (Green Street Advisors, 2009). This implies that the deleveraging would help controlling and solve negative effects from the crisis. Therefore, this paper does not only concentrate on the effect of leverage in general, it also takes into account the crisis period of 2008, to reveal if there is a different influence of leverage in boom and bust situations. This is also important because the 2008 crisis, was largely caused by the housing market (mortgage market).

The research on this subject for the REIT market might be limited, research on leverage in the financial world is widely available. The first paper to propose a theory

(6)

6

concerning financial leverage and returns, comes from Modigliani and Miller (1958). Their paper and especially their second proposition (MM proposition II), serves as a starting point for this thesis. They developed a relatively simple, but still widely accepted, model for the effect of leverage on returns. Their theory predicts a positive relationship between the two variables. After MM proposition II, many studies have examined this relationship, with mixed results. Many recent papers have found a negative relationship between leverage and

returns, like George and Hwang (2010). They state that the financial distress costs a firm faces are an important factor influencing this relationship. According to the paper of Sun et al. (2013) financial distress also plays a role in the realization of REIT returns. Over the years, several measures of financial distress have been developed. Ohlson (1980) states that his O-score predicts the probability of default of a firm, based on its financial distress costs. Ohlson builds on the earlier developed Z-score (Altman, 1968). Slightly other measures have also been introduced, like the Kaplan and Zingales (1997) index for financial constraint. In this research the O-score and KZ-index are included.

This paper is conducted to add results to the list of the effect of financial leverage on stock returns in general. But, especially, to add new information to the relatively scarce literature on the effect of financial leverage on real estate returns. Next to that, the paper takes the crisis period into account and uses two different measures for financial distress. For this paper a sample is used of 52 US equity REITs. The REIT market in the United States is chosen, because it is the most mature market, with the best data availability, and therefore is the best comparable to regular stocks. The dataset is a combined set of quarterly return data and stock information from CRSP and quarterly accounting figures from the SNL database.

The approach to answering the research question consists of two separate parts. First, a model is constructed and estimated with returns as the dependent variable and leverage as the most important factor. This model is estimated with, besides leverage, a set of control variables which emerge from the existing literature on both regular stocks and REIT stocks. Also a crisis variable is included, to measure the possible difference between the pre- and post-crisis period. The second part of the methodology uses the approach of the paper of Giacomini et al. (2015). This paper is the first on the specific topic of REITs, to investigate the relationship between leverage and returns and also take a crisis component into account. Their paper conducts a cross-country analysis, this paper will focus on the US REITs only. This paper adds supportive or contradictive evidence to the paper of Giacomini and also uses a

(7)

7 slightly more recent time period.

The paper starts with a literature review, which is divided in a part about leverage research for regular stocks, a part about the financial distress component and a REIT specific part. At the end of the literature section, the hypotheses are derived. In section three, the dataset is introduced in more detail and the methodology is made more concrete, including the equations that are estimated in both parts of the methodology. Section four contains the descriptive statistics for the two approaches. Section five includes the results of the

estimated regressions and provides an analysis and interpretation of the results. The results for the second part of the methodology are compared to the paper of Giacomini et al. (2015) in this section. Finally, section six concludes the main findings of the paper and discusses possible limitations and opportunities for further research in this field.

2. Literature review

This section contains the related literature concerning the effect of financial leverage on returns. First the theory that serves as starting point of leverage theory is discussed. Second, some articles which confirm or contradict this article are reviewed. The third subsection introduces several measures of financial distress. After that the link to Real Estate Investment Trusts (REITs) is made and some relevant results from that sector are discussed. Finally the hypotheses that are tested in the models are derived in the last part of the literature section.

2.1 Modigliani and Miller (1958)

The theoretical base of this paper is the Modigliani and Miller paper (1958). This paper introduces the first theoretical framework for the effect of leverage on returns. The authors aim to provide a theory to explain the effect of financial structure on market valuation. The paper starts with the idea that a firm has two goals that will be reached by rational decision making. The goals are maximizing profits and maximizing market value. In the case of certain outcomes, these two are equivalents. However, in case of uncertainty, they are not, in this case multiple outcomes become possible and outcome dispersion arises. The authors state that in this, more realistic case, debt may increase the expected return at the cost of more dispersion. Proposition II is represented by the following formula:

(8)

8

This formula states that the expected yield of a share consists of a capitalized rate p plus a premium of the debt-to-equity ratio times the spread between p and r. Following this formula, a higher leverage ratio (Dj/Sj) will result in a higher expected return. Next to the

stated theory the paper also shows some empirical results. It is expected that the coefficient on the leverage ratio should be significantly positive. Data about oil companies and electric utilities is used, significant positive effects of 5.1% and 1.7% are found respectively. Meaning that a 1% higher leverage ratio leads to an increase in shareholder return of 5.1% in the oil companies and 1.7% for the electric companies. The authors claim that these results support their theoretical proposition. When the model is extended with the square terms of leverage, to account for possible curvature, the coefficients remain positive and significant. The square terms have a negative coefficient, indicating negative curvature, this result is opposite to the existing theories at the time. The authors state that more extensive testing is required and the model probably neglects other factors influencing returns. One factor mentioned is the dividend pay-out ratio.

Although MM proposition II is based on several simplifications, it is considered the first widely accepted theory concerning the influence of leverage on returns. A theoretical paper from Gomes and Schmid (2010) builds on the MM framework. Although this paper relaxes some of the assumptions, the idea of a positive relationship between leverage and returns remains. After the Modigliani and Miller paper in 1958, many empirical papers have been dedicated to this subject, the most interesting results will be listed in part 2.2.

2.2 Empirical research on leverage and returns

The empirical evidence on the effect of leverage on returns shows mixed results. The overall image is that earlier studies in the seventies, eighties and nineties find positive results, while more recent papers find a negative relationship. Combining the positive results with the high leverage ratios in the real estate sector twenty years ago, implies that debt was seen as positive. One of the indirect questions that this paper will address is, if that should still be the case. Considering the fact that REITs have been deleveraging, debt might be less desired at this moment. But before shifting to the case of REITs, papers that investigate the leverage effect on stocks or other financial products are discussed.

(9)

9

and returns is the study of Christie (1982), the results show a significantly positive

relationship between leverage and returns. Their dataset consists of 379 American firms in the period of 1962 until 1978. Later work of Bhandari (1998) supports the findings of Christie and finds a positive relationship as well. Their empirical analysis shows a positive relationship between stock returns and the debt-to-equity ratio. The analysis controls for Beta and firm size. Their measure for leverage is built by taking the book value of the total assets minus the book value of the total common equity and divide that over the market value of common equity. To test the model, the authors use a Fama-MacBeth model (Fama & MacBeth, 1973). The same authors, found a positive relationship themselves in their paper in 1992 (Fama&French, 1992).

