The capital structure of US REITs before and after the 2007-2008 financial crisis.

(1)

The capital structure of US REITs before and after the 2007-2008

financial crisis

Name: Harmanpreet Singh Student number: 0515884 Date : 13 – 10 – 2013

Supervisor: Erasmo Giambona Second reader: Zacharias Sautner

Faculty: Economics and Business

Program: Master Science in Business Economics Specialization: Finance and Real Estate Finance

(2)

Dedication

For my late father, Darshan Singh, who always inspired and encouraged me to education and always believed in my potential. May his soul rest in peace.

For my mother, Amarjit Kaur, who kept her patience and believe in me and is proud of me no matter where I end up, as long as I do it with honesty and fulfillment.

(3)

Acknowledgement

First of all I would like to thank my family and friends, who are pushing me forward in life and through this research. I would not reach this far if they had not supported me. They also kept reminding me that education is just a source to reach higher goals and not a goal by itself. I think they make me a better person.

I would like to thank all the tutors , assistants, professors and students of the University of Amsterdam who provided me with an excellent education combined with humor, patience, discipline and love. I gained so much knowledge and skills , which is not just restricted to one field (profession) , but many other fields that are socially related.

I am thankful in particular to Herman ten Napel. He inspired me and supported me

throughout my whole period as a student of the University of Amsterdam. I hope one day I can excite students and approach other people with just the fraction of power by which he did.

I would like to thank Erik de Wit for triggering me to go deeper into this research. I would like to thank Hans van Ophem for guiding me and polishing my econometric knowledge. Every time I came by and visited him for some advice I left with hope and positive energy. He is an excellent tutor. I would like to thank Mohammad Arab as a study counselor he was always someone I could reach out to and mentally supported me. Rob van Hemert for upgrading mine thesis by providing it with liable and constructive feedback, and polishing my academic writing skills. Erasmo Giambona and Zacharias Sautner for grading me thesis. Last but not least, I would like to thank Floris van Dorp, with his help I was able to collect the necessary data and learned how to format large data files. Without the data I would not have been able to finish this thesis.

Thank you all.

(4)

1. Introduction

Under rationality long term assets should be financed with long term liabilities and short term assets with short term liabilities. Before the crisis there was opportunistic behavior in the financial asset market which encouraged most to finance long term assets with short term liabilities, also known as rollover debt, refinancing every time the short term period ends with new short term liabilities. Because of the sub-prime mortgage crisis money providers such as banks stopped lending and the financial system came to a hold and the problems for

refinancing became apparent. From this period on the need of stricter regulation worked their way through different institutions, corporations and companies.

REITs are companies that own income producing real estate, they invest mostly in direct real estate and these are long term investments thus these should be financed with a long term perspective. This empirical research will try to examine if there has been any change in the capital structure of REITs in the United States after 2007-2008. To look at the effect of the changes and because regulation changes need time to implement the subject years are 2009, 2010, 2011 and 2012. The research question of this thesis: Are there noticeable changes in the debt capital structure of REITs after the crisis in 2007-2008?

Previous studies that have been done mostly consist of one country with a small sample of REITs and a relatively small time frame. Erol & Tirtiroglu (2011) conducted a study on the capital structure of REITs in Turkey. Turkish REITs do not have to pay dividends and still enjoy corporate tax exemption. The paper shows significant effects of the 2001 financial meltdown whereas a shift takes place from the short-term to the long term debt market. The testing period is 1998 till 2007. However the REIT sample is considerably small with just 13 Turkish REITs. Mamoru (2009) investigates the Japanese REITs (J-REITs) between 2003 and 2008 and finds that J-REITs with high ratios of real estate investment assets in highly liquid regions, that is regions where the trade frequency per unit area is high, have high debt to equity ratios and debts of long-term maturity. J-REITs with high concentration ratios of small real estate assets, that are traded as residential properties, also have high debt-to-equity ratios and debts-to-long term maturity. Chikolwa (2011) investigates the determinants of capital structure for 34 Australian listed REITs (A-REITs) between 2003 and 2008 and finds that profitability, growth opportunity and operational risk are negatively related to leverage while size has a positive effect. However, even this study relatively uses a small time frame and a small sample of REITs. There are also studies with large samples, but the focus in these papers the focus lies on finding the proper capital structure theory that fits with

(6)

the REITs capital structure.1 This research is different in its structure by comparing two time periods for a relatively large sample of REITs. One period consists of a time frame before the crisis 1999-2006 and the other the time period after the crisis 2009-2012. The period of the financial crisis is considered to be 2007 and 2008 and is left out in the analysis, therefore a proper comparison can be made between the two time periods. By comparing two time periods it is possible to find variables changing from period to period, which indicates a change in the REIT capital structure.

Most of the universities have no access to Bloomberg database and therefore the use of the Bloomberg database in this research is unique to others.2 The relevant variables for

determining the leverage structure of US REITs are collected from Bloomberg database. The data is formatted into a balanced panel data. The regression for panel data consists of the fixed effects model and the random effects model. The appropriate model will be determined using a Hausman and Fixed Likelihood test. Furthermore, in this paper there will be three proxies to measure the leverage, respectively total debt to total assets, long term debt to total assets and short term debt to total assets. These models assist in answering the research question, but will also give insight in whether or not the shift was significant for the total, long term or short term debt.

This thesis is structured as followed. A brief description on the history and development of REITs will be presented which will indicate the characteristics and the growing importance of REITs. Following, a literature review on the capital structure theories with respect to REITs shall be presented. Thereafter the data and methodology will be discussed in section 4. In section 5 will contain the statistical and empirical analyses. In the last part there will be a discussion and conclusion to the research question.

1_{Dolde and Knopf (2010) have a sample of 243 REITs between 1994 and 2006 and test the agency}

problems related with insider ownership, risk and leverage in REITs. Because the trade-off and pecking order theory have contradictory predications with respect to insider ownership and risk they conduct an empirical analysis. The study finds no significant effects of risk, but do find empirical evidence that insider ownership is negatively related to leverage, thus in favor the trade of theory. Harrison et al. (2010) have a sample 473 REITs over a period of 1990 till 2008 and find that asset tangibility is positively related to leverage whereas profitability and market-to-book ratios are negatively related. Favoring the market timing and trade-off theory over the pecking order theory.

2

Previous studies mainly used SNL, Datastream or Compustat. I acquired this data sample through Global Property Research, Floris van Dorp.

(7)

2. History of REITs

Stocks, bonds, cash and real estate are rather well-known asset classes in general. The sub asset classes of these investments are less known. Within real estate a sub asset class is Real Estate Investment Trust (REIT). In this section the up come , development, rules and regulations regarding the REITs will be described, which will highlight the role of REITs in the financial world.

2.1 From the 60’s till the mid 80’s

REITs came about in 1960, when the U.S. Congress decided that smaller investors should also be able to invest in large-scale, income-producing real estate (Congressional Record, 2010). It determined that the best way to do this was to follow the model of investing in other industries and corporations by the purchase of equity. Therefore real estate investment vehicles were created. These brought more liquidity and transferability to an asset class which was until then considered to be only accessible for large investors.

