AND THE EFFICIENT CAPITAL STRUCTURE FOR D UTCH HOUSING ASSOCIATIONS ALM

ALM

AND THE EFFICIENT CAPITAL STRUCTURE FOR

D

UTCH HOUSING

ASSOCIATIONS

Rijksuniversiteit Groningen

Abstract

This thesis explores the existence of efficient financing policies for Dutch housing associations while complying with borrowing constraints set by regulators. This is achieved by studying the risk and return on assets and equity associated with different target debt ratios. This relationship is analysed by using a Monte Carlo simulation model to simulate the future cash flows, asset value, financing policy, and investment policy of a Dutch housing association. Results are in line with findings from studies analysing the benefits of leverage on real estate investments. Dutch housing associations can set efficient target debt ratios without violating the borrowing constraints. Furthermore, violating the constraints and using too much debt can be devastating as returns start to decrease and risk start to increase.

JEL classification: C15, G11, G32

MSc Thesis Finance

Frank B. Verbeek, s1914073

University of Groningen, January 10th, 2014

Supervisor RuG: Prof. Dr R.E. Wessels

2

Contents

1.INTRODUCTION ... 3

2.DUTCH HOUSING ASSOCIATIONS ... 5

3.LITERATURE REVIEW ... 6

3.1DEBT FINANCING ... 6

3.2CAPITAL STRUCTURE DECISIONS... 7

3.3MONTE CARLO SIMULATION... 8

4.SIMULATION MODEL ... 10

4.1BALANCE SHEET ... 11

4.2OPERATING CASH FLOWS ... 12

4.3ASSET VALUE ... 14

4.4INVESTMENT POLICY ... 17

4.5LIABILITIES ... 18

4.6AGENCY COSTS ... 19

4.7SENSITIVITY ANALYSES ... 20

5.RISK AND RETURN ... 20

5.1RETURN MEASUREMENT ... 20 5.2RISK MEASUREMENT ... 21 6.DATA ... 23 6.1SIMULATION PARAMETERS ... 23 6.2MODEL INPUT ... 24 7.RESULTS ... 25

7.1EFFICIENT TARGET DEBT RATIOS ... 25

7.2RESTRICTIONS CFV AND WSW ... 30

8.CONCLUSION ... 30

8.1RESEARCH QUESTION ... 31

8.2FURTHER RESEARCH ... 32

REFERENCES ... 34

APPENDIX A:CASH FLOW OVERVIEW ... 38

APPENDIX B:MODEL INPUTS ... 39

3

1.

I

In finance, the capital structure of companies has been widely studied. Modigliani and Miller (1958) commenced with studying the impact of capital structure on firm value with their debt-irrelevance proposition. Since Modigliani and Miller several articles have studied the existence of an efficient capital structure (Brennan and Schwartz, 1984; Fischer Heinkel and Zechner, 1989; Cannaday and Yang, 1996; Boyd, Ziobrowski, Ziobrowski and Cheng, 1998; McDonald, 1999; Hennessy and Whited, 2005). The general consensus is that an efficient capital structure decision is driven by a trade-off between tax benefits and potential distress cost.

Whilst the advantages of debt financing are thoroughly studied for listed companies, research is scarce for non-listed companies. This study is aimed at studying the advantages of debt financing for non-listed companies. More specifically, companies which cannot issue equity and are required to change their investment policy to alter their capital structure, more explicitly Dutch housing associations. The advantages of debt financing are studied by analysing the risk and return on assets and equity associated with different target debt ratios. Maintaining a target debt ratio is familiar among financial managers. As Graham and Harvey (2001) illustrate, in a survey among CFOs, 81% of the respondents indicate that they maintain a target debt ratio. Previous research in the field of real estate investment suggests that efficient target debt ratios exist (Cannaday and Yang, 1996; Boyd, Ziobrowski, Ziobrowski and Cheng, 1998; McDonald, 1999). According to McDonald (1999), the use of debt adds to firm value if companies are required to pay taxes and have positive expectations about the real estate investment. Moreover, Van der Spek and Hoorenman (2011) find that efficient capital structures exist. They conclude that portfolios with debt ratios up to 40% are efficient, adding additional debt is likely to decrease returns. Possible advantages associated with debt financing exist for Dutch housing associations as they are required to pay taxes. However, attracting too much debt can generate diminishing returns due to costs associated with financial distress.

Goal of this thesis is to analyse the existence of an efficient capital structure for Dutch housing associations without violating borrowing constraints set by the regulators. This should result in strategic management advice regarding the capital structure, specifically, the existence of efficient target debt ratios within the restrictions imposed by the Central Housing Fund (CFV) and the Dutch Social Housing Guarantee Fund (WSW). Target debt ratios are inefficient when higher returns can be realized at the same or lower risk, or when the risk can be lowered without reducing returns. Resulting in the following research question:

4

With the central hypothesis:

“As adding debt increases risk there must be a point where adding debt becomes inefficient, the risk added is not fully compensated by the returns generated.”

To prevent housing associations from taking excessive risks, they are governed by the Central CFV and the WSW. The CFV and WSW impose restrictions on the way housing associations attract debt, value their dwellings, and the level of risk being taken. To incorporate the borrowing restrictions and to measure if the restrictions allow housing associations to set efficient target debt ratios the following hypothesis will be tested:

“An efficient capital structure is within the borrowing constraints imposed by the CFV and WSW”.

To answer this question a Monte Carlo simulation model, based on the model developed by Van der Spek and Hoorenman (2011) and the dynamic trade-off model developed by Titman and Tsyplakov (2007), is used to simulate future cash flows generated by the housing association’s financing and investment policy. According to Millet (1988) the use of a Monte Carlo simulation helps in forecasting the financial and economic environment in preparation for subsequent decisions. Kelliher and Mahoney (2000) argue that Monte Carlo simulation is preferred over alternative approaches as it offers great possibilities for understanding risks associated with strategic decisions. The relationship between the financing policy and risk and return is studied by simulating the future cash flows of housing associations with different target debt ratios and with a dynamic investment policy. Cash flows are created by renting out, constructing, renovating, and selling dwellings. Furthermore, after each year in each scenario the value of the assets, based on market value in rented state, is calculated by simulating the cash flows from this year for the next 15 years as if the dwellings are owned by a rational commercial investor. Additionally, to find inefficient financing policies, an analysis in the risk-return field is necessary. As return measures I use the annualized Total Return (TR) on assets and equity. Moreover, the use of a Monte Carlo simulation results in a distribution of possible outcomes which provides for a measure of risk. Additionally, Sharpe and Sortino ratios are used to measure the returns generated per unit of risk. To test whether efficient target debt ratios are within the borrowing constraints the probability of violating the constraints set by CFV and WSW is calculated. This probability is calculated as the number of scenarios with a restriction violation divided by the total number of scenarios, the so-called path-probability.