Contradicting results with Modigliani and Miller have also been found in following papers. Two more recent articles contradict MM proposition II. George and Hwang (2010) find a significantly negative relationship between returns and leverage, especially for risk-adjusted returns. As a motivation for their research George and Hwang mention a “leverage puzzle”. This puzzle arises from the paper of Penman, Richardson and Tuna (2007). In that research the book-to-market ratio is divided into a separate asset and leverage part. This separation provides a positive relationship between the asset part and returns, but a negative relation for the leverage part. This relationship holds under controls for : size, estimated beta, return volatility, momentum and default risk. George and Hwang (2010) state that these results are abnormal and contradicting, hence, a puzzle. After investigating the puzzle, the authors support the findings of Penman et al. (2007), concerning the negative relationship between leverage and returns. But compared to Penman, they add a possible explanation for the negative relationship, which would not be expected according to the MM proposition. George and Hwang (2010) include a measure of financial distress. The idea behind this is that the capital structure decisions of firms partly depend on distress costs. When asset payoffs are low, the financial distress costs add to systematic risk. Firms with higher distress costs will take lower leverage, which results in a smaller chance of defaulting. But part of the costs of systematic risk remain, so firms with little leverage have a lower chance of distress and a bigger exposure to systematic risk than firms with higher leverage. This is the explanation for the negative relationship between leverage and returns. Like many other articles these papers look at leverage as a part of the price ratio or book-to-market ratio.

(10)

10

Comparing these articles shows conflicting results, this thesis adds new results to this field of study by using a different dataset and a more updated time period. An important feature of the George and Hwang (2010) paper, the financial distress component, is also used in this paper.

2.3 Financial distress

An important part of the contradicting results in the previous section have to do with

financial distress costs. George and Hwang (2010) claim that these costs lead to the different effect of leverage that arises from their empirical analysis. If their findings are correct, this issue cannot be ignored in this paper. Even more important is the fact that this paper will focus on the pre- and post-crisis period. In the case of an economic downturn, such as the 2008 financial crisis, differences arise between firms regarding the financial distress costs. This idea is supported by the view of Opler and Titman (1994), they find that highly leveraged firms suffer the most in economic downturns when distress costs are high. The more conservative firms, with lower leverage ratios, perform better in such periods. Complementing to this is the paper of Garlappi and Yan (2011), who state that financial distress costs have a negative effect on returns. Furthermore, due to the unique regulations concerning REITs, financial distress costs are probably the most important element

influencing the effect of leverage. The reason for this is that REITs, in contrast to other financial institutions, do not have to deal with corporate taxes. Modigliani and Miller (1958) already include these taxes as an important factor. The absence of corporate taxes for REITs will probably increase the influence of other factors, such as the financial distress costs. More about the special regulations concerning REITs can be found in section 2.4. Combining these facts, financial distress costs have to be part of the empirical model in this thesis.

In financial literature, several measures have been presented to account for financial distress costs. Altman (1968) created the Z-score, a measure which consists of multiple financial ratios. The combination of the individual ratios leads to a measure with greater statistical significance compared to all the separate ratios. This model was updated by the paper of Altman, Haldeman & Narayanan (1977), who introduced the “ZETA-analysis”. Also contributing to this subject is the paper of Ohlson (1980), it includes a variation of the Z-score, which is referred to as the O-score. The main attribution of this paper is the sample size of over 2000 firms in the empirical analysis, compared to the paper of Altman, which

(11)

11

only included 66 firms. The O-score is built from the predictions of the separate ratios, combining these enabled Ohlson to come up with a general equation that provides the O-score of a given firm on a given date. Some ratios that are included in these measures are: total liabilities divided by total assets, working capital divided by total assets, current liabilities divided by current assets, net income divided by total assets and some dummies which are one if a certain level is exceeded. An example of such a dummy is the “OENEG” in the O-score, which takes the value of one if the liabilities are larger than the assets. In later sections of this paper, the model and the chosen variables for the financial distress measure are explained.

A third option for measuring financial distress is given in the paper of Kaplan and Zingales (1997). They have created the KZ-index, which measures the extent to which firms rely on external financing. Firms with higher KZ scores, have a higher probability of

experiencing difficulties in economic downturns. In that sense, it can be seen as a surrogate for the other financial distress measures. The index is built as a five factor model, including cash flows, debt, dividends, book-value and cash.

Concluding this section, there are several options for measuring the probability of financial distress. Where the O-score is a more extended and reliable version of the Z-score and the KZ-index provides a slightly different approach as a third option.

2.4 REITs

The first three sections of the literature part have focused on financial stocks in general, but because the 2008-crisis was largely caused by the real estate sector, the following two sections concentrate on real estate. Within the real estate sector, one of the investment vehicles which are comparable to stocks, are REITs. The abbreviation stands for Real Estate Investment Trusts, which are companies that invest (almost) solely in real estate. These firms allow investors to invest in real estate as part of their portfolio, without the capital

restrictions and liquidity risks that are associated with directly investing into real estate. REITs are becoming more and more like regular financial stocks (Glascock et al., 2000). For that reason this investment vehicle is chosen to be able to compare the results of this paper with the previous studies.

US equity REITs, which are used in this paper, were introduced in 1960. As mentioned before, REITs are subject to a different tax regime and regulations. First of all REITs are

(12)

12

required to pay out at least 90% of their taxable income as dividends to their shareholders. These firms do not have to pay corporate taxes when they meet the 90% pay-out

requirement, but the profits of the shareholders are taxed. Another restriction is that a REIT has to have a minimum number of shareholders. This restriction determines that one single shareholder cannot control the company on his own. There are no restrictions concerning leverage for US REITs, however, market pressure might cause firms to keep their leverage ratio at a certain level (Green Street Advisors, 2009). After this short introduction section 2.5 discusses the findings about the effect of leverage in the REITs sector found in recent

literature.

2.5 Leverage research in the REIT sector

To shift the topic to REITs, the report from Green Street Advisors (2009), provides some insight in the relationship between REITs and leverage. The report describes the start of the de-leveraging process which began during the financial crisis. It also states that a large debt percentage was considered normal in the real estate sector twenty years ago. In general, REITs tend to have higher leverage ratios than other types of firms. A diagram in the report shows that in 2009 the median leverage ratio for REITs in the US was 54%, which is much higher than the median of the S&P500 companies, which is 21%. In the sample used in the paper of Feng, Ghosh and Sirmans (2007), the average debt ratio is even up to 65%. The high leverage ratios itself are quite surprising, because REITs do not profit from the tax benefits associated with debt. Feng et al. (2007) state that the high ratios, despite the lack of tax benefits, are due to adverse selection costs of raising equity. Following the Green Street Advisors report, the level of debt seems to have been decreasing over time and that process is enhanced by the crisis.

In most cases, in leverage research, stock returns are mentioned, but as REITs behave more or less like normal stocks, this is generalizable. The study of Glascock, Lu and So (2000) investigate the link between stock, bond and REIT returns. Their main finding is that REITs have become more like stocks over the years, especially since the 1990s. Implying that in current times, the differences are small enough to compare results for stocks to results for REITs.