President Dwight D. Eishenhower signed the legislation for the REIT Act law. This law allowed REITs to enjoy tax breaks which were similar to those of mutual funds

(Congressional Record, 2010). After the creation of the REIT Act the National Association of Real Estate Investment Funds (NAREIF) was incorporated which would later become known as the National Association of Real Estate Investment Trusts (NAREIT). During 1961 the first REITs were created and three of those REITs currently still exist, First Union Real Estate (now Winthrop Realty Trust), Pennsylvania REIT and Washington REIT, which indicates their long sustainability. Shortly afterwards, in 1965 Continental Mortgage Investors was the first public traded REIT that was listed on the NYSE (REIT, 2010).

In 1969 the European REIT legislation was passed in The Netherlands. This opened and marked the beginning of the global spread of the REIT structure(REIT, 2010). In 1970 Realty Trust Review was the first periodical journal specialized in public real estate securities. The latest news, research and feature reports on real estate securities were published in this journal. Until 1972, when the NAREIT REIT Index came, investors had no benchmark tools available for benchmarking price and total return performances in REITs. The rising

popularity of REITs is noticeable when in 1985 the first open-end mutual fund, which was specialized in REITs and other real estate securities came available on the market (REIT, 2010).

(8)

2.2 Late 80’s till beginning of the next century

The development of the real estate and REITs market was not just booming throughout history. In 1989 the real estate bubble busted. This bubble was caused by rising interest rates and over development in the commercial real estate sector. Many Savings and Loan (S&L) lent far more money than was competent and sound for real estate lending. The Tax Reform Act of 1986 helped deregulation of savings and loan lending which exacerbated the loan lending. It was the worst real estate downturn since the Great Depression of the 1930s (FDIC , 1997, p.141). In 1991 the market showed some signs of improvement and Kimco Realty Trust concluded the first successful equity REIT IPO in many years. In that same year for the first time ever a publicly traded REIT (New Plan REIT) achieved an equity market capitalization of $1 billion (REIT, 2010). In 1993 the President of the United States, Bill Clinton, signed the Omnibus Budget Reconciliation Act. This act made it easier for big

investors such as pension funds to invest in REITs (GPO, 1993). In the same year one of the largest REIT corporations of present date, Simon Property Group, made what was until then the biggest REIT IPO and raised $839.9 million. The IPO of Simon Property Group remained the largest IPO until 1997 when Boston Properties Inc. made an IPO and raising $902.8 million (REIT, 2010).

As part of the amendments to simplify REIT legislation, President Clinton signed the Simplification Act of 1997 (Govtrack, 1997). This Act allowed REITs to provide a small amount of non-customary services to its tenants which also meant REITs could have some additional income besides rents. In the same year the Treasury Department changed and updated some of its tax laws to ensure most foreign investors paid a 15 percent tax on dividends and provided them access into U.S. REITs.

2.3 1999 till present

January 1999 a small revolution found place as the NAREIT introduced a real-time index, the NAREIT Real-Time REIT Index. This allowed investors to trade with real time pricing, more transparency and efficiency of the market. As public traded real estate investment was attracting more and more attention from Europe, the European Public Real Estate Association (EPRA) was formed in 1999. Over the years NAREIT and EPRA have consistently maintained a close working relationship in promoting the public traded real estate and their member companies (REIT, 2010).

In 1999 another REIT Modernization Act was signed by President Clinton which would become effective from 2001 (Edwards, 1999). REITs would be allowed to form one or more taxable REIT subsidiaries (TRS) that can perform services to REIT tenants and others. REITs were until then restricted in their customary real estate services, for example providing

(9)

housekeeping services or cable television to residents in a residential REIT. Because of these constraints many REITs were prevented from competing with non-REIT real estate firms. REITs had to established third-party subsidiaries (TPS) to provide such services, which later are known as the taxable REIT subsidiaries (TRS), as these were not exempted from tax payments. These TRS are subject to the corporate tax, but not to the REIT

diversification tests (Matheson, 2001).

In 2000 the first REIT Exchange Traded Funds (ETFs) came on the market the iShares Dow Jones Real Estate Index. At present there are more than 20 REIT ETFs on the market. As of 2001 the Standard & Poor’s opens it indexes to REITs and NAREIT, EPRA and Euronext launch their EPRA/NAREIT Global Real Estate Index, a benchmark index for the global real estate sector. In 2003 the U.S. and U.K. Tax Treaty is ratified, which means that U.K. pension funds can invest in U.S. REITs for a reduced withholding rate of 15 percent and in certain cases without any taxes withheld on REIT dividends. Later the American Senate approved this ratification with other countries such as Japan, Mexico and Australia (NPB, 2003).3

In October 2004 President Bush signed another REIT Improvement Act (RIA). The impact was that it would eliminate a discriminatory barrier to foreign investors investing in U.S. REITs. Another impact was that RIA allowed REITs the opportunity to avoid REIT

disqualification for non-intentional REIT test violations. REITs got the opportunity to either fix a mistake or pay a monetary penalty of the violation of the REIT tax rule (Edwards &

Bernstein, 2005).

The Asia Pacific Real Estate Association (APREA) is founded in 2005 and is similar to the U.S. NAREIT and Europe EPRA establishments. APREA represents the Asia-Pacific public real estate sector on the global property investment market. The formation shows a spread and interest of real estate investment in all areas of the world. In the years following allot of academic and industry based research papers on global REIT and listed property investment highlight the importance of REITs as an investment tool.4

2.4 Current set of rules and definitions

Certain regulations and rules were set up to help qualify real estate companies as REITs, subsequently leading to tax breaks. Other rules and regulations were set up to help small, large, domestic or foreign investors, invest in REITs. The rules have been moderated and changed over the last fifty years. Some of the changes that are implemented are the Tax Reform Act of 1976 which allowed for REIT simplification amendments so REITs could be

3

More information about these cases can be found in NPB (2003).

(10)

established as corporations instead of business units. The “Five or Fewer” rule which made it easier for institutions and pension funds to invest in REITs. Finally, the REIT Modernization Act of 1999, which allows for taxable REIT subsidiaries and reduces the income distribution requirement from 95% to 90%. The latest qualitative rules as used by the U.S. Securities and Exchange Commission (SEC) can be found on the SEC website.5 The current SEC rules are the following:

1) At least 90% of the taxable income needs to be distributed to the shareholders as dividends.

2) A REIT must be managed by a board of directors or trustees. 3) Shares must be fully transferable.

4) At least 100 unique shareholders (after the first year).

5) 5 or fewer individuals may not hold more than 50% of the shares. 6) At least 75% of the total assets is invested in real estate.

7) At least 75% of the REIT income must be derived from real estate sources including rents, mortgages, and capital gains on real estate.

8) At least 95% of a REITs gross income must be derived from “passive’ financial investments, including rents, dividends, interest, and capital gains. This “passive”

strategy involves limited buying and selling, buying with the intention of long-term appreciation and limited maintenance.

9) Maximum of 25% of its assets may consist of non-qualifying securities or stock in taxable REIT subsidiaries.