5

The remainder of this thesis is organised as follows. Section 2 presents a brief description of the role and functioning of housing associations. Section 3 contains a literature review on relevant prior research regarding the capital structure. Additionally, Section 3 presents a literature review on the use of Monte Carlo simulations for real estate valuation. The description how the simulation model is build up is given in Section 4. The methodology used to analyse the risk and return associated with a target debt ratio is presented in Section 5, whereas the data and parameters used are given in Section 6. Section 7 contains the results of the Monte Carlo simulations and the sensitivity analysis performed. Section 8 concludes.

2.

D

Similar to most West European countries, the Dutch government has retreated from the housing market. This development led to a reduction in financial support for the social rental sector and, consequently, housing associations needed to adopt a more market oriented approach (Kramer and Van Welie, 2001). Nowadays, Dutch housing associations are non-profit organisations that are required to operate in the interest of housing (Kramer, Kronbichler, and Van Welie, 2011). According to Boelhouwer (2007), while Dutch housing associations do have social responsibilities, they cannot be regarded as government organisations, nor are they pure market parties.

Primary task of the associations, stipulated in the Social Rental sector Management Order (BBSH), is to provide good, affordable housing for those who are unable to find a dwelling in the market (Elsinga, Haffner, and Van der Heijden, 2008). Furthermore, associations are required to deliver good quality housing and liveable neighbourhoods. These social objectives are to be funded by their capital surplus. This capital surplus is primarily represented by the increase in value of the dwellings they own. To be able to get funds to invest in their social objectives, housing associations are required to either sell part of their housing stock or take up a loan (Kramer, Kronbichler, and Van Welie, 2011). According to Boelhouwer (2007) there are no rules on the use of the proceeds from the sale of rented dwellings, the only restriction is that the proceeds should be spent on housing related activities.

6

Furthermore, as repayment of short term debt involves a higher liquidity risk, the WSW limits associations in the use of short-term debt.

From a financial perspective Dutch housing associations are interesting subjects as they do not have shareholders and thus cannot issue equity to finance investments. Net worth grows through an increase in the value of the assets. This limitation implies that financing gaps need to be filled by either retained earnings, the sale of dwellings, or attracting more debt. Moreover, as there are no shareholders, retained earnings will not be distributed as dividends but, instead, will be used to invest in more dwellings or other social objectives. Additionally, although Dutch housing associations are non-profit organisations and not pure market parties, they are required to pay corporate tax leading to a tax shield when using debt financing.

3.

L

After the pioneering work of Modigliani and Miller (1958) there have been various studies on the effects of using debt financing. The next section gives an overview.

3.1 DEBT FINANCING

In the traditional trade-off model, developed by Modigliani and Miller (1963), the main benefit of debt arises from tax deductibility. Additionally, according to Miller (1977), primary costs are associated with financial distress. Van Binsbergen, Graham, and Yang (2010) argue that, in line with Modigliani and Miller, tax savings due to the deductibility of interest is one of the main benefits of using debt financing. Downside of using debt is increasing the risk of the investment, thus the amount of risk an investor is willing to take is of importance.

7

of leverage on a private real investment trust (REIT). They find, in line with previous studies, that efficient capital structures exist where portfolios with debt ratios up to 40% are efficient. Adding additional debt is likely to decrease returns.

This thesis adds to the existing literature by determining the efficient capital structure for corporations invested in real estate and limited in equity financing, in this case Dutch housing associations. Building on the conclusions of previous research, as Dutch housing associations are required to pay taxes, adding debt can potentially increase returns. Different from the Tyrell and Bostwick (2005), the model developed in this paper does not incorporate increasing interest rates if leverage increases. Loans obtained by Dutch housing associations are backed by the WSW and I assume that this will prevent interest rates to increase. Furthermore, like Van der Spek and Hoorenman (2011), I will employ a simulation model to analyse the existence of efficient debt ratios.

3.2 CAPITAL STRUCTURE DECISIONS

Several articles formulate dynamic trade-off models to study capital structure decisions. For example, Bradley, Jarrel, and Kim (1984) use a dynamic trade-off model to find a firm’s optimal capital structure. They argue that an optimal capital structure involves a trade-off between tax advantages generated by debt and various leverage related costs. Furthermore, they demonstrate that financial distress costs are significant and that optimal firm leverage is inversely related to the variability of firm earnings. Strebulaev (2007) tests data generated from a dynamic trade-off model with empirical data from other studies. Strebulaev concludes that proper study of capital structure requires a dynamic trade-off model that endogenizes financing and investment decisions. Models developed by Brennan and Schwartz (1984), Hennessy and Whited (2005) and Titman and Tsyplakov (2007) employ an endogenous financing and investment policy. Brennan and Schwartz and Titman and Tsyplakov assume in their model that free cash flows are distributed to shareholders. Brennan and Schwartz analyse a firm’s intertemporal investment and capital structure policy. Titman and Tsyplakov examine how being able to dynamically adjust the debt ratio affects the deviation of actual ratios from target debt ratios. Moreover, Titman and Tsyplakov address the possibility that economic shocks can move firms away from their target debt ratios and may also cause target debt ratios to change over time. Hennesy and Whited (2005) develop a dynamic trade-off model with an endogenous choice of leverage and investment decisions to study the relationship between target debt ratios and firm value. Hennesy and Whited show that there is no preferred target debt ratio, firms can be savers or heavily levered. Additionally, they show that leverage is path dependant.

8

structure. According to Fischer, Heinkel and Zechner, the optimal capital structure decision depends on the benefits of debt financing, underlying asset variability, the riskless interest rate, and the size of the costs of recapitalizing. They model the value of a levered firm as a function of the market value of the unlevered assets and of its debt ratio. Furthermore, Fischer, Heinkel and Zechner, unlike other research, generate predictions about firms’ capital structure decisions while allowing the actual debt ratio to deviate from the target ratio. Kane, Marcus and McDonald (1985) investigate capital structure choices by using an option valuation model based on a competitive market for houses. In their model, a firm can adjust its capital structure by the amount of outstanding debt becoming due. They argue that the advantage of debt is the extra rate of return earned by optimally levered firms relative to an otherwise identical unlevered firm.