While, there are not a lot of studies examining financial leverage and REITs, there have been papers investigating the capital structure choices of real estate companies in

(13)

13

general. Examples of these kinds of papers are: Howe & Shilling (1988), Maris & Elayan (1990), Capozza & Seguin (2000), Brown & Riddiough (2003), Giambona, Harding and Sirmans (2008), Boudry, Kallberga and Liu (2010) and Harrison, Panasian and Seller (2011). However, these studies do not address the relationship between financial leverage and REIT returns. Among the papers that do investigate this particular subject are Cheng and Roulac (2007), who find a weak negative relationship on this subject, while Pavlov, Steiner and Wachter (2013) do not find a significant relationship at all.

The most recent article on this subject, Giacomini et al. (2015), find strong significant evidence for a positive relationship between leverage and REIT returns. Also, this paper takes into account the crisis effect, by including a dummy for the years 2008-2009. This dummy is also highly significant, but as expected, with a negative sign. An interaction variable of the crisis dummy and the leverage variable, also gives a significantly negative result, indicating that higher leverage in a crisis period leads to stronger negative results. Like George and Hwang (2010), Giacomini et al. (2015) include a measure of financial distress. In their paper they use the KZ-index from Kaplan and Zingales (1997), as mentioned before in section 2.3. The KZ index used in the Giacomini paper consists of five factors: cash flow, Tobin’s Q, debt, dividends and cash. Important to note, is that the KZ index is not significant in their US sample, also it does very little to alter the results for the leverage variable. This is

contradicting to the paper of George and Hwang (2010), who claim that the financial distress measures has a significant impact on the effect of leverage on returns. The paper of

Giacomini conducts their analysis for 8 different countries with a relatively developed REIT market, this paper will focus only on the most developed market of all, the US REIT market. The results of this paper are compared to the results for the US in their paper. An important part of this thesis, will build on the methodology of Giacomini, to compare results.

As Feng et al. (2007) state, there is little research on how REITs choose their external financing. Their own study shows that REITs with high market-to-book ratios often have high leverage ratios for long periods of time. Chan, Hendershott and Sanders (1990) find that the effect of changes in inflation and changes in interest rates have a larger impact on highly leveraged REITs. This Is what would be expected, because leverage in general exposes a firm to more risk. That is also why this thesis will be interesting, in times of economic prosperity leverage is seen as a positive tool, while after a crisis it is considered as risky. By taking both periods, this research can focus on differences in the two types of economic periods.

(14)

14

Some additional research focuses on target leverage for REITs. This concerns the question of why REITs choose for more leverage. Ooi, Ong and Li (2010) find that market timing is more important than a previously set target leverage. This would explain the deleveraging of REITs since the 2008 crisis.

Concluding from this section, there is a lot of research on real estate firms which involves leverage, less studies concerning REITs specifically and almost no study so far has directly researched the effect of leverage on returns. The very recent paper of Giacomini et al. (2015) is the only paper that directly investigates the relationship between leverage and REIT returns and finds a significant result. Contrasting with other recent literature on the subject, they find a positive effect. Because of the similarities in research questions and methodology, the Giacomini paper is the most important paper for this thesis. Part of the methodology is the same, so results can be compared. This thesis adds a slightly more recent dataset, with focus on US REITs, instead of the cross-section of 8 countries. Also a separate set of regressions are run, with different independent variables and a different financial distress measure.

2.6 Hypotheses

From the studied literature, it seems that leverage, In theory, should have a positive effect on returns. However, several papers have found opposite results, showing a significantly

negative relationship between leverage and returns. An important factor would be the financial distress costs as mentioned by George and Hwang (2010). Recent literature in general finds negative results for the effect of leverage on returns, however, the research that is closest to this thesis, the paper of Giacomini, gives a significantly positive result for leverage. However, the Giacomini et al (2015) paper is one of the few giving a positive result for leverage in recent studies. Therefore, a negative effect for leverage is expected, following the recent papers. The second hypothesis concerns the crisis, it is expected that the crisis enhances the negative effect. For the first part of the methodology, this means that the crisis variable (stating the period from 2007 and later) is negative and the same is expected for leverage in the crisis period.

Considering the literature, the following hypotheses are tested in this paper, mostly based on the recent papers:

(15)

15

(2) The relationship between leverage and REIT returns is negative, and stronger during and after the 2008 financial crisis.

(3) Adding financial distress to the equation (measured by O-score/KZ-index) will alter the negative leverage effect substantially.

These hypotheses are tested in two parts. In the first part a series of regressions is run with independent variables that emerge from existing literature, to measure the effect of leverage on returns. In the second part, the methodology of Giacomini et al. (2015) is copied and the results are compared with their US REIT market results. The expectations for part two, are simply that the results of this paper will be similar to those of Giacomini’s regressions for the US REITs sample. The only difference for the hypotheses is that Giacomini finds a positive relationship between leverage and returns. Meaning that on the point of the leverage effect, a contradicting result is expected compared to Giacomini.

3. Data and Methodology

The dataset used for this research is a combined set between return and price data from the Ziman REIT database which is found in CRSP and financial data from the SNL database. The company SNL is a provider of financial data on different business sectors, including the REIT sector. The dataset consists of 52 US REITs in the time-period ranging from 2004 until 2014. The dataset consists of quarterly data, in the existing literature different frequencies are used, but as Giacomini et al. (2015) states, their results show no significant differences when the frequency is changed from monthly to quarterly data. REITs are seen as a relatively young asset class, which is not fully developed yet in many countries, the US REITs have existed since 1960. Due to their long history US REITs may be assumed to be the best fit to the idea of Glascock et. al (2000), which states that REITs behave largely like normal stocks. Taking this into account, US REITs are the best option regarding the comparison with other papers. The time-period is determined to capture the financial crisis of 2008 and include some pre- and post-crisis years into the dataset.

The methodology consists of two parts, the first part is an OLS regression including a leverage ratio and some control variables which emerge from existing literature. In the first part, O-score will be used as a measure of financial distress, just as in the paper of George

(16)

16

and Hwang (2010). The second part follows Giacomini et al (2015) paper which uses the Fama & French factors in their regression and the KZ index instead of O-score.