The SEC classifies REITs in to three categories, Equity REITs, Mortgage REITs and Hybrid REITs. Most of the REITs are equity REITs. Equity REITs typically own and operate income-producing real estate. Mortgage REITs, on the other hand, provide money to real estate owners and operators either directly in the form of mortgages or other types of real estate loans, or indirectly through the acquisition of mortgage-backed securities (MBS). Hybrid REITs generally are companies that use the investment strategies of both equity REITs and mortgage REITs. At the beginning of the REIT development mortgage REITs dominated the REIT market, equity REITs were not popular. This was mainly because of the ownership and management of assets which had to be separated. After the Tax Reform Act of 1986 this changed, REITs were allowed to manage and own their properties as vertically integrated companies and become more popular among investors. As of 2012 (NAREIT), equity REITs

5_www.sec.gov/

(11)

formed a 91% of the publicly traded US REITs market. Mortgage and hybrid REITs were respectively , 7% and 1%.

REITs that are registered with the SEC and are traded on a stock exchange are known as publicly (listed) traded REITs. There are also REITs that are not publicly traded (non-listed) and known as traded REITs. Publicly listed REITs are more liquid compared to non-traded REITs as they are non-traded on the stock exchange. To buy or sell a non-non-traded REIT a broker that has been engaged to participate in the non-traded REIT can come through. Besides the listed and the non-listed REITs there is also a REIT mutual fund, which works in the same way a normal mutual fund would, but it mainly invests in REITs. Nowadays it is also possible to buy an REIT ETF (exchange-traded fund) or REIT Tracker that tries to follow the underlying asset as precise as possible. Where the REIT mutual fund will try to

outperform a certain benchmark, the REIT ETF or Tracker will try to follow the benchmark. To come back to the public REITs , anyone can buy the shares in a publicly traded REIT. The Shareholder has the benefits of real estate ownership without needing to manage or care for the real estate property. Other benefits include liquidity and diversity. Unlike real estate property the shares of a REIT can be quickly and easily sold on the stock market. Mostly the REIT does not possess a single real estate property, but portfolio of properties which in turn provides the investor diversification. A company that qualifies as a REIT is permitted to deduct dividends paid to its shareholders from its corporate taxable income. As a result, most REITs historically remit at least 100 percent of their taxable income to their shareholders and therefore owe no corporate tax. The taxes are paid by the shareholders on the dividend they receive. This structure prevents double taxation similar to a mutual fund. It is important to keep in mind that a REIT cannot pass any tax losses through to its investors (Griffith, 2011, p.317).

The U.S. REIT market is a mature market that has developed since the 1960s in to a highly liquid market for real estate that is accessible for investors worldwide. The NYSE capitalizes a REIT market of over $300 billion and trades $4 billion worth of REIT shares a day (NYSE magazine, 2013). Besides the listed REITs, the unlisted U.S. REITs manages assets of more than $70 billion. The U.S. and world REIT market nowadays provides investors with dividends, portfolio diversification, transparency, competitive performance and off course valuable liquidity. Due to the development and growth of the REIT industry REITs are considered being a mainstream investment.

(12)

3. The Capital Structure Theory

The following section describes the capital structure theories. These theories will be put in line with Real Estate and especially REITs considering whether these theories indicate an optimal structure for REITs. This overview will cover the Modigliani and Miller theorem (1958;1963) who state that an optimal capital structure is irrelevant. The trade-off theory and pecking order theory are follow up theories on the Modigliani Miller theorem, as both the theories introduce corporate taxes, cost of debt, information asymmetry and agency costs. The first three theories are rather popular and the two theories that follow, respectively, the market timing theory and the asset match theory are rather new. These theories shall be discussed as they may proof valuable for the near future in new upcoming researches and studies.

3.1 Modigliani Miller Theorem

One of the first financial theorems on the Capital Structure (CS) of a firm is that of Modigliani and Miller (MM). MM theorem simply states that the Market Value (MV) of a firm is not determined whether or not they use debt or equity, but on its earnings and the risk of the underlying assets. The three most common ways of financing are: Issuing shares (new equity), borrowing (debt) or using retained earnings/profits from previous years that were kept in the firm. MM (1958) propose that there is a CS irrelevance, but these propositions have some major assumptions that are not practical in reality. The basic theorem states that the market follows a random walk, basically assuming that the market can go up- or

downwards with the same probability. The theorem states further that there are no taxes. In reality, this is not the case as firms are confronted with corporate taxes as well as personal taxes on income and dividends. For now this assumption holds. There are no bankruptcy costs. Simply implying firms can shut down with no additional costs. The market economy is efficient as firms generating losses can simply stop their operations with no further

implications. There are no agency costs. The agent (management) handles in the best interest of the principle (shareholders), and does not do any actions for his personal benefits at the cost of the firm. There is total transparency in the market and the information in the market is equally spread among all the participants. Basically, this means that there is no asymmetric information and the market is efficient. When all of the assumptions mentioned above are met, the value of a firm is unaffected by its CS. It does not matter if the firm raised capital by issuing stock, selling debt or what the firm’s dividend policy is.

(13)

Value ( levered ) = Value ( unlevered ) = Present Value of Future Cash Flows

In general, any increase or decrease in the Value of the firm comes at the account of the firm’s stockholders (Brealey et al., 2006), but the Value of the firm is unaffected by its choice of CS. A simple example will show the MM neutrality in CS.

V = D + E E = V – D

In the case of the unlevered firm the value of D = 0. So that V = E. If the firm is fully levered E = 0. So the V = D. Important to note is that even though the structure of the capital has changed, the Value never changed. It is like a pie, where the pie remains equal, but how you cut the slice can differ, but has no impact on the size of the pie. An important assumption however is that the cost of capital of both the unlevered and levered firms are equal. See figure 1. There is no optimal CS for the firm that maximizes the value.

(14)

After this first proposition of the MM Theory other propositions introduced corporate taxes and interest tax deductibility in to the model, but no personal taxes, no financial distress and other costs of debt (Modigliani and Miller 1963 and Miller 1977). By ignoring the costs of debt, the model looks a linear function where the usage of debt is positive and the optimal CS policy is 100% debt as that maximizes the firm value. When taxes are taken into account the levered firm gets a tax shield that is summed up with the unlevered firm value. Firms can deduct interest with the tax rate times the value of debt (Tc*D). It is assumed that the debt is perpetual (Modigliani and Miller, 1963). This tax shield causes a positive linear function as the amount of debt increases. See the formula in figure 2. The value of the unlevered firm is denoted by Vu and the Value of the levered firm with tax shield is denoted with VL.