The model developed in this thesis is similar to those developed by Brennan and Schwartz (1984), Hennesy and Whited (2005), and Titman and Tsyplakov (2007) as I use a dynamic simulation model with endogenous investment and financing decisions. The model developed in this thesis shows similarities with the model developed by Fischer, Heinkel and Zechner (1989) and Kane Marcus and McDonald (1985). In my model I allow the actual debt ratio of Dutch housing associations to deviate from their target ratio within certain boundaries and, moreover, the housing association can only adjust its capital structure with the amount of outstanding debt becoming due.

This thesis extends the existing dynamic capital structure literature as Dutch housing associations are limited in attracting equity. Housing associations cannot issue equity and do not have shareholders. Different from the model developed by Brennan and Schwartz (1984) and Titman and Tsyplakov (2007), cash flows are not distributed to shareholders, but instead are invested in assets. Key difference between my model and previous research is that, to simulate future cash flows and the development of assets and equity, I employ a Monte Carlo simulation.

3.3 MONTE CARLO SIMULATION

9

Millet (1988) states that scenario analysis aims at two purposes, the first is forecasting the environment in preparation for subsequent decisions, the second purpose is to be able to evaluate strategic choices against the generated scenarios. Thus, simulation can be used to aid the selection of a strategy through developing a range of forecasts (Bunn and Salo, 1993). Baroni, Barthélémy, and Makrane (2007) conclude that simulating cash flows to value assets in real estate investments, based on a Monte Carlo simulation model, permits the user to estimate the portfolio’s price distribution for any time horizon and allows for easy value at risk calculations.

In order to get reliable estimates of risk and return a large number of economic scenarios need to be simulated. A way to obtain estimates of risk and return is by using a Monte Carlo simulation which is often used in Asset Liability Management (ALM) to model the economic risk and return factors (Kramer and Van Welie, 2001). Monte Carlo simulation generates a large pool of future economic variables by quantifying the relevant risk factors. The advantages of a Monte Carlo simulation are that it is robust in analysing the interactions between uncertain inputs that are represented by a range of possible values, and with data that is not normally distributed (Kelliher and Mahoney, 2000). It is preferred over alternative approaches as it offers great possibilities for understanding risks and assists in the understanding why certain decisions are the right ones. Furthermore, Monte Carlo simulation has strong visual aspects leading to better understanding for the ones making the calls. The accuracy of the simulation depends primarily on the accuracy of the input variables used in the model (French and Gabrielli, 2005).

10

4.

S

M

The simulation model developed in this paper uses an endogenous dynamic investment policy - which is affected by the association’s net cash flows and debt ratio - to explore the relationship between a target debt ratio and the risk and return associated with this target ratio. The corresponding risk and return spectrum for each target debt ratio is found by inserting different target ratios as input and juxtaposing the associated risk and returns. For each target ratio 500 different scenarios are used. Each scenario represents a possible future state of different economic and financial variables for the next 15 years. These variables affect an association’s cash flows and asset value. For each scenario the investment policy in each year is determined, associated cash flows are calculated, financing policy is determined and checked if it remains within the constraints, and asset value is computed. The asset value is calculated by calculating the net present value (NPV) of simulated expected future cash flows generated by the dwellings as if they are owned by a commercial investor. The simulation, using 100 scenarios, to determine the asset value is started in each year in each scenario within the main simulation. This simulation within a simulation leads to an explosion of the calculation time needed to determine the risk and return spectrum1_{. A detailed} description of the association’s response to different scenarios is given in the next paragraphs. Like Titman and Tsyplakov (2007), I start by giving a short summary of the simulation model by displaying a timeline:

At year t = 0

I assume that the housing association already exists and the value of the dwellings owned at year t = 0 is determined by calculating the market value in rented state. Moreover, the target debt ratio is being set. I assume that the housing association’s capital structure matches the set debt ratio at the start of the simulation. If, for example, the target debt ratio is set at 20% I assume that the starting capital structure contains 20% liabilities and 80% equity. Per simulation cycle the debt ratio will be increased with 10%-points, starting at a debt ratio of 0%, to get the corresponding risk and return outputs based on the values obtained at the end of year 15.

At each subsequent year

The firm realizes the cash flows which are affected by the current economic state, last year’s debt ratio, and resulting investment policy. The market value of the dwellings in rented state will be recalculated taking the median of the NPV resulting from a simulation

1

11

with 100 scenarios for the future cash flows of the dwellings as if they are owned by a commercial investor. Based on the end of year debt ratio, inter alia affected by the change in asset value, extra financing to pay for cash shortfalls, and the acceptable debt ratio range, next year’s investment policy will be determined.

At year t = 15

The asset and equity value at the end of year 15 will be used to calculate the return achieved over the selected time period. The value of equity is determined by the difference between asset value and value of debt.

The next section will give a description of how the model is set up. The description starts with the assumptions regarding the structure of an association’s balance sheet followed by how cash flows are simulated, asset values are calculated, liabilities are incorporated, and how the investment policy changes over time.

4.1 BALANCE SHEET

The use of Dutch housing associations enables me to make some assumptions that make the model more comprehensible. For example, the asset side of a housing association’s balance sheet consists almost entirely out of the dwellings owned and the liability side is dominated by long-term debt (Kramer and Van Welie, 2001). This brings me to the assumption that housing association’s asset value is determined by dwellings and cash only, no other assets are taken into account. Furthermore, I assume that the debt portfolio consists out of short-term one year corporate bonds and ten year corporate bonds. Equity arises through an increase in the value of the assets minus the value of debt outstanding. In this model it is assumed that an association defaults if the value of equity equals zero or if all dwellings are sold.

12

important to note that the cash flows are discounted solely to determine the asset value at a certain point in time; I am not calculating the NPV of the whole housing association. Further details and the calculation of market value in rented state are presented in Section 4.3.