3.1 Part one: effect leverage on returns, including O-score

In the first part an OLS regression is conducted, with returns as the dependent variable. The other side of the equation consists of the leverage ratio, a dummy for the pre-and post-crisis period, the financial distress measure and some control variables. These regressions (eq. 2-6) provide insight in the relationship between leverage and returns and also take the difference between the pre-and post-crisis years into account, as one of the assumptions made in this paper is that the effect of leverage will differ between those periods. These five equations are estimated:

2. Ret = β0 + β1Rett-1 + β2Size + β3BookMarket + β4Momentum + β5Leverage + ε

3. Ret = β0 + β1Rett-1 + β2Size + β3BookMarket + β4Momentum + β5Leverage + β6Oscore + ε

4. Ret = β0 + β1Rett-1 + β2Size + β3BookMarket + β4Momentum + β5Leverage + β6Oscore +

β7CrisisDummy + ε

5. Ret = β0 + β1Rett-1 + β2Size + β3BookMarket + β4Momentum + β5Leverage + β6Oscore + β7Crisis

+ β8Crisis*Leverage + ε

6. Ret = β0 + β1Rett-1 + β2Size + β3BookMarket + β4Momentum + β5Leverage + β6Oscore + β7Crisis

+ β8Crisis*Leverage + β9Crisis*Oscore + ε

In these equations, variable Leverage is the leverage ratio measured as total debt divided by total assets, Size is the firm size measured by market capitalization, BookMarket is the book-to-market ratio measured by dividing the book value over the market value, Momentum is a variable created by Fama and French. Together with the lagged value of return, these factors are the basis of the regressions. All these variables are frequently used when investigating the relationship between leverage and returns. Size and book-to-market ratio are included, because these are commonly used in other papers like Bhandari (1998) and Penman et al. (2007). The momentum variable is used in several studies, like George and Hwang (2004). After the first basic regression, they are extended with the Crisis dummy, which takes the

(17)

17

value of 1 when the observation took place during or after the 2008 financial crisis and zero otherwise. The measure to include financial distress in the regression is the O-score

introduced by Ohlson (1980). The O-score is a measure that has been proven to have good predictive power for bankruptcy (Dichev, 1998). Next to the O-score and the before

mentioned Z-score (Altman, 1968) there have been several other measures like the option-theoretic measure from Merton (1974) and the hazard model by Campbell, Hilscher and Szilagyi (2007). However, Chava and Purnanandam (2010) conclude that all the different measures show a negative relation between distress and returns. Combining the fact that George and Hwang (2010) use the O-score and the fact that all measures have the same impact on results, the O-score is used to measure financial distress risk in this paper. Due to lacking data in both COMPUSTAT and SNL concerning some of the accounting variables to compute O-score, the final O-scores have been manually put into the dataset using numbers of Y-charts1, a website with data and analytics used for investment decisions. In the final two regressions, interaction terms between Crisis and Leverage, and Crisis and Oscore are added to the equation. These terms measure the effect of leverage and O-scores in when the observation is in the crisis period.

Modigliani and Miller (1958) mention the dividend pay-out ratio as a possible

influence on returns. However, due to the specific regulations for REITs, which require a pay-out ratio of at least 80%, this variable is ignored in this section of the methodology. The differences in dividend pay-out ratio will be very small, only ranging from 80 until a 100%.

3.2 Part two: Giacomini et al. (2015)

The second part of the methodology refers the methodology of Giacomini et al (2015). . On the left side of that equation are the monthly returns, which in this paper are quarterly REIT returns. Giacomini conducts regressions for eight different countries and one regression for the whole sample, in this paper only the US REITs are examined. The results are compared to the results for US REITs in the Giacomini paper. The regression is stated in equation 7:

7. Ret = β0 + β1MktRFt-1 + β2SMBt-1 + β3HMLt-1 + β4Momentumt-1 + β5FirmLiqt-1 + β6Inflationt-1 +

β7Crisis + β8Leveraget-1 + ε

(18)

18

The independent variables on the right side of the equation consist of MktRF, which is the quarterly excess return between market returns and the risk-free rate. Also included are the other Fama and French factors, SML, HML and Momentum. Furthermore, there is a measure for firm liquidity (FirmLiq), which is computed by dividing the amount of shares traded on the last day of the previous period over the total amount of outstanding shares. Variable Inflation is the US inflation rate. The two variables which are of interest in this research are the Crisis dummy, taking the value of 1 if the observation is in the year 2008 or 2009. The leverage ratio is the final variable, again measures as total debt over total assets. The only variable that is left out in this research is the LocalMRP, which measures the local excess return in the Giacomini-regressions. This variable is mostly used to compare different countries and is dropped for the regressions in this paper.

After estimating the first regression, the authors add an interaction term named Leverage*Crisis, to account for the effect of leverage during the crisis period. In their last equation a measure for financial distress is added. Unlike the O-score from George and Hwang (2010), they use the KZ-index of Kaplan and Zangali (1997). This index is computed as follows in equation 8:

8. KZ-index = -1.002*CashFlow + 0.283*Tobin’sQ + 3.319*Debt – 39.368*Dividends – 1.315*Cash

Finally, in each regression, six lagged return values are included, based on AIC-criterion. Furthermore, all regressions are conducted using firm and time fixed effects. This thesis will follow the methodology of Giacomini et al. (2015), to be able to compare the results in the best way possible.

4. Descriptive statistics

4.1 Statistics main variables: Leverage and Returns

First of all the two main variables for this research are discussed in this section, Leverage and Returns. The summary statistics for the Leverage variable are reported in Appendix A1. The numbers show a relatively steady development through the years in the sample. In the pre-crisis years, average leverage in the sample increases from around 49.6% in 2004 up to around 54.7% in 2007 and 2008. After the crisis, a decrease is visible, breaking the leverage

(19)

19

ratio down again to around 49.4% in 2014. The same image is visible for the average standard deviation and the spread between the minimum and maximum value over the years. Both the spread and the standard deviation increase towards the crisis years and decrease afterwards. While the differences may not be as strong as might be expected from the Green Street Report (2009), a clear pattern arises from the statistic. The average leverage ratio for the whole dataset is 51.63%, which as mentioned earlier, is relatively high compared to other types of companies. Compared to the sample used in Giacomini et al. (2015), the leverage ratio for this sample is higher as well, the Giacomini sample has an average of about 44.6%. This different could exist due to a slightly different dataset and a minor difference in computing leverage.

The table for the development of Returns (Appendix A2) through the years in the dataset shows positive average returns for seven years in the dataset, excluding the crisis years 2007 and 2008, and the years 2011 and 2013. The negative returns in the crisis years are as expected, considering the economic downturn. The other two negative years could be explained as a lagged result of the crisis. Also notable is the relatively large standard

deviation in 2009, indicating that the spread between returns also increased as a reaction to the crisis years.

4.2 Part one

The summary statistics for the control variables in part one of the methodology can be found in Appendix A3. The sample size for most of the variables is around 1500, with the exception of the O-score variable. Due to a lack of data this variable only has 667 observations, which should still be sufficient for a reliable analysis. As can be seen from the minimum and maximum values, the value of O-score has a very large spread. The scores range from -37.34 up to 35.32, averaging at 3.187. Leverage ranges from 6% up to 96%, with an average of 51.6%. The average book-to-market ratio is .495.