Figure 2 MM theory with taxshield

The MM theory looked from the present set of rules and regulations for a REIT shows that only one assumption holds ground and that is the assumption of no taxes. REITs are exempted from paying corporate taxes if 90% of the earnings are paid out as dividends. Al the other assumptions cannot be accounted for. With respect to bankruptcy costs a REIT

(15)

cannot easily leave or enter a market. As a REIT is a company that owns, and in most cases, operates income-producing real estate, shutting down the operations does mean additional costs. Immediate liquidation will expect to bring along costs and time pressure so that the Liquidation Value(LV) will be lower than the Market Value (MV).6 REITs own many types of commercial real estate, ranging from office and apartment buildings to warehouses,

hospitals, shopping centers, hotels and even timberlands. This commercial real estate is managed by several managers, from a property manager to an asset and portfolio manager. Whether these managers always act in the best interest of the principle and do not do any actions for personal benefit could be discussed. Graff and Webb (1997) show that the commercial real estate market is inefficient and the major cause of this inefficiency is the agency cost. The study done by Dolde and Knopf (2010) shows there is a significant relationship between the ownership structure of REITs and the risk structure of REITs. The REIT structure was made to provide investors in real estate diversification similar to that of mutual funds and stocks and because they are publicly traded at the stock exchanges they are liquid assets as compared to direct real estate. These previous mentioned papers show that the REITs market is not transparent and there is asymmetric information. Considering the fact only the tax exemption holds ground from the MM theory, it can be concluded that it does matter for REITs if they are unlevered or levered, and that there is a optimal capital structure out there.

3.2 Trade-off Theory

In reality it would make no sense to fully leverage a firm with debt. Too much leverage could put stringent financial and control restrictions on a company that in case it could not meet with it is financial obligations it would slide of into financial distress. This financial distress causes poor performs and results in liquidation of the firm’s assets. The costs associated with financial distress are the direct costs of bankruptcy and indirect costs associated with the circumstances of where the firm is momentarily standing. Those indirect costs can be debt overhang costs, which causes the inability to finance any more positive NPV projects which under better circumstances would have been executed. Excessive risk taking, or gamble behavior, win all lose all mentality, even to the point that the related firms, such as the suppliers and customers, can be scared off (Opler and Titman, 1994). Figure 3 shows how the value of the firm would look like when financial distress is introduced and that there is a optimal capital structure available.

6_{Shilling, Benjamin and Sirmans(1990) have shown this in their paper that Liquidation Value in most}

(16)

One of the weak points in the classical tax-deductibility benefit of leverage is that it gives a large incentive to choose debt over equity, and if the debt chosen has a high interest, the tax shield will be larger. This incentive can be misleading as choosing a high debt interest rate, also suggest a higher probability of financial distress. The research paper of Cheng and Tzang (1988) shows that REITs are sensitive to changes in short and long term interest rates. In a stable economic period between 1973-1979 REITs showed sensitivity towards long term interest rates. During the period of 1980-1985 where there was economic growth and higher inflation REITs were also sensitive to short term interest rates besides the long term interest rate. This could be explained by the fact that managers become more

opportunistic during economic prosperous periods and tend to finance long term assets with short term debt.

REITs do not pay corporate income tax, because of their unique regulatory environment. The tax-shield incentive should be proportional, so that REITs should use debt minimal according to the trade-off theory (Howe and Shilling, 1988). The trade off theory suggests that there is tradeoff between financing with debt or equity. When the benefits of financing with debt are higher than the cost of debt the value of the firm is increased (Kraus and Litzenberger, 1973). As the proportion of debt increases the marginal benefits decrease at a point from where on there is more marginal cost of debt then benefit. The benefits of debt are the tax benefits, and the costs of debt are the bankruptcy and indirect costs. Basically, the level of debt level is chosen to balance interest tax shields against the costs of financial distress. In figure 3 this pointed out as the optimal CS. The followers of the trade-off theory believe the firm sets a

(17)

target debt-to-value ratio and moves towards this target (Myers, 1984), this target is considered to be the optimal CS.

3.3 Pecking Order Theory

The pecking order theory is introduced and promoted by Myers and Majluf (1984). The pecking order theory basically states that there is order of financing. The order is based upon costs as asymmetric information, agency and financial distress. The financing order first prefers internal finance. Internal finance can be done with internal funds that are available for the firm. These are mainly cash, marketable securities, retained earnings and reserves. When there are no sufficient internal funds available to finance the firm steps over to debt financing. Debt financing proceeds/continues until it is not sensible any more to finance further with debt. This could occur when the debt service coverage or marginal costs of debt increase the benefits of marginal debt. Basically assuming the costs of financial distress discourages anymore debt. The last resort for financing is the equity financing. Equity financing in most cases means issuing new shares. Equity financing is seen as an external financing source as it expands the company’s current ‘owners’ with the new shareholders. It is assumed managers have an information advantage over investors and thus the source they use to finance actually signals the market about under and over valuation of the stock of the firm. If the stock is undervalued, it means the market is pricing the stock lower than what the manager considers being a fair price and will not try to finance with new equity. The issuance of Initial Public Offering (IPO) or Secondary Public Offering (SPO) is usual already done with a discount, and giving a discount on a price what is already considered to be at discount would mean the market would lower the price. So the manager prefers debt financing. In case the stock is overvalued the manager prefers to finance with equity as the manager does not mind going for an IPO or SPO with a discount as he knows the market is overvaluing the firm stock price. However this behavior is known to the market and its investors and they anticipate on this behavior. The financing signals the market and the market reacts. So when the board of a firm decides to issue debt for a project it signals the board’s confidence that a project or investment is profitable and the current market

circumstances are not in favorable to finance with equity. Vice versa when the board of a firm decides to issue equity this would lead to a drop in share price.

There are some industries where there are allot of intangible assets. Pharmaceutical, high-tech industries for example. It is easy to recognize these companies as they have high degree of expenditure on Research and Development, and there main and future business is highly dependable on this research and development. These industries show different

(18)

reactions to capital raisings in the market.

Frank and Goyal(2003) show that the pecking order theory fails for small firms where it should hold. Small firms are more information asymmetric compared to larger firms. So the signaling of financing from the firms’ board should lead to more stronger and noticeable reactions from the market. On the other hand Fama and French (2002) show that the pecking order theory can explain more features of the data then the trade of theory and that the use of the pecking order theory in practice should be favored over the two theories. Myers and Shyam-Sunders (1999) also had the same conclusion, and found the pecking order theory been preferred over the trade of theory.

The pecking order theory states that high profitability leads to lower debt leverage, because the firm has more internal sources available. So with respect to REITs, profitable REITs should have lower debt levels relatively to the less profitable REITs. Dividend policy is adaptable to investment opportunities. Sticky dividend policy with a high deviation in profits shows the need of financing from internal and external sources period to period. In case of external financing the firm prefers debt, then convertible bonds and equity as last resort.

3.4 Market Timing Hypothesis

Market timing is a theory based on the fact the market can be timed. This simply means that there are several strategies that allow you to profit in a bullish market as well as in a bearish market and firm’s decision making in financing with debt or equity are based upon these market circumstances (Baker and Wurgler, 2002). So basically the market is determinant for the use of debt or equity at certain points in time. If the market favors debt, the firm uses debt financing. If the market favors equity financing, the firm uses equity. There for the capital structure of a firm represents the cumulative decisions made in the past till present. Recent number of studies supporting this theory are done by Boudry et al. (2010), Li et al. (2010) an Huang and Ritter(2004). Supports of this theory in the real estate are done by Li, Ong and(2010 ; 2007), which suggest that target leverage behavior plays a role to market timing behavior in the financing decisions of REITs.