4.2 OPERATING CASH FLOWS

According to Kramer and Van Welie (2001) the main cash flows generated by housing associations are concomitant to rent income and interest expenses. The methodology presented in Baroni, Barthélémy and Makrane (2007) will be used as a starting point to model the cash flows. The operating cash flow can be shown as:

( ) ( ) (4.1)

W

= cash inflow in in year in scenario = cash outflow in year in scenario = corporate tax rate

The input parameters used to determine the cash flows depend on the simulated economic parameters. The next section will elaborate on the operating cash in- and outflows. An example of a cash flow overview for a Dutch housing association is presented in Appendix A.

4.2.1 OPERATING CASH INFLOWS

Following Baroni, Barthélémy and Makrane (2007) the cash inflows generated by a real estate investment generally take the form of rent payments. This entails that cash inflows are determined by total potential rent income minus rent loss due to vacancy. The cash inflows can be depicted as:

( ) ̅ (4.2)

W

= vacancy rate in scenario

̅ _{= average number of dwellings in year in scenario}

The rent increases for social rents are regulated by the government and equals the price inflation plus an additional increase to compensate for a new tax regulation2. This rent increase is only applicable for rent levels below the maximum social rent allowed. Furthermore, housing associations cannot raise rents to their desired rent level unless a tenant turnover takes place. In this model I assume that the housing association maintains three rent levels. A low rent level to provide

2

13

housing for the lowest income classes, a medium rent level, and a rent level equal to the maximum social rent level allowed; the liberalization limit. When a tenant turnover takes place within a certain rent class the rent will be increased to the liberalization limit until a set division between the low, medium, and maximum rent level asked is achieved. This division of rent classes can be displayed as:

( ) (4.3)

Where:

[( ) ] (4.4) [( ) ] (4.5)

And: = low rent in year t in scenario

= medium rent in year t in scenario

= maximum social rent in year t in scenario

= percentage of dwellings at low rent level in year t in scenario = percentage of dwellings at medium rent level in year t in scenario

= tenant turnover rate in year t in scenario = social objective for low rent level

= social objective for medium rent level

The social objective of a housing association is a fixed input and represents the maximum level of dwellings with the maximal social rent, i.e. a certain percentage of dwellings will be rented out against a rent lower than market rent to provide for payable housing for the low income classes. The inflation rate will be simulated and the tenant turnover rate depends on the house price index where the house price index will be simulated.

4.2.2 OPERATING CASH OUTFLOWS

The cash outflows are determined by summing the expenses which can be categorized into maintenance, turnover, wages, and other costs.

[ ] ̅ (4.6)

W

= average turnover costs per dwelling for year in scenario = average wage costs per dwelling for year in scenario

14

Maintenance costs are incurred to maintain dwellings and keep them in a good state. These costs are assumed to grow with the maintenance index. Turnover costs are incurred when a tenant turnover takes place and is assumed to grow with the maintenance index. Wage costs are the expenses related to management of the dwellings. Examples of these costs are administrative costs, costs associated with handling complaints, and cost for the management of the property. It is assumed that wage costs grow with the wage index. Miscellaneous costs are, for example, associated with insurance and is assumed to grow with inflation. The maintenance, inflation, and wage index will be simulated.

4.3 ASSET VALUE

Unlike other research the value of the assets of Dutch housing associations will not be determined by calculating the net present value (NPV) of the future cash flows generated by the dwellings according to the association’s policy. Instead, the value of the unlevered assets, i.e. the dwellings, will be calculated using the market value in rented state. This market value can be calculated by taking the maximum NPV of two scenarios. The first is the going concern scenario in which the dwellings are assumed to be rented out and remain in possession of a commercial investor indefinitely. In the second scenario, part sale, dwellings are sold when tenants leave (CFV, 2011). In both scenarios a DCF model over a period of 15 years is being used. Due to the changing economic environment and the dynamic investment policy, the input values to determine the market value will change having the result that the market value in rented state will be calculated for each scenario for each year. Determining the market value for each scenario and for each simulation year, results in a simulation model within the main simulation model, increasing the calculation time considerably. The DCF model to calculate the asset value as the maximum NPV of both scenarios for each year in each economic scenario can be shown as the median of:

[ ] [( ∑ ( ) ) ( ∑ ( ) ) ] (4.7)

W

= net cash flow at year in scenario

= discount rate

15

Barthélémy and Makrane, 2006). To provide a more robust valuation and incorporate uncertainty in the valuation I will simulate the input variables using 100 scenarios which are, at each different scenario in each year, randomly picked from a set of 2000 possible future scenarios. Section 6 describes how the different economic and financial scenarios are generated.

4.3.1 MARKET CASH FLOWS

The cash flows used to calculate the market value in rented state are very similar to those described in Section 4.2. The main difference between the cash flows in the going concern scenario and association cash flows is that in the going concern scenario I assume that as a tenant leaves, the commercial investor will always ask the market rent, thus the commercial investor is not bothered with the social objective to provide for affordable housing. Following CFV (2011), I assume the market rent to equal 5% of the vacant value of a dwelling. The main difference between the cash flows in the part sale scenario and association cash flows is that in the part sale scenario I assume that as a tenant leaves, the commercial investor will sell the dwelling at the vacant value, which is assumed to equal the WOZ3-value of the dwelling. Moreover, commercial investors tend to be more efficient on costs. To reflect this efficiency the costs are based on the Dutch real estate operating costs standards (VEX standards) (CFV, 2011).

The final cash flow in year T is composed of two parts. The first part consists of the cash flows generated by rental income and costs. The second part of the final cash flow is the revenues from a fictitious sale. Combining these parts results in the net cash flow at year T:

( ) ( ) (4.8)

Where: = terminal value at year in scenario

In this thesis I will use the exit yield method to determine the terminal value following Baroni, Barthélémy, and Makran (2007) and Hoesli, Jani, and Bender (2006). The terminal value is determined by calculating an exit yield, discount factor minus a growth rate, and dividing the final net operating cash flow by this exit yield. To account for aging of the dwellings I incorporate renovation costs in the net operating cash flow. A drawback of the exit yield method is that the terminal value is highly dependent on the net cash flow in the final year. The terminal value can be displayed as: ( ) (4.9) 3

16

= discount rate

= growth factor

In this research the growth rate is set equal to the expected long-term inflation. The underlying assumption is that rent increases equal inflation and the economic parameters relating to the operating costs are assumed to grow towards expected long-term inflation. The expected long-term inflation is assumed to be equal to the target inflation of the European Central Bank (ECB), which is set at 2% (ECB, 2013).