4.3 Part two: Giacomini

To compare the sample of this paper to the sample of Giacomini et al. (2015), the table of summary statistics on the regression variables is copied in this paper. The Leverage variable has already been discussed in section 4.1 and is higher (0.516 over 0.466) compared to the Giacomini sample. The other variables do not show substantial differences, only due to the difference in data frequency. The Fama and French variables, MktRf, SMB, HML and

(20)

20

Momentum have been converted from monthly to quarterly data, this explains some of the differences. Inflation is equal, which is not surprising, considering the largely similar time period. The measure of firm liquidity FirmLiq is lower than in the Giacomini sample, interesting to see is that the maximum score in their paper is substantially higher. The

KZIndex measure for financial distress also differes a lot between samples, averaging at 1.4 in this sample compared to 0.316 in the Giacomini paper. This difference can be explained by data selection. Giacomini used accounting figures from the DataStream database, this database is not available for this research. This could mean that slightly different accounting measures are used in the formula for the KZ-Index. It is also possible that the different average is caused by some outliers in the dataset. As can be seen, the spread between the minimum and maximum value in the dataset for this research is much wider. Finally, the number of observations lies around 1450 to 1500 for most variables, excluding the KZ-index, for which not all information was available for every observation. The Giacomini sample also had this restriction, although the difference was smaller. Nevertheless the number of

observations for KZ-index should be sufficient for the regression analysis.

5. Results and Analysis

This section of the paper includes the results and the analyses of the results from part one and two of the described methodology. Part one being the equations with control variables emerging from existing literature and part two of the methodology includes the regressions similar to the Giacomini et al. (2015) paper.

5.1 Results part one

The results of the first part of the regression, are displayed in table 1. The basic regression is in column 1 of the table, including the lagged value of returns, a size factor, a momentum factor and the variable of interest, leverage. In column 2 O-score is added to the equation and in column 3 the crisis dummy is added, which takes the value of 1 when the observations has taken place during or after the financial crisis. The last two columns report results with interaction dummies for leverage, o-score and crisis. As mentioned before, due to a lack of data on the financial distress variable, columns (2) until (5) rely on less observations

(21)

21

Table 1: Regressing US REIT Returns on Leverage, O-score and control variables from existing literature

(1) (2) (3) (4) (5) Rett-1 0.179*** 0.197*** 0.197*** 0.197*** 0.197*** (0.0417) (0.0505) (0.0506) (0.0506) (0.0502) Size 0.0022 0.0030 0.0032 0.0031 0.0029 (0.0014) (0.0026) (0.0027) (0.0027) (0.0027) Book-to-Market ratio -0.0321*** -0.0312** -0.0306** -0.0306** -0.0305** (0.0102) (0.0138) (0.0138) (0.0138) (0.0138) Momentum -0.977*** -1.369*** -1.380*** -1.382*** -1.384*** (0.0849) (0.178) (0.179) (0.179) (0.179) Leverage -0.0149 -0.0487** -0.0475** -0.0258 -0.0170 (0.0108) (0.0215) (0.0215) (0.0291) (0.0277) O-score 0.0006 0.0005 0.0005 -0.0011* (0.0008) (0.0009) (0.0009) (0.0006) Crisis dummy -0.0078 0.0053 0.00216 (0.0059) (0.0192) (0.0184) Crisis * Leverage -0.0253 -0.0388 (0.0366) (0.0357) Crisis * O-score 0.0021* (0.0012) Constant -0.0179 -0.0211 -0.0201 -0.0283 -0.0194 (0.0340) (0.0631) (0.0625) (0.0624) (0.0625) Observations 1429 652 652 652 652 Adjusted R2 0.247 0.310 0.310 0.309 0.311

Table 1 shows the regression outputs for the regressions measuring the effect of Leverage on US REIT returns. The basic equation estimated in column (1) is: Ret = β0 + β2Rett-1 + β3Size + β4Book-to-Market + β5Momentum +

β6Leverage + ε. The dependent variable Ret is the quarterly return for US REITs. The variable of interest is Leverage measured by the ratio of total debt divided over total assets. The control variables include a measure

for momentum the lagged variable of return and a book-to-market ratio, measures as the book value divided over the market value. In column 2 Ohlson’s O-score is added as measure of financial distress, column (3) adds a crisis dummy taking the value of 1 when the observation is during or after the 2007-2008 financial crisis. Interaction terms between crisis and leverage/O-score are added in columns (5) and (6). Standard errors are robust and displayed in parentheses below the models estimates. The stars stand for significance at the 10(*), 5(**) and 1(***) percent respectively.

compared to the complete sample. Including O-score causes a decrease in observations from 1429 to 652, this is still sufficient for a reliable analysis. The results show that the one

quarter lagged value of return is highly significant throughout all of the five estimated

regressions. The sign is positive, which is as expected. Other widely used variables in leverage research, the book-to-market ratio and the momentum variable are also highly significant in

(22)

22

all five equations. The size factor, measured by market capitalization, is not significant in any of the equations at the five percent level. This is remarkable, because it is common to include size in equations concerning the effect of leverage. Cheng and Roulac (2007) for example, find a significant relationship in two of the three time periods they include in their research. The variables of interest, leverage and O-score, show mixed results. Leverage is only

significant in column (2) and (3) of table 1. This

shows that adding O-score to the regression, causes the leverage variable to become

significant at the five percent level. The sign for leverage is negative in all five regressions. As expected in hypothesis one, based on recent studies on this subject, there is a negative relationship between leverage and REIT returns. Although this relationship is confirmed by the results, it is only significant in two of the five regressions. The financial distress measure of Ohlson (1980) is only significant at the ten percent level in one of the four equations in which it is included. The only significant outcome is in column (5) when also the interaction term between crisis and O-score is added to the equation. Adding this interaction term also causes the sign on O-score to switch from positive to negative. In line with George and Hwang (2010) a clear result for the effect of O-score is not found, only a weak negative relationship in one of the regressions in this paper. The crisis dummy does not show any significant result, meaning that the effect of being in the period after the crisis, does not differ significantly from zero. Also the interaction between crisis and leverage does not show any significant effects. These results do not confirm hypothesis number two. The estimate for the interaction term between crisis and O-score does show a positive significant sign in column (5). This result is remarkable, because it would be imply that firms with higher O-scores, meaning more financial distress, would have higher returns in a crisis period. The adjusted R-squared is relatively constant around .310 in all regressions. Adding O-score to the equation in column (2) causes an increase in adjusted R squared from .247 to .310, however this is probably due to the changing number of observations in the sample.

As a robustness check, the equations reported in Table 1, are re-estimated with a differently computed leverage variable. Instead of computing leverage as a ratio of total debt divided over total assets, leverage is computed following Modigliani and Miller (1958). They use the debt-to-equity ratio as a measure for leverage. Consequently, also the interaction term between crisis and leverage is changed to an interaction term with the debt-to-equity ratio as leverage measure. The results are reported in Appendix B1. The results are very

(23)

23

similar to the results in Table 1, however, the significance of the leverage variable in column (2) and (3) of Table 1 has disappeared when using the deb-to-equity ratio. The debt-to-equity ratio itself does not show any significance throughout all five regressions. The results for the variables that were significant in Table 1, O-score and Crisis*O-score in column (5), are not altered by the different leverage measure. Although the sign for the leverage remains negative, consequent with hypothesis 1, the significance of the results has changed. This would imply that it does matter for leverage research, which measure of leverage is used.