However the empirical evidence for this theory is still not sufficient and is not able to explain many occurrences in the market. In the market at same moment in time some firms issue debt while other firms issue equity. There are no studies found that explain this simple situation within the market timing model. Unlike the trade of theory and the pecking order theory, the market timing hypothesis still leaves allot of room for completion with additional research and literature. The market timing hypothesis is sometimes considered to be a part

(19)

of behavioral finance, which in itself is considered relatively new field in finance. The market timing theory shall not be further discussed as if falls outside the scope of this thesis.

3.5 The Asset Match Theory

According to Frank and Goyal (2007) there is no capital structure theory model available that is capable of explaining the wide range of reasons for the use of corporate debt financing. In the next section a new theory will be introduced which is not yet discussed or written about. This theory is named the Asset Match Theory and will provide a logical and rational

framework for corporate financing and capital structuring.

The Asset Match Theory is a new theory based on the capital structure of firms and how it preferably should be structured. Basically, the optimal capital structure of the firm should reflect the Asset configuration (activa) matched by its investment possibilities on the pasiva side of the balance sheet (equity and liabilities). This theory can be used a reference framework by which a judgement about the quality of the governance and opportunistic behavior on firm level can be made. With this theory you can either tell if the management or board of directors is handling in the best interest of the firm or is trying to implement

opportunistic behavior. The agents leading the firm should act in the best interest of the firm. The main importance of a firm is continuity, which is often taken for granted. The firm is often seen as an organic process taking place over time. The asset match theory (AMT) states that firms should match the assets according to the exposure they have with those assets. For example a long term asset should be financed with a long term investment proposition. This could either be a long term debt or equity. Similarly a short term asset should not be financed with a long term debt or equity. In either case if the firm does not match the assets correctly there is an asset mismatch.

In this thesis we will try to show that REITs currently are better matching their assets then before the crisis according to the AMT. Up to the financial crisis there was a lot of

opportunistic behavior noticeable in the market. Refinancing was a normal term in the financial world, which led to this behavior. It actually led people and firms to believe they could finance with a mismatch and make a profit (opportunistic behavior). The downside was never considered. If there was an appropriate match between assets in the first place there would maybe be a downfall in the market economy, mainly because of macro- economic factors, but not due to a “bubble”, which also busted during the previous crisis.

(20)

If firms enhance a good investment strategy, the effects of the AMT should be noticeable. So that there is a better match between the assets they have on their balanced sheets and the way they are capitalized. In this thesis we will try to show that, with the level of leverage that should have become less relatively to the equity financing. REITs are perfect for testing this new theory, because as they own real estate properties and a large part of their balance sheet consists of them, REITs should have become better matching firms to test the AMT.

I will analyse the leverage composition according to the previous described capital structure theories and try to get an insight of the REITs capital structure before and after the financial crisis. In the following section the methodology and data will be described.

(21)

4. Data and Methodology

In the following section the data, variables, model and descriptive statistics will be described in detail with respect to its source, features, format and limitations. First, the data

and thereafter the variables of the model. Third, the model and the expected hypothesis will be presented. The methodology which describes the panel data estimation techniques used to compute the estimates of the parameters of the model. The leverage models for total, long and short term debt with respect to total assets will be specified. Finally, the statistic analysis of the variables with respect to their mean, median, standard deviation, minimum and

maximum will be discussed.

4.1 Data

The data set runs from 31 December 1998 (1998Q4) till 30 September 2012 (2012Q3). Within this period the years 2007 and 2008 are removed as it is assumed that this period marks the financial crisis. As such the data springs from 2006Q4 to 2009Q1. The data is divided in two time periods, Period 1 and Period 2. Period 1 reflects the period before the crisis from 1998Q4 to 2006Q4 with twenty-three quarters, and Period 2 reflects the period after the crisis from 2009Q1 to 2012Q3 with fifteen quarters. Where previous studies were constrained by the use of annual data, this studies extracts quarterly frequency data including 48 quarterly intervals. The usage of quarterly data allows to capture the capital structuring activities more dynamically as financing occurs often throughout a year. As seen from the financial crisis a fall of prices in the capital markets can take place within a short time frame and increase the need of financing.

The analysis is restricted to the United States (US) and equity REITs, as the prime interest lies in REITs which own and hold investment properties. The identifications used for the US equity REITs are their Bloomberg ticker codes, because all the REIT specific data is

extracted from the Bloomberg database. A total of 292 US equity REITs are included in the dataset. Some of the REITs did not exist the total time frame from 1998Q4 till 2012Q3, but are still included in the data, which eliminates the survivor bias.7 Missing values have been omitted. This also means the initial dataset consists of an unbalanced panel.

7

Survivorship bias can lead to overly optimistic beliefs because failures are ignored, such as when companies that no longer exist are excluded from analyses of financial performance. It can also lead to the false belief that the successes in a group have some special property, rather than just

(22)

4.2 Independent and dependent variables

In this subsection the variables will be discussed and whether or not they are included in the analysis. First of all the dependent variable: Leverage. This thesis uses three common leverage ratios: total debt to total assets, long term debt to total assets and short term debt to total assets. Similar leverage ratios can be found in Feng, Ghosh and Sirmans (2007),

Elayan and Marts (1990), and Rajan and Zingales (1995). In the following section the

independent variables will introduced. These are size, dividend pay-out ratio, profitability ratio and the volatility of the earnings defined as operational risk.

4.2.1 Size

There are several interpretations for size. Dolde and Knopf (2009) use the log of total assets as a measure of size for REITs and find a positive relation between size and leverage, whereas Ertugrul and Giambona (2010) define size as the book value of the firms real estate assets. However their findings show that REITs with a market share in the upper segment market share quartile use less leverage, thus showing a negative relation between size and leverage. Here I will use the definition of Dolde and Knopf (2010) as the measure for size, that is the log of total assets. Size is relevant because it carries some important information. The larger the firm is the greater the ability the firm has to finance on the market (Fama and French, 1995). This suggests a positive relation between size and leverage and fits the trade-off theory. Myers and Majluf (1984) state that larger firms are more transparent and therefor need less monitoring, this argument suggests a negative relation between size and leverage and fits the pecking-order theory. According to the AMT size is irrelevant, the only importance is how the size is configured. If the firms total asset are largely fixed and long term assets the CS should be skewed towards long term financing, that is long term debt and equity. Because the sample is build out of equity REITs the effect of size will most likely be positive. This is similar to the findings of Feng, Ghosh, Sirmans (2007), Feng (2007) and Bond (2006).