4.3.2 DISCOUNT RATE

There has not been any conclusive research on the value of a discount rate for unlisted unlevered Dutch real estate. As I need to calculate the net present value of unlevered assets I will derive a discount rate from the market in line with Kramer (2013b). The discount rate is calculated using Dutch listed real estate as a proxy. The expected geometric return on listed levered real estate will be calculated based on the CAPM beta, which is estimated on rolling 12-month returns over the past ten years. To be able to determine the expected unlevered return, the expected return for listed real estate needs to be corrected for leverage. Following Pagliari, Scherer and Monopoli (2005) and Kramer (2013b), I derive expected return on non-listed and unlevered real estate by using:

[ ] [ ] ( ) [ ] (4.10)

Where: [ ] = expected return on unlisted unlevered real estate

[ ] = expected cost of debt

[ ] = expected return on listed levered real estate = loan to value

According to Pagliari, Scherer and Monopoli (2005), listed real estate companies have a LTV of 40% to 50% on average. For the cost of debt I follow Kramer (2013b) by assuming an AA rating with a LTV of 40% and an A rating with a LTV of 50%. Furthermore, as the expected return on unlisted real estate will be used as a discount rate I will need to use the simple arithmetic return (Cooper, 1996). To convert the geometric return to arithmetic return, under the assumption that the returns follow a random walk, the following formula will be used:

[ ] [ ] (4.11) Where: = standard deviation associated with the return on unlisted

real estate.

17

[ ] = expected geometric return on unlisted real estate

4.4 INVESTMENT POLICY

Investment cash flows associated with a housing association’s investment policy typically take the form of the selling proceeds, renovation costs, construction costs and demolition costs of dwellings. Demolition costs are ignored in this research as they only represent a small fraction of the investment cash flows. Renovation of a dwelling normally takes place after a renovation cycle. For simplicity, I use average renovation costs per dwelling each year by dividing the total renovation costs per dwelling with the renovation cycle. The underlying assumption is that the portfolio of dwellings is assumed to be, in real terms, constructed equally over time and that dwellings are renovated on a linear basis. Moreover, costs associated with renovation are tax deductible. Renovation costs are assumed to grow with the construction input index. The next paragraph elaborates on the policy regarding the building and selling of dwellings.

Different target debt ratios are set at the start of the simulation. During the simulation the actual debt ratio can deviate from the target ratio due to changing asset value or changing debt value resulting from financing needs. Next to the target debt ratio, following Fischer, Heinkel and Zechner (1989), a range is given within which the debt ratio is expected to remain. This limit is given by an upper bound and a lower bound . If the actual debt ratio exceeds debt will have to be repaid, a housing association cannot issue equity to alter its capital structure. To be able to generate the cash flows needed to redeem debt the investment policy will be altered following a priority list. The first option is to use internal operating cash flows and foregoing investments in new dwellings, i.e. building dwellings. If the first option provides to be insufficient, like Hennessy and Whited (2005), capital will be sold at a discount. This sale of assets can be viewed as a form of financial distress.

18

Where: = net new production in year in scenario

= debt ratio in year in scenario

= net cash flow in year in scenario

= debt redemption long-term debt at year in scenario = newly attracted long-term debt in year in scenario = construction cost new dwelling in year in scenario = discounted sale value of dwelling in year in scenario

Additionally, is limited to the amount of debt maturing at year in scenario . Furthermore,

construction costs and the sale value of a dwelling are assumed to grow with the construction input cost index and house price index respectively.

4.5 LIABILITIES

Liabilities in this model consist entirely of debt. As input I assume that the debt portfolio consists of short-term one year and long-term ten year secured AAA rated corporate bonds. AAA rated bonds are used because loans obtained by the housing associations are backed by the Dutch Social Housing Guarantee Fund (WSW) which is AAA rated (Conijn, 2011; Moody’s, 2012). Due to the backing of the WSW, debt attracted by the housing associations can be viewed as risk-free. The WSW imposes a restriction on the use of short-term debt, debt with a maturity shorter than two year, as short-term financing involves a different risk profile compared to long-term financing. Repayment of short term debt involves a higher liquidity risk which negatively impacts the liquidity for long-term financing. To control for this risk the WSW limits the use of short-term debt to a maximum of 7.5% of the value of the assets.

19

nine years ago, 10% with a remaining maturity of two year with a ten year interest rate stated eight years ago etc. Moreover, I assume that the long-term debt will be rolled over when it matures.

The amount of long-term debt being attracted each year varies with the financing need and the beginning of year actual debt ratio. If the debt ratio exceeds the threshold debt will be redeemed with cash flows from the changed investment policy. As in Kane, Marcus and McDonald (1985), debt redemption is maximized to the level of debt maturing in the specific year. If the debt ratio still exceeds the threshold more debt will be redeemed, again to a maximum of debt maturing, in the next year. Debt will be redeemed until the debt ratio is below the threshold . If the debt ratio is below the threshold debt will be attracted and invested. The amount of debt attracted equals the amount necessary to regain a debt ratio within the boundaries set.

The interest rate on debt will be calculated using the spread between euro AAA corporate bonds versus euro AAA sovereign bonds. The risk premium Dutch housing corporations pay over their loans differs from the market premium. This difference exists because loans obtained by the housing associations are backed by the WSW (Conijn, 2011). Due to this assurance I assume that the loans entered by the associations are secured AAA corporate bonds. The backing of the WSW only applies to loans used for social housing. As I assume that all dwellings are social housing I can assume loans to be backed by the WSW. Interest rates on one and ten year secured AAA corporate bonds will be simulated. Interest cost for year in scenario can be displayed as:

(

∑

) ( )

Where: = interest rate for short-term debt at year in scenario = interest rate for long-term debt at year in scenario

= short-term debt outstanding in year in scenario

= long-term debt outstanding with maturity in scenario

4.6 AGENCY COSTS

20

to watch the firm’s decisions. This reduced supervision can cause a possible moral hazard problem as managers are tempted to take excessive risks or start empire building (Hart, 1993).