Concluding the analysis of part one, it can be stated that a weak negative relationship between leverage and quarterly REIT returns is found. This relationship is not persistent through all different equations. In fact, when including the (also insignificant) interaction term between crisis and leverage, the estimate of the leverage variable does not differ significantly from zero. When replacing the leverage ratio (total debt over total assets) by the debt-to-equity ratio (Modigliani and Miller, 1958), all results become insignificant. The difference between the pre- and post-crisis period, is not found in this sample. Hypothesis two, as stated in the literature section 2.6, is not confirmed by the results.

5.2 Results part two: Giacomini

The results for the second part of the methodology, following the approach of Giacomini et al. (2015) are shown in Table 2. Columns (2), (4) and (6) are directly comparable to the results of Giacomini, these columns are estimated including the year dummies. The other three columns are estimated without the year dummies, to get a broader image of the estimated effects. Columns (1) and (2) contain the basic regression, The interaction variable combining the crisis dummy and the leverage variable are added in column (4) until (6). In column (5) and (6) the KZ-index (Kaplan and Vingales, 1997) is included as a measure of financial

distress. The sample contains 1171 or 1170 observations in the first four columns, adding the KZ-Index causes a decrease to 581 observations, due to lack of available data.

The results in Table 2 show that the Fama and French control variables and the inflation measure are significant at the one percent level throughout all six specifications. The momentum variable (MOM) is only significant in two of the six regressions. The measure for firm liquidity shows significantly positive estimations in all equations. However, in the equations excluding the year dummies, the estimation of Firm Liquidity is significant at the one percent level, while in the other three columns with the year fixed effects it is only

(24)

24

Table 2: Giacomini regressions

(1) (2) (3) (4) (5) (6) (Mkt-Rf) t-1 0.870*** 0.661*** 0.878*** 0.668*** 0.935*** 0.780*** (0.0651) (0.0932) (0.0655) (0.0937) (0.0962) (0.129) SMB t-1 -0.975*** -0.769*** -0.982*** -0.774*** -1.112*** -0.999*** (0.153) (0.195) (0.149) (0.191) (0.204) (0.257) HML t-1 -1.104*** -0.949*** -1.085*** -0.933*** -1.093*** -1.023*** (0.155) (0.136) (0.151) (0.132) (0.183) (0.180) Momentum t-1 -0.272*** -0.0614 -0.273*** -0.0677 -0.183 -0.0792 (0.0794) (0.102) (0.0782) (0.0988) (0.114) (0.139) Firm Liquidity t-1 0.984*** 0.653* 0.963*** 0.634* 1.096*** 0.789* (0.347) (0.373) (0.342) (0.366) (0.423) (0.434) Inflation t-1 5.309*** 8.167*** 5.340*** 8.181*** 5.708*** 7.689*** (0.479) (0.748) (0.477) (0.729) (0.658) (0.887) Crisis -0.0177*** -0.0574*** 0.0339*** -0.0062 0.0265** 0.0031 (0.0051) (0.0098) (0.0128) (0.0147) (0.0134) (0.0193) Leverage t-1 -0.0077 -0.0073 0.0162 0.0162 0.009 0.0107 (0.01) (0.0096) (0.0114) (0.0109) (0.0166) (0.0168) Crisis * Leverage t-1 -0.0957*** -0.0945*** -0.0937*** -0.0934*** (0.0247) (0.0240) (0.0252) (0.0256) KZ Index t-1 -0.0001 -0.0001 (0.0002) (0.0002) Constant -0.0014 0.0190** -0.0133** 0.0070 -0.0068 0.0042 (0.0060) (0.0079) (0.0065) (0.0084) (0.0074) (0.0103) Observations 1171 1171 1170 1170 581 581 Adjusted R2 Year Dummies Return Lags 0.468 N Y 0.508 Y Y 0.475 N Y 0.515 Y Y 0.517 N Y 0.546 Y Y This table shows the results for the reproduction of the regressions used in Giacomini et al. (2015). The following basic regression is reported in column (1) and (2): Ret = β0+ β1Mkt-Rft-1 + β2SMBt-1 + β3HMLt-1 +

β4Momentumt-1 + β5FirmLiqt-1 + β6Inflationt-1 + β7Crisis + β8Leveraget-1 + ε. The dependent variable Ret stands for

the quarterly US REIT returns in the 2004-2014 sample. Three Fama and French factors are included (Mkt-Rf,

SMB, HML) as well as the momentum variable Momentum. FirmLiq is measured by traded shares on the last day

of the subsequent period, divided over the total shares outstanding. In every regression a crisis dummy is included, taking the value of 1 from januari 2007 until februari 2009. The variable of interest is Leverage, measured as total debt over total assets. All variables are lagged one period, except for the crisis dummy. Columns (3) until (6) expand the basic model with an interaction term between crisis and leverage and the Kaplan and Zingales (1997) index as a financial distress measure. Six return lags are included in all regressions, based on the AIC-criterion, and column (2), (4) and (6) include year dummies (excluding the crisis period and the year 2002). All standard errors are robust and shown in parentheses. The stars stand for significance at the 10(*), 5(**) and 1(***) percent level respectively.

significant at the ten percent level. The crisis dummy variable reports mixed results, highly significant in the equations without year fixed effects and only significant in column (2) with

(25)

25

these effects. This shows that including the interaction term, causes a major change in the significance of the crisis effect on its own. The leverage variable is not proven to differ

significantly from zero in all estimated equations. While the leverage variable on its own does not influence returns in this sample, the interaction variable shows highly significant results in all four equations. The sign is negative, indicating that having higher leverage during the crisis period has a very negative influence on returns. Adding the measure of financial distress in the last two columns, does not have any significant influence and does not alter the results for Leverage and the interaction term.

Comparing the results of the variables of interest to the outcomes of Giacomini et al. (2015) provides the following insights. Starting with the equation in column (2), differences with the outcomes of Giacomini are visible. Although the Crisis variables show the same sign in both papers, the effect in the sample of Giacomini is larger. The interpretation is the same, and significant at the same one percent level, but their estimate is .240 and the estimate in this paper is only .057. Concerning the leverage variable, results are contradicting, the sign in the Giacomini regression is positive and the estimate is significant at the one percent level. The sign and estimate in this paper are negative and not significant.

The second equation which can be compared, is column (4), which adds the

interaction term Crisis*Leverage. It is interesting to see, that adding the interaction dummy does not change the significance of Crisis and Leverage in the Giacomini paper, while in this paper, Crisis cannot be proven different from zero anymore. Again the positive significant relationship for leverage is not found in this paper. Both papers do show a one percent significant negative estimate for the interaction variable.

The last equation to be compared is reported in column (6), which matches the regression of Giacomini including the KZ-index. The similarity between the results of both papers in this equation is that adding the KZ-index does not alter the results reported for the regression without this financial distress measure. The variable for the KZ index itself is insignificant and cannot be proven to differ from zero.