4.2.2 Profitability ratio

This thesis incorporates the measure of firm profitability. Brealey et al. (2008) use the return on assets as a measure, which is simply the profit of the firm divided by its total assets. Because there are many accounting ways to lower profitability in the book such as high depreciation and amortization subtractions. The EBITDA is taken instead of the profitability, and the EBITDA is divided by the total assets to indicate a profitability ratio. The trade-off theory relates high profitability with higher incentive to increase leverage and thus the tax shields. However the pecking order theory suggest that firms with high profits have less appetite to finance with leverage and equity. The main source for finance should come from

(23)

internal funds (Myers and Majluf, 1984). REITs however are interesting to look at as

according to the regulations they are submitted to >90% pay-out policy. This leaves les ability to finance with internal funds if opportunities arise. With respect to the AMT the internal or external funds is relevant as long as the investment opportunity is financed with the same perspective. For REITs the investment opportunities are new real estate assets to acquire or build, so the financing should be with long term financing as real estate is a long term asset. This suggest a negative relationship between profitability and short term debt, a positive relationship with long term debt and equity.

4.2.3 Earnings risk

The earnings risk reflects the variation in earnings period by period. If this variation is high leverage will be low and vice versa. A stable earnings pattern indicates a stable income and the probability of defaulting on the debt is less likely to happen, but if the earnings are very volatile the likelihood of defaulting in one period is greater. Basically a higher variation in earnings means less debt, and lower variation in earnings means less debt. Both the pecking order theory and trade-off theory predict this relationship. In this paper the earnings risk is formulated by taking the percentage change period by period.

4.2.4 Equity change

Following Baker and Wurgler (2002), the change in leverage will be decomposed into a change in equity, which is either by equity issuance, retained earnings or both. A positive change in equity will have a negative effect on leverage. As more equity means less leverage in the total assets. In this thesis the change in equity will be considered a dummy variable. The measure of the change in equity will simply be calculated as the percentage change. If the change is less than 1% the value will be 0. This is done because the change may well be coming from bonus giving to the management in shares. If the change is 1% or larger then it is assumed that the change came from either equity issuance, retained earnings or both.

4.2.5 Dividend pay-out ratio

REITs have to out more than 90% of their profits as dividends. The high dividend pay-out means less retained earnings. According to both the pecking order theory and trade-off theory this means a positive relationship between dividend pay-out and leverage. Les

retained earnings, means there is a higher incentive for borrowing. The dividend pay-out ratio is calculated as the dividend paid divided by the equity.

(24)

4.2.6 Excluded variables

Some of the variables, such as tangibility, growth or investment opportunities, firm specific factors such as sector, that are used in other papers are excluded in this thesis for several reasons which I will explain. Because Equity REITs are studied it is assumed that the firms have significant real estate properties in their balance sheets and thus the tangibility is high. High level of tangibility means higher level of leverage as the collateral and liquidation value of these assets is higher. I found that growth and investment opportunities are mostly approximated by the market-to-book ratio of firms in previous studies of Feng, Ghosh and Sirmans (2007) and Mamoru (2009). However what the financial crisis learned us was that these ratios contain a lot of noise and vague information, which creates a bubble. To exclude this bubble and its noise and vague information, I do not look at the market-to-book ratio. Another potential way to look at the investment opportunities was to look at the capital expenditure (capex), but even this capex figure sometimes contains information which is not related to investment opportunity. For example, the capex also included capital spend on maintenance, renovation or rebuilding of existing real estate properties, what clearly is not an expansion of the real estate assets the REIT has. This paper is indifferent about the sector. Real estate is real estate, whether the property is an office building, a retail shopping center or a residential complex. It is assumed that real estate is long term assets and the financing of these long term assets should be done accordingly. For these reasons these variables have been left out of this study.

A summary of all the capital structure theories with respect to the variables can be found in Appendix E (table X). The expected signs and the actual signs found through studies are also highlighted in this table.

4.2.7 Expectations of the debt ratios and variables

First, I expect the TDR to be lower in Period 2 than in Period 1. If this is true then REITs are more equity financed (long term financed) then debt financing. However because the shift can also take place from the STDR to the LTDR, so that my second expectation is: STDR in period 2 is lower than the STDR in period 1. If this is true then there is more long term

financing, which can either be equity of long term debt. Finally, I expect the LTDR in period 2 to be higher than in Period 1 as this indicates a shift towards long term financing.

The regression model expectations of the models TDR , LTDR and STDR in Period 1 and 2 are also presented in Table (9). If the expectation hold it is safe to assume that REITs have begun to better match their capital structure to the assets they hold. Especially with equity and long term debt financing, which are a long term financing instrument.

(25)

5. Statistical and Empirical Analysis

In this section I will present and analyze the results first for the descriptive statistics in Period 1 and 2 of the sample, secondly the models that are used in this thesis are presented, and thereafter the emperical results of these models. Finally, the comparison of two sample data to find any structural break that indicates that the capital structure of REITs changed.

5.1 Descriptive statistics and analysis

All the variables are expressed in US dollars. The descriptive statistics of the sample for Period 1 are displayed in Table 1.8 The results of the sample for Period 2 are displayed in Table 2. The Period 1 REITs have an average total asset value of $2,691 million with a minimum of $3.5 million and a maximum of $52,894 million. The average total debt for Period 1 is $1,750 million, the average long term debt is $879 million and the short term debt

averages at $640 million respectively. The mean of the equity is $843 million. REITs in this sample have an average of about $57.863 million in earnings before interest, taxes and depreciation. Finally, REITs in the sample on average paid out $21.062 million in dividends.

Table Y for the Period 2 sample indicates that REITs have an average total asset value of $4,606 million. The average total debt is $3,120 million, whereas the average for long term and short term debt is $1,235 million and $1,460 million. The average of the equity is $1,418 million. The average of the earnings before interest, tax and depreciation is $80.551 million. Finally, the dividends paid out in the Period 2 sample have an average of $27.925 million.

Table 3 shows us the comparison between the sample of Period 1 and Period 2. The mean of each variable is compared between the periods and the change is presented in

percentage. The results show that the total asset changed 41,57% from Period 1 to Period 2. Total debt (TD) increased with 43.91% and long term and short term debt increased with 28.8% and 56.16%. The mean of the Equity increased 40.55% from the sample of Period 1 to 2. The earnings before interest, taxes and depreciation increased with 28.17%, whereas the dividends increased with 24.58%.

The descriptive statistics of the dependent variable leverage for both Period 1 and 2 are displayed in Table 4. The total debt ratio increased, from a mean of 0.572453 in Period 1 to a mean of 0.599135 in period 2, so the REITs increased the leverage in this period. The mean of the long term debt ratio decreased from 0.353826 in Period 1 to 0.339759 in Period 2.

(26)

REITs decreased their long term debt. The mean of the short term debt ratio increased from 0.159365 to 0.177495. There was more short term borrowing by REITs in Period 2 then in Period 1. The equity ratio (ER) Table 4, shows that the ratio dropped from 0.427547 in Period 1 to 0.400865 in Period 2. Meaning that the equity decreased relatively from Period 1 to Period 2. Overall, the long term financing that consists of the long term debt and equity dropped from 0.781373 (0.427547 + 0.353826) in Period 1 to 0.740624 in Period 2. I will test these findings in a regression framework, first I will discuss the empirical model and the assumptions in the next subsection. Finally in the last subsection the results of the empirical analysis will be presented and discussed.