An example of an agency cost is, for example, the use of derivatives. While using derivatives to hedge a housing association’s interest exposure can be a good strategy, the mismanagement of derivatives can have severe effects. A Dutch housing association Vestia, for example, used derivatives for hedging purposes but neglected the accompanying liquidity risks. This mismanagement led to a total loss of around 700 million euros. Furthermore, due to less market supervision, managers of housing associations could be tempted to destroy value by entering into high profile prestigious investments. For instance, the Dutch housing association Woonbron acquired an old cruise ship which they wanted to renovate and use for housing purposes. The redevelopment costs of the ship were estimated at around 6 million euros but, due to mismanagement, amounted to a total cost of over 200 million euros (Kramer, Kronbichler, and Van Welie, 2011). However, while agency cost can have a serious impact on capital structure decisions, to reduce complexity in my simulation model I ignore possible agency problems. Jensen and Meckling (1976), Jensen (1986), and Hart (1993) elaborate on the subject of agency problems accompanying optimal capital structure decisions.

4.7 SENSITIVITY ANALYSES

To measure the impact of changing the range from which the actual debt ratio may deviate from the target debt ratio, I will repeat the simulation with different boundaries for the thresholds and . The deviation range from the target debt ratio is set at 5%-points and 10%-points respectively. An increase in the deviation range results in the association being able to postpone a change in its investment policy.

5.

R

To find inefficient financing policies an analysis in the risk-return field is necessary. A policy is inefficient when higher returns can be realized at the same or lower risk, or when the risk can be lowered without reducing returns. The right choice of risk and return measurements is crucial as these are used to compare different financing strategies (Kramer and Van Welie, 2001).

5.1 RETURN MEASUREMENT

21

return (IRR). The IRR does incorporate the time value of money. The downside with the IRR is that it is unreliable if cash flows are fluctuating between positive and negative values (Hillier, Grinblatt, and Titman, 2002; 345). The TR is preferred over the IRR measurement as cash flows of housing associations will fluctuate over time. The TR on assets can be displayed as:

[ ]

(5.1)

Where: = value of assets at end of year in scenario

= starting value assets

The TR on equity can be presented as:

[ ] (5.2) Where: (5.3)

And: = value of equity at end of year in scenario = value of debt at end of year in scenario

= new loans entered in year in scenario = loan redemption in year in scenario

= starting value liabilities

The TR on assets is applied because when the association’s equity value increases, more debt will be attracted and invested in building additional dwellings. The costs associated with building a dwelling are usually higher than the market value in rented state of the newly build dwelling, thus destroying equity value. Looking at the aim of housing associations, providing housing, it is sensible to look at the TR on assets.

5.2 RISK MEASUREMENT

22

(5.4)

Where: = Returns generated by association

= expected long-term inflation = Standard deviation of return

A drawback of using the Sharpe ratio is that it penalizes both positive and negative deviations from the median return generated. According to Sortino and Van der Meer (1991), standard deviation captures the risk associated with achieving the mean return, but that standard deviation can be unrelated to bad outcomes. Sortino and Van der Meer argue that a distinction must be made between good and bad volatility. Where bad volatility is below the minimum acceptable return (MAR) and good volatility is the dispersion above the MAR. The MAR is a pre-determined reference point used to compare returns achieved. As Dutch housing associations do not have shareholders and do not have a required rate of return, I will set the MAR equal to the expected long term inflation. The Sortino ratio incorporates downside risk and utilizes the MAR to give a measure of risk. For the analysis in the risk-return field both the Sharpe and the Sortino ratio will be used. The Sortino ratio can be depicted as:

(5.5) Where: [ ∑( ) ( ) ] ( ) ( ) (5.6)

And: = Downside risk

= total number of scenarios

23

The ICR displays an association’s ability to earn its interest costs with operational cash flows. Furthermore, it is used to determine to which extent the operational cash flows can decrease without bringing the association in financial distress. The equally weighted DSCR can be displayed as:

̅̅̅̅̅̅̅̅ _∑

Where: = total amount of debt outstanding at year in scenario

= net interest costs at year in scenario

The WSW requires housing associations to incorporate fixed debt redemption within the calculation of the DSCR. This debt redemption, which is set at 2% of total debt, represents a housing association’s ability to repay its debt within the coming 50 years. Moreover, the DCSR helps to depict if there are sufficient operational cash flows available to meet interest and loan redemption requirements. A ratio equal to one indicates that the operational cash flows are exactly offset by the interest costs and loan redemption. The WSW requires a lower ratio compared to commercial investors. Commercial capital providers normally require an extra margin on the DSCR to be able to offset potential setbacks.

6.

D

The next section describes the economic input parameters used to simulate the economic environment and the input values used as a starting point in the simulation.

6.1 SIMULATION PARAMETERS

24

Table 1: Economic parameters used as input in the simulation - Consumer price index (inflation) - House price index

- Wage index - 3 month Corporate AAA Euro bond interest rates - Construction input cost index - 1 year Corporate AAA Euro bond interest rates - Construction output cost index - 10 year Corporate AAA Euro bond interest rates

This table shows the economic parameters which are used as input in the Monte Carlo simulation model to determine the expected future cash flows and asset value. These parameters are simulated by the Real World Dynamic Scenario Generator developed by Ortec Finance.

The economic parameters will be simulated by the Real World Dynamic Scenario Generator (DSG) developed by Ortec Finance. The DSG employs a frequency domain methodology to construct time series models that give a joint description of the empirical long and short term behaviour of economic and financial variables. According to Steehouwer (2010), this methodology brings together the empirical behaviour of the economic and financial variables as observed at different horizons and observation frequencies. Furthermore, this methodology gives understanding of the corresponding dynamic behaviour, both in terms of empirical time series data and of the time series models used to describe this behaviour. Using the DSG guarantees that the statistical characteristics of the generated scenarios correspond with the statistical characteristics observed in the past (Steehouwer, 2010).

6.2 MODEL INPUT

Input values are based on an average Dutch housing association. The CFV publishes data on each association annually (CFV, 2012). To calculate the input values a weighted average is calculated based on the number of dwellings owned by an association. The input values used in the simulation are depicted in Appendix B. The next paragraphs present how the turnover rate is determined and the results from the calculation of the discount rate.

Data for Dutch turnover rates are scarce. Turnover rates for Dutch rental houses owned by housing associations are available from 1999 to 2011 (13 observations) displayed in Figure 1. This lack of data prevents me from estimating a statistically sound probability density function. According to Kramer (2013a) the turnover rate is strongly correlated to the change in house prices. In markets with increasing prices the turnover rate is expected to rise and vice versa. These findings are in line with the research from Deng, Gabriel and Nothaft (2003). As such, I assume that the average turnover rate is dependent on the house price change. According to Kramer the model to predict future turnover rates with only the change in the house price index as explanatory variable is:

(6.1)

Where: = turnover rate

25

The results from the calculation to determine the discount rate are presented in Table 2. The average arithmetic return, displayed in the last column, will be used in the DCF model to calculate the net present value of the assets.