Considering the models in general, the year dummies seem to have an important effect and are rightfully included into the regression. The models in Table 2 have also been estimated with firm fixed effects, like in the Giacomini paper. Results of those regressions are reported in Appendix B2. The results do not differ from the results in table 2, with the

(26)

26

the six specifications. This equation was estimated without the year dummies and is

therefore not considered as a contradicting result. A reason for choosing the output in Table 2 over the regressions with firm fixed effects is that they do not alter the results and do not increase the adjusted R-squared. The adjusted R-squared for the models in this paper is substantially larger than for the regressions of Giacomini. Values range from .468 to .546 in this paper compared to values .273 and .284 for Giacomini. This means that the model in this paper explains more of the volatility in the sample than the models of Giacomini.

Concluding this second part, the results show similarities as well as differences with the Giacomini et al. (2015) paper. One of the most important similarities is the significantly negative sign for the interaction term Crisis*Leverage. One of the main focus points of this paper concerns the effects of a leverage in the crisis period. The results support the findings of Giacomini. Most important difference is the lack of significance found for the leverage variable itself in this paper. Discrepancies in results can occur, due to different data

frequencies, although Giacomini states that they have run the same regressions for quarterly data and this did not alter the results. Furthermore, the slightly different dataset can be a cause, as well as minimal differences in computing the variables, or the databases which the data come from.

6. Conclusions

In this paper the relationship between financial leverage and REIT returns is examined. The research has special attention for the 2008 financial crisis and the addition of financial distress costs as an important factor influencing the described relationship. The motivation for this research lies in the fact that the REIT sector relatively unexplored is in some areas, one of those areas being financial leverage and returns. For regular stocks, there is a long history of leverage research starting with the Modigliani and Miller (1958) paper. The results are mixed, with some recent additions concerning REITs in particular. This paper adds to those recent results by estimating a self-constructed model and reproducing the approach of Giacomini et al. (2015).

The strongest result found in this paper concerns hypothesis two. Strong significant evidence is found for a negative relationship between leverage and returns during a financial

(27)

27

crisis. This result supports the finding of Giacomini et al. (2015) and Opler & Titman (1994). The regressions from the Giacomini paper, which are replicated in part two of the

methodology in this paper, show significantly negative estimates for all regressions in which the interaction between crisis and leverage was included. This shows that having a higher leverage ratio is more risky in times of economic downturn, firms with higher leverage get punished more in these periods.

In hypothesis one a negative result for the leverage component was expected in relation to returns, based on the most recent studies. In this paper, only weak evidence is found for such a negative relationship. In the first part of the methodology, leverage was significantly negative at the five percent level in two out of five estimated regressions. In all other equations, including the six following Giacomini, the effect of the leverage component could not be proven to be different from zero.

The third important finding concerns hypothesis three, the effect of the financial distress measures. Two different measures were used in this paper, Ohlson’s O-score (1980) and the KZ-Index (Kaplan and Vingales, 1997). The expectation was that these measures would have a substantial influence in altering the effect of leverage. In both parts of the research, this influence has not been found. O-score only proved to be only weakly significant in one of the five equations of part one. The KZ-index, which was included in the Giacomini-regressions, did not have any effect at all. The question which could be raised here , is if these measures are suited for measuring financial distress in the REIT sector.

The main conclusions of this paper are that, like earlier research, the outcomes are inconclusive about the effect of leverage. There is only weak evidence for a negative relationship between leverage and returns, and most of the time no evidence at all. In the crisis period of 2007-2008, however, having higher leverage has a significantly negative influence on REIT returns. This result is strong and shows that firms with high leverage are more sensitive to economic downturns. The financial distress measures did not proof to be of any influence on the examined relationship.

The limitations for this research mostly concern data issues. The data used in the Giacomini et al. (2015) paper came from DataStream. That database was not available for this research, therefore data was used from a combination of WRDS and SNL databases. This could possibly influence the results generated in part two of the methodology. Some of the differences between this paper and the Giacomini results could be caused by the data.

(28)

28

A possible opportunity for further research would be working with a bigger timespan, which includes more boom and bust situations. Or possibly including more crises caused by housing markets. Looking at the effect of leverage in boom and bust situation is limited if it only includes the 2008 financial crisis, like in this research. That is an opportunity to broaden and strengthen the results found in this paper. A second possible further research could focus on the financial distress measure. So far, all measures have been designed for regular stocks. It is questionable if, due to the insignificant results, the existing measures are fit for

measuring financial distress within REITs.

This paper does not provide a final answer to the question if leverage is a positive or negative factor for returns. It does contribute to the existing literature and to the REIT research in particular. Leverage will always be an issue and an interesting figure for investors, however, it remains dependent of the circumstances how it will influence returns.

(29)

29 References

Altman, E.I., 1968. Financial Ratios, Discriminant Analysis and the Prediction of Corporate Bankruptcy, The Journal of Finance 23(4), 589-609.

Altman, E.I., Haldeman, R.G. and P. Narayanan, 1977. ZETA analysis: A new model to identify bankruptcy risk of corporations, Journal of Banking & Finance 1(1), 29-54. Bhandari, L., 1988. Debt/Equity Ratio and Expected Common Stock Returns: Empirical

Evidence, The Journal of Finance 43(2), 507-528.

Boudry, W.I., J.G. Kallberg, and C.H. Liu, 2010. An Analysis of REIT Security Issuance Decisions, Real Estate Economics 38(1), 91-120.

Brown, D.T., and T.J. Riddiough, 2003. Financing Choice and Liability Structure of Real Estate Investment Trusts, Real Estate Economics 31(3), 313-346.

Campbell, J.Y., J. Hilscher and J. Szilagyi, 2007. In search of Distress Risk, The Journal of Finance 63(6), 2899-2939.

Capozza, D.R., and P.J. Seguin, 2000. Debt, Agency, and Management Contracts in REITs: The External Advisor Puzzle, The Journal of Real Estate Finance and Economics 20(2), 91-116.

Chan, K.C., Hendershott, P.H. and A.B. Sanders, 1990. Risk and Return on Real Estate: Evidence from Equity REITs, Real Estate Economics 18(4), 431-452.

Chava, S. and A. Purnanandam, 2010. Is Default Risk Negatively Related to Stock Returns? Review of Financial Studies 23(6), 2523-2559.

Cheng, P., and S.E. Roulac, 2007. REIT Characteristics and Predictability, International Real Estate Review 10(2), 23-41.

Christie, A.A, 1982. The stochastic behavior of common stock variances: Value, leverage and interest rate effects, Journal of Financial Economics 10(4), 407-432.

Dichev, I., 1998. Is the Risk of Bankruptcy a Systematic Risk?, The Journal of Finance 53(3), 1131-1147.

Fama, E.F. and J.D. MacBeth, 1973. Risk, Return and Equilibrium: Empirical Tests, Journal of Political Economy 81(3), 607-636.

Fama, E. F., 1976. Inflation Uncertainty and the Expected Return on Treasury Bills, Journal of Political Economy 84, 427-448.

Fama, E.F. and K.R. French, 1992. The cross-section of expected stock returns, The Journal of Finance 47(2), 427-465.