5.2 Empirical Model

Using the information from the previous parts, I construct leverage models in this part. The models for Period 1 and Period 2 are tested to answer the question: Did REITs change their Capital Structure after the financial crisis ? The question evolves looking in the fact whether the leverage variables have a structural break and are different from before and after the financial crisis in 2008 and 2009.

The general leverage model :

Model 1: Leverage = f(size, profitability, earnings risk, dividend pay-out ratio,

dequity)

Al the models with one represent the parameters for data of Period 1, al the models with two represent the parameters for data of Period 2. The model 1.1 is the total debt to total asset ratio for Period 1, model 1.2 is the long term debt to total assets for Period 1, model 1.3 is the short term debt to total assets for Period 1. The same goes for the models 2.1 , 2.2 and 2.3 they respresent the results for Period 2, that is after the crisis, whereas Period 1 is before the crisis.

The models that will be regressed are :

(1) Leverage = a + b*taa + c*Divr + d*Ebitdar + f*Pebitda + g*Pe + Є

Leverage = ( TD/TA , or LTD/TA, or STD/TA )

(1.1)TD/TA = a + b*taa + c*Divr + d*Ebitdar + f*Pebitda + g*Pe + Є

(27)

(1.3)STD/TA = a + b*taa + c*Divr + d*Ebitdar + f*Pebitda + g*Pe + Є

(2.1)TD/TA = a + b*taa + c*Divr + d*Ebitdar + f*Pebitda + g*Pe + Є

(2.2)LTD/TA = a + b*taa + c*Divr + d*Ebitdar + f*Pebitda + g*Pe + Є

(2.3)STD/TA = a + b b*taa + c*Divr + d*Ebitdar + f*Pebitda + g*Pe + Є

The abbreviation TD stands for total debt, LTD for long term debt, STD for short term debt, taa for size, Ebitdar for the profitability ratio, pebitda for the earnings risk, Divr for dividend pay-out ratio, pe for the percentage change in equity. The model has an interception variable a, and c till g are coefficients of the respective variables. The residual error term is defined by E.

Since the dataset consists of a time series panel of REITs a Classical Linear Regression Model with Ordinary Least Square(OLS) will not be sufficient as the estimates will not be efficient. To improve this efficiency of the coefficients estimated, the panel data estimation technique Panel Least Squares(PLS) is used. This technique accounts for the effect that the same cross-sectional relationship of individual REITs is observed at different points in time. Two widely used techniques are the Fixed Effects Model and Random Effects model. The Hausman and Fixed Effects Likelihood tests evaluates the significance of whether it should account for cross section and period effects in the model and whether the effects in the models are significant with random effects or fixed effects. In this paper I will use the Fixed Effects model for both the individual REITs and the time periods.9

As pe is rather a new variable it may not be a perfect variable for the leverage models, however it may increase the fit of the models and thereby give additional information. Therefore I will add Pe into the models if it increases the adjusted R2.10 The results of the adjusted R2 with and without Pe in the model can be obtained in Appendix X.

The models 1.1, 1.3 and 2.3 show lower adjusted R2, when the variable pe is included therefore I conclude that the variable Pe is excluded from these models as it decreases the fit of the model. Whereas the models 1.2 , 2.1 and 2.2 show a higher adjusted R2, therefore I conclude the variable Pe is included in these respective models, because it increases the fit of the model.11

Table 5 shows a correlation matrix for Period 1 and Period 2 independent variables. If there are correlations which are greater than 0.7 there is potential problem of dependency in the model, but as one can see from Table 5 there is no high correlation between any variable,

9

See results of the Hausman test and Fixed Effects Likelihood test in the Appendix G.

10

The Adjusted R2 can be used as a tool to validate variables. See Brooks (2002, p.137)

(28)

meaning that the variables are independent of one and other.

To exclude heteroskedasticity in the model of the ordinary PLS the standard errors and covariance are period weighted (PCSE) and corrected for the degree of freedoms which has no significant effect because the sample is relatively large. I assume that there is no

endogeneity problem in my models. Basically meaning, that there is no correlation between the independent variables and the error term in the regression models and the coefficients in the PLS with PSCE are not biased.

If there is no significant change between period 1 and 2, the coefficients for panel A and panel B regressions should be equal (Braker and Wurgler, 2002). The interaction will be measured with a Chow-test. The Chow-test is a statistical and econometric test commonly used in time series analysis to analyze and signal for the presence of a structural break.12 The results of the chow test will follow the section of the results of the empirical analysis.

5.3 Empirical Results and Analysis

The regressions in this section are set out to test the findings from the descriptive statistics table 4 in a regression framework with all the independent variables and to look whether there is an empirical significant change in Capital Structure of REITs before and after the financial crisis. I will also look at the determinants and whether or not they match their theoretical expected signs as presented in table 9.13 The regression models were estimated by PLS with Panel Corrected Standard Errors(PSCE).14

Tables AA and BB contain empirical results for TDR, LTDR and STDR for Period 1 and 2. In the sample of Period 1 the models explain between 79% and 85% of the within sample variance in the dependent variables TDR, LTDR and STDR. The standard errors and p-values show that the dependent variables are overall strongly significant for Model 1.1 and 1.2. In model 1.3 almost all the variables are insignificant, however they are still included as they represent the economic capital structure theory variables. The F-statistics show all the models are strongly significant. In model 1.2 Pe variable is included and in the models 1.1 and 1.3 it is excluded as they showed a better fit without the variable.15 The coefficients for the TDR model 1.1 are all positive, meaning that size has a positive effect on leverage. As

12

Read Chow (1960) to learn more about the fundamentals of the chow test. See Appendix X for the Chow test statistic and the way it is calculated.

13_{The summary table can be found in Appendix H} 14

This model technique is first used by Prais-Winsten. More information about this model can be found in Trend Estimators and Serial Correlation by Prais andWinsten(1954).

(29)

the size increases the leverage will increase as well. The Divr coefficients shows that if the pay out of dividends gets larger, the leverage increases, this follows that if REITs pay out more of its profits in dividends the lesser funds are available to finance new projects with internal funds, according to the pecking order theory leverage is the next best option. Ebitdar shows that if profit increases the leverage also increases. This is not in line with the trade of theory nor the pecking order theory. Pebitdar is the earnings risk of the REITs and has a positive effect that is almost close to zero. The variable is not significant at the 10%,5% and 1% respectively with a p-value of 0.4787.

Model 1.2 for the LTDR shows the same signs as model 1.1. Only the variable Pe has a negative effect on long term debt. Pe stands for the increase in equity, and in the periods that the Pe increases there is less need of long term leverage. Same as Pebitdar in model 1.1 the negative effect on long term debt is small and close to zero and the variable is only

significant at the 10% and 5% level with a p-value of 0.0451. Pebitdar has a positive effect close to zero and is insignificant at the 10%,5% and 1% with a p-value of 0.1661. In model 1.3 for the STDR size has a negative effect, meaning that if the size increases the short term leverage decreases and vice versa. This means that smaller REITs lend more short term to finance projects, and larger REITs finance more with equity or long term debt.