Table 2: geometric and arithmetic returns for Dutch unlisted unlevered real estate

LTV Cost of debt R[llRE] R[ulRE] Arithmetic Return Average Arithmetic Return

40% 4.75% 6.089% 5.553% 5.850%

5.870%

50% 5.10% 6.089% 5.594% 5.891%

This table shows the results from the calculation of the discount rate. The discount rate is determined by unlevering the geometric expected return derived from Dutch listed real estate.

7.

R

The next section displays the results from the Monte Carlo simulation model. The following paragraphs discuss the existence of efficient financing policies and the relating target debt ratios. Moreover, the probabilities of not meeting the restrictions imposed by the CFV and WSW are presented and discussed.

7.1 EFFICIENT TARGET DEBT RATIOS

Table 3 depicts the results from the simulation. The results show a clear pattern in the risk and return spectrum for both assets and equity. From Table 3 it becomes apparent that using debt financing increase total returns up to a target debt ratio of 50%. For target debt ratios set higher than 50%, however, average and median annualized total return start to decrease. In addition, per incremental step in the target debt ratio risk starts to increase. Up to a target debt ratio of 40%, the added risk is compensated with extra returns. Higher target debt ratios lead to inefficient capital structures as risk increase and returns start to diminish. The results for target debt ratios of 0% to 60% have been plotted in Figure 2 and Figure 3. These figures give a graphical illustration of the

26

development of the risk and return per target debt ratio. The graphs give a clear illustration of the tipping point from where adding additional debt becomes inefficient.

Table 3: Average and median of total return on assets, total return on equity and relating standard deviation per target debt ratio.

Target debt ratio 0% 10% 20% 30% 40% 50% 60% 70% 80% 90%

5%-point deviation range from target debt ratio

Average TR on Assets 0.0309 0.0320 0.0346 0.0363 0.0374* 0.0373 0.0310 0.0068 -0.1123 -0.2129 Median TR on Assets 0.0297 0.0299 0.0343 0.0351 0.0395 0.0396* 0.0354 0.0334 0.0326 0.0227 Standard deviation 0.0153 0.0155 0.0164 0.0188 0.0201 0.0230 0.0546 0.1075 0.1857 0.3439 Sharpe ratio 0.7124 0.7742 0.8902* 0.8670 0.8657 0.7522 0.2015 -0.1228 -0.7124 -0.6772 Downside risk 0.0022 0.0022 0.0023 0.0024 0.0025 0.0039 0.0677 0.1722 0.3834 0.6288 Sortino ratio 4.9545 5.4545 6.3478 6.7917 6.9600* 4.4359 0.1625 -0.0767 -0.3451 -0.3704 Average TR on Equity 0.0309 0.0328 0.0372 0.0390 0.0399 0.0403* 0.0307 -0.0110 -0.1831 -0.3771 Median TR on Equity 0.0297 0.0317 0.0366 0.0376 0.0393 0.0398 0.0402* 0.0389 0.0385 0.0121 Standard deviation 0.0153 0.0164 0.0177 0.0201 0.0221 0.0267 0.0974 0.1637 0.3402 0.5079 Sharpe ratio 0.7124 0.7805 0.9718* 0.9453 0.9005 0.7603 0.1099 -0.1894 -0.5970 -0.7818 Downside risk 0.0022 0.0022 0.0023 0.0025 0.0026 0.0051 0.0926 0.2163 0.4791 0.6586 Sortino ratio 4.9545 5.8182 7.4783 7.6000 7.6538* 3.9804 0.1156 -0.1433 -0.4239 -0.6029

*Maximum TR, Sharpe, and Sortino ratio.

This table displays the risk and returns on assets and equity resulting from the Monte Carlo simulation with a 5%-point deviation range. Furthermore, the Sharpe and Sortino ratios are shown, which depict the extra returns generated per unit of risk.

27

Looking at the Sharpe ratios, depicted in Table 3, the conclusion can be drawn that target debt ratios up to 20% remain efficient. However, looking at the Sortino ratios, it can be concluded that target debt ratios up to and including 40% are efficient. For target debt ratios above 50% the Sharpe and Sortino ratios decrease sharply, indicating that these target debt ratios become inefficient as the extra risk is compensated with lower returns. Difference between the Sharpe ratios and Sortino ratios exist due to the Sortino ratio making a distinction between good and bad volatility. Table 3 shows that there is low downside risk associated with target debt ratios up to 40%. An explanation for the low downside risk stems from the method used to value the assets.

28

The simulation results with a 10%-point deviation from the target debt ratio are shown in Table 4. Increasing the deviation range should result in an association being able to postpone a change in its investment policy, and possibly change the returns generated. From the results it becomes clear that postponing a change in the investment decision leads to increased returns on equity. However, the pattern in the risk and return spectrum for both assets and equity is similar to that of a 5%-point deviation range. The Graphs displaying the risk and return spectrum for 10%-point deviation range are shown in Appendix C. Furthermore, the Sharpe and Sortino ratios show the same results as the results generated with a 5%-point deviation from the target ratio. Looking at the value of the Sortino ratio it can be seen that increasing the deviation range results in a lower Sortino ratio for the average annualized total return on assets while increasing the Sortino ratio for the average annualized total return on equity.

29

and redemption requirements starts to decrease significantly from a target debt ratio set at 50% and higher.

Table 4: Average and median of total return on assets, total return on equity and relating standard deviation per target debt ratio with a 10%-point deviation range from target ratio.