Feng, Z., C. Ghosh and C.F. Sirmans, 2007. On the Capital Structure of Real Estate Investment Trusts (REITs), The Journal of Real Estate Finance and Economics 34(1), 81-105. Garlappi, L. and H. Yan, 2011. Financial Distress and the Cross-section of Equity Returns.

The Journal of Finance 66(3), 789-822.

George, T.J., and C.Y. Hwang, 2004. The 52-week High and Momentum Investing, The Journal of Finance 59(5), 2145-2176.

George, T.J., and C.Y. Hwang, 2010. A resolution of the distress risk and leverage puzzles in the cross section of stock returns. Journal of Financial Economics 96(1), 56-79. Giacomini, E., D.C. Ling, and A. Naranjo, 2015. Leverage and Returns: A Cross-Country

Analysis of Public Real Estate Markets, Journal of Real Estate Finance and Economics 51(2), 125-159.

Giambona, E., J.P. Harding, and C.F. Sirmans, 2008. Explaining the Variation in REIT Capital Structure: The Role of Asset Liquidation Value, Real Estate Economics 36(1), 111-137.

(30)

30

and Stock Returns, Journal of Real Estate Finance and Economics 20(2).

Gomes, J.F., and L. Schmid, 2010. Levered Returns, The Journal of Finance 65(2), 467-494. Green Street Advisors, 2009, Capital Structure in the REIT sector.

Harrison, D.M., C.A. Panasian, and M.J. Seller, 2011. Further Evidence on the Capital Structure of REITs, Real Estate Economics 39(1), 133-166.

Howe, J.S., and J.D. Shilling, 1988. Capital Structure Theory and REIT Security Offerings, The Journal of Finance 43(4), 983-993.

Kaplan, N., and L. Zingales, 1997. Do Investment-Cash Flow Sensitivities Provide Useful Measures of Financing Constraints? The Quarterly Journal of Economics 112(1), 169-215.

Ling, D.C., and A. Naranjo, 2015. Returns and Information Transmission Dynamics in Public and Private Real Estate Markets, Real Estate Economics 43(1), 163-208.

Maris, B.A., and F.A. Elayan, 1990. Capital Structure and the Cost of Capital for Untaxed Firms: The Case of REITs, Real Estate Economics 18(1), 22-39.

Merton, R.C., 1974. On the pricing of corporate debt: The risk structure of interest rates, The Journal of Finance 29(2), 449-470.

Modigliani, F. and M.H. Miller, 1958. The Cost of Capital, Corporate Finance and the Theory of Investment. The American Economic Review 48(3), 261-297.

Ohlson, J.A., 1980. Financial Ratios and the Probabilistic Prediction of Bankruptcy. Journal of Accounting Research 18(1), 109-131.

Ooi, J.T.L., Ong, S.E. and L. Li, 2010. An Analysis of the Financing Decisions of REITs: The Role of Market Timing and Target Leverage. The Journal of Real Estate Finance and

Economics 40(2), 130-160.

Opler, T.C. and S. Titman, 1994. Financial Distress and Corporate Performance. The Journal of Finance 49(3), 1015-1040.

Pavlov, A.D., E. Steiner, and S.M. Wachter, 2015. Macroeconomic risk factors and the role of mispriced credit in the returns from international real estate securities, Real Estate Economics 43(1), 241-270.

Penman, S., S.A. Richardson and I. Tuna, 2007. The Book-to-Price Effect in Stock Returns: Accounting for Leverage. Journal of Accounting Research 45(2), 427-467.

Sun, L., S. Titman, and G. Twite, 2013. REIT and commercial real estate returns: A post-mortem of the financial crisis, Real Estate Economics 43(1), 8-36.

(31)

31 Appendix A: Descriptive Statistics

A1: Appendix A1 describes the development of the average leverage ratio in the sample by year. The leverage ratio is computed by taking total debt divided over total assets. Data are from the SNL database

A2:

Appendix A2 describes the development of the average return in the sample by year. The returns are US REIT returns collected from the CRSP database.

Leverage Year Ratio (average) Standard deviation Minimum value Maximum value 2004 0.496 0.159 0.007 0.791 2005 0.509 0.167 0.007 0.811 2006 0.523 0.169 0.018 0.831 2007 0.547 0.169 0.04 0.859 2008 0.547 0.171 0.04 0.941 2009 0.526 0.171 0.025 0.966 2010 0.510 0.157 0.034 0.86 2011 0.506 0.144 0.046 0.753 2012 0.516 0.139 0.053 0.903 2013 0.501 0.147 0.011 0.872 2014 0.494 0.14 0.048 0.728

Year REIT return (average) Standard Deviation Minimum value Maximum value 2004 0.027 0.032 -0.06 0.134 2005 0.011 0.037 -0.057 0.112 2006 0.023 0.037 -0.044 0.091 2007 -0.043 0.035 -0.113 0.026 2008 -0.023 0.078 -0.46 0.097 2009 0.036 0.128 -0.27 0.677 2010 0.025 0.035 -0.076 0.146 2011 -0.01 0.041 -0.201 0.061 2012 0.01 0.016 -0.016 0.06 2013 -0.003 0.021 -0.044 0.068 2014 0.03 0.022 -0.071 0.062

(32)

32 A3:

Appendix A3 provides summary statistics for the variables used in the first part of the methodology. The O-scores were collected from www.ycharts.com , a website for analyzing financial stock data. The momentum variable was collected from the Fama and French database in WRDS. The other variables were collected, or computed by, using data from the SNL database.

A4:

Appendix A4 provides summary statistics for the variables used in part two of the methodology. This table shows the same information as reported in Giacomini et al. (2015). Variables Mkt-Rf, SMB, HML and Momentum were obtained from the Fama and French factor database in WRDS. The other variables were collected, or computed by, using data from the SNL database.

Variable names N Mean Standard deviation Minimum value Maximum value O-score 667 3.187 5.502 -37.34 35.32 Leverage 1,481 0.516 0.161 0.006 0.966 Momentum 1,496 -0.000 0.034 -0.140 0.052

Size 1,496 4.012e+09 5.124e+09 9.730e+07 4.971e+10

Book-to-Market 1,491 0.495 0.406 -0.425 8.708 Variables N Mean Standard deviation Minimum value Maximum value Leverage 1,481 0.516 0.161 0.007 0.966 KZ-index 742 1.397 4.242 -86.49 66.43 Mkt-Rf 1,496 0.002 0.0301 -0.078 0.053 SMB 1,496 0.003 0.0123 -0.028 0.029 HML 1,496 0.003 0.0185 -0.047 0.046 Momentum 1,496 -0.000 0.0336 -0.140 0.052 Inflation 1,496 0.002 0.00399 -0.013 0.008 Firm Liquidity 1,444 0.011 0.00872 0.001 0.084

The effect of financial leverage on REIT returns : pre- and post 2008 financial crisis

The effect of financial leverage on REIT returns: pre-and post

2008 financial crisis

Table of Contents