Table AA: Period 1 regressions

Variable Model 1.1 = TDR Model 1.2 = LTDR Model 1.3 = STDR

C 0.136703 (0.025506) -0.203211 (0.038544) 0.205651 (0.029327) TAA 0.060464 (0.003640) 0.080945 (0.005651) -0.007166 (0.004193)**** Divr 0.178500 (0.031239) 0.147644 (0.044456) 0.003090 (0.034918)***** Ebitdar 0.735548 (0.135584) 0.769142 (0.204523) 0.015401 (0.155543)****** Pebitdar 3.02E-06 (4.26E-06)* 9.67E-06 (6.98E-06)** 5.10E-06 (4.84E-06)****** Pe -4.05E-05 (2.02E-05)*** Observations 5027 4486 4963

(30)

R-squared 0.844814 0.812447 0.794507 F-statistic 101.0191 80.50774 53.36318

F-prob 0.00 0.00 0.00

*p-value = 0.4787 ** P-value = 0.1661 *** P-value=0.0451 **** P-value = 0.0875 *****P-value = 0.9295 ****** P-value= 0.9211 *******P- value = 0.2918

Between the apprentices are the standard errors of the coefficients. C represent a constant. TAA is the log of total assets (ta). Divr is the dividend pay-out ratio calculated as dividend paid out divided by ta. The profitability, Ebitdar is calculated by dividing the EBITDA by ta. Pebitdar is the earnings risk which is measured as the percentage change in EBITDA. Pe is the percentage change in equity.

Table BB: Period 2 regressions

Variable Model 2.1 = TDR Model 2.2 = LTDR Model 2.3 = STDR

C 0.678292 (0.065552) 0.347785 (0.075801) -0.091727 (0.034355) TAA -0.017608 (0.008744)* -0.002454 (0.010113)**** 0.036704 (0.004564) Divr 1.347000 (0.313831) 0.540720 (0.359858)***** -0.056174 (0.038873)******** Ebitdar 1.694232 (0.397222) 2.012515 (0.467299) -0.287563 (0.164315)********* Pebitdar 5.21E-05 (0.000183)** 0.000317 (0.000237)****** -0.000298 (0.000147)********** Pe -0.001664 (0.000819)*** -0.001710 (0.000961)******* Observations 1976 1975 2481 R-squared 0.918341 0.921687 0.921520 F-statistics 114.2394 120.2336 120.0249 F-prob 0.00 0.00 0.00

*P-value = 0.0442 **p-value= 0.7765 ***p-value= 0.0424 ****p-value= 0.8083 *****p-value= 0.1331 ******p-value= 0.1812 ******* p-******p-value= 0.0753 ******** p-value = 0.1486 ********* p-value = 0.0802 ********** p-value = 0.0421

In the sample of Period 2 the models explain between 91% and 93% of the within sample variance in the dependent variables TDR, LTDR and STDR. The standard errors and p-values show that the dependent variables are overall significant for Model 2.1 and 2.2. In

(31)

model 2.3 almost all the variables are insignificant, however they are still in the model for the same reason as in model 1.3. The F-statistics show all the models are strongly significant. In model 2.3 is excluded as it showed a better fit without the variable.16 In model 2.1 for the TDR size has a negative effect. If size increases the leverage decreases and vice versa. After the crisis REITs decreased their leverage if they got larger as expected from the AMT. The variable size is only significant at the 10% and 5% level with a p-value of 0.0442. Pe has a negative effect which is consistent with a priori expectations, the results in the models of Period 1 and is significant for the 10% and 5% level with a p-value of 0.0424. The coefficient for Pebitdar is highly insignificant with a p-value of 0.7765. Model 2.2 has a negative size effect, which is also as expected by the AMT. As REITs get larger they finance less with long term debt. In model 2.2 only C and Ebitdar are strongly significant. Size, Divr, Pebitdar and Pe are in significant with p-values of 0.8083, 0.1331, 0.1812 and 0.0753 respectively. Except for Pe which is significant at the 10% level , the variables are

insignificant. Model 2.3 shows that the STDR increases as the size increases. This was not expected by the AMT. After the crisis it was expected that there is less short term financing. So the correlation of size and short term debt was expected to be negative. Increasing Divr and Ebitdar have a negative effect on short term borrowing as expected by the AMT. REITs pay out more, but also borrow less short term and as profitability increases the need of short term borrowing decreases.

When comparing the models in both sample I find the following. In sample Period 1 size had a negative effect on STDR whereas in Period 2 size has a positive effect on STDR. If I take in to consideration the results from table Z that the TA changed by 41,57% and size is measured as the log of ta, the size of the REITs in Period 2 is relatively larger than in Period 1. The STD grew with 56,16% compared to Period 2 (Table 3). That is the main reason that there is more STD in Period 2 (Table 1 and 2) and a positive correlation between STDR. The regression model 1.3 and 2.3 show this change in coefficients. These findings do not support the AMT, which expected a negative sign of size in Period 2(Table 9). The trade-off theory does predict positive effects of size and the pecking order theory negative effects. Previous studies done by Rajan & Zingales (1995), Fama & French(2002), Feng, Ghosh, Sirmans (2007), and Bond & Scott (2006) found positive effects of size.I found mixed results as size in model 1.1 and 2.1, 1.2 and 2.2 shows that the coefficient changed from positive in period 1 to negative in Period 2, meaning that in sample Period 1 size had a positive effect on the TDR and LTDR and in sample Period 2 size had a negative effect on these leverage ratios. This change does not support the AMT, because to get a better fit with the assets it is

(32)

actually expected that ER and LTDR increase. The negative size effect is acceptable if the shift took place from STD to LTD and E, but this is not the case in Period 2 (see also table 3 and 4).

The dividend pay-out, profitability and earnings risk also changed coefficients. From positive signs in model 1.3 to negative sign in model 2.3. Implying that a higher dividend pay-out, profitability and earnings risk decrease the STDR which is also in line with AMT(Table 9). As REITs should decrease their STD to get a better match with their assets that are long term. All the other coefficients except for these discussed were similar in both samples.

To look at whether the Period 1 models statistically differ from the Period 2 models I perform 3 chow test analysis, which are presented in Table 6, 7 and 8. Table 6 shows that the TDR model 1.1 and model 2.1 have structural break, so the coefficients for both models are different. The outcome of the chow test statistics follow a f-distribution and the p-value of the f-statics at 1% is 2,80471074 and the outcome from the chow test statistic is 718,7233. Table 7 shows that the LTDR model 1.2 and 2.2. have a structural break, so the coefficients for both models are different. The chow test statistic with a p-value of 1% is 2,80471365 and the outcome of the chow test for the respective models is 554,8244. Finally, the STDR model 1.3 and 2.3 in table 8 shows that the chow test statistic with a p-value of 1% is 2,80474986 and the outcome of the statistic for the models is 317,9959 so that the models have a structural break and the coefficients of both models are not equal. Overall, this means that the capital structure model of REITs did change before and after the crisis.

The capital structure of US REITs before and after the 2007-2008 financial crisis.