Target debt ratio 0% 10% 20% 30% 40% 50% 60% 70% 80% 90%

10%-point deviation range from target ratio

Average TR on Assets 0,0309 0,0319 0,0338 0,0346 0,0356 0,0368* 0,0324 0,0245 -0,0549 -0,2126 Median TR on Assets 0,0297 0,0298 0,0323 0,0356 0,0377* 0,0376 0,0360 0,0362 0,0357 0,0221 Standard deviation 0,0153 0,0152 0,0157 0,0174 0,0187 0,0221 0,0514 0,0843 0,1067 0,3201 Sharpe ratio 0,7124 0,7829 0,8790* 0,8391 0,8342 0,7602 0,2412 0,0534 -0,7020 -0,7266 Downside risk 0,0022 0,0023 0,0024 0,0025 0,0026 0,0031 0,0647 0,1120 0,3094 0,4913 Sortino ratio 4,9545 5,1739 5,7500 5,8400 6,0000* 5,4194 0,1917 0,0402 -0,2421 -0,4734 Average TR on Equity 0,0309 0,0331 0,0382 0,0402 0,0416 0,0427* 0,0395 0,0313 -0,0474 -0,1797 Median TR on Equity 0,0297 0,0317 0,0385 0,0388 0,0410 0,0418 0,0428 0,0441* 0,0438 0,0220 Standard deviation 0,0153 0,0164 0,0176 0,0196 0,0212 0,0249 0,0297 0,1180 0,2228 0,4765 Sharpe ratio 0,7124 0,7988 1,0341* 1,0306 1,0189 0,9116 0,6566 0,0958 -0,3025 -0,4191 Downside risk 0,0022 0,0022 0,0022 0,0024 0,0025 0,0033 0,0648 0,1127 0,3100 0,4920 Sortino ratio 4,9545 5,9545 8,2727 8,5957 8,6400* 6,8788 0,3009 0,1003 -0,2174 -0,4059

*Maximum TR, Sharpe, and Sortino ratio.

This table displays the risk and returns on assets and equity resulting from the Monte Carlo simulation with a 10%-point deviation range. Furthermore, the Sharpe and Sortino ratios are shown, which depict the extra returns generated per unit of risk.

The results from the simulation show that the use of debt creates extra returns. The use of debt generates extra returns not only due to the tax deductibility, but also due the positive development of the value of assets with a relative low downside risk. With respect to the returns achieved on assets and equity, adding debt when the debt ratio is above 50%, within both deviation ranges, result in an inefficient capital structure as the relative risk associated with the return starts to increase faster relative to a capital structure with less debt. Furthermore, the Sortino ratios indicate that target debt ratios set up to 40% remain efficient. The presence of this efficient frontier, for all deviation ranges, results in accepting the first hypothesis:

30

Table 5: Probability not meeting requirements stated by the CFV and WSW per target debt ratio.

Target debt ratio 0% 10% 20% 30% 40% 50% 60% 70% 80% 90%

5%-point deviation from target debt ratio

Probability violating ICR 0.000 0.000 0.002 0.010 0.108 0.290 0.910 1.000 1.000 1.000

Probability violating DSCR 0.000 0.000 0.002 0.018 0.028 0.200 0.528 1.000 1.000 1.000

Default probability 0.000 0.000 0.000 0.000 0.000 0.004 0.008 0.044 0.216 0.420

Probability violating a constraint 0.000 0.000 0.002 0.018 0.108 0.292 0.914 1.000 1.000 1.000 10%-point deviation from target debt ratio

Probability violating ICR 0.000 0.000 0.002 0.010 0.050 0.240 0.902 1.000 1.000 1.000

Probability violating DSCR 0.000 0.000 0.002 0.012 0.014 0.144 0.440 1.000 1.000 1.000

Default probability 0.000 0.000 0.000 0.000 0.000 0.000 0.004 0.012 0.092 0.232

Probability violating a constraint 0.000 0.000 0.002 0.014 0.052 0.240 0.906 1.000 1.000 1.000 This table depicts the probability of not meeting the constraints set by the CFV and WSW per target debt ratio and per deviation range.

7.2 RESTRICTIONS CFV AND WSW

Table 5 depicts the probabilities of not meeting the requirements set by the CFV and WSW. The probabilities of violating the constraints set by the CFV and WSW increase significantly from a target debt ratio above 40% with a deviation range of 5%-points, and from a target debt ratio above 50% with a deviation range 10%-points. With a target debt ratio set at 50% an association has got a 29.2%, and 24% probability of not meeting a constraint with a deviation range of 5% and 10%-points respectively.

The constraints imposed by the CFV and WSW are set such that housing associations are able to set target debt ratios that are efficient. The risk of violating the restrictions is significant when using an inefficient target debt ratio set at 50%. While target debt ratios between the 0% and 40% range are efficient, the target ratio an association is willing to use depends on the association’s risk preferences. As an efficient capital structure is within the constraints imposed by the CFV and WSW this result in accepting the second hypothesis:

“An efficient capital structure is within the borrowing constraints imposed by the CFV and WSW”.

8.

C

31

8.1 RESEARCH QUESTION

Several articles studied the advantages of using debt financing for real estate investments. This study is aimed at studying the advantages of debt financing by analysing the risk and return associated with various target debt ratios for non-listed companies invested in real estate, more explicitly Dutch housing associations. The risk and return spectrum is analysed by using a simulation model, similar to the models developed by Van der Spek and Hoorenman (2011) and Titman and Tsyplakov (2007), with an endogenous financing and investment policy. As Dutch housing association are non-listed companies they cannot use the equity market to get additional financing or to alter their capital structure. Financing gaps need to be filled by either retained earnings, the sale of dwellings, or attracting more debt. The simulation model used in this thesis employs a Monte Carlo simulation to simulate different future states of an average Dutch housing association with different target debt ratios and deviation ranges set for the next 15 years. Furthermore, the simulation model values the assets owned by the housing association in each scenario in each year as if they are owned by a rational commercial investor, and as such simulates and discounts the expected future cash flows to get the net present value of the assets owned.

Findings from previous research suggest that efficient target debt ratios exist (Cannaday and Yang, 1996; Boyd, Ziobrowski, Ziobrowski and Cheng, 1998; McDonald, 1999). As the simulation model illustrates, in accordance with studies in real estate investment, Dutch housing associations have advantages resulting from using debt financing. The simulation results show that the use of debt can result in higher returns. These advantages arise due to the tax deductibility of interest expenses and the value development of assets. The valuation of dwellings, by using the market value in rented state, results in an option like structure that reduces the downside risk of the returns. The lower downside risk and higher potential upside on the value of the assets support the findings of McDonald (1999), who states that the use of debt adds value if investors pay taxes and have positive expectations about the real estate investment. Furthermore, while interest rates do not increase with the use of extra debt, diminishing returns are generated due to financial distress.

32

It should be taken into account that risk increase disproportionally for every incremental step in target debt ratios over 50% due to costs associated with financial distress. Returns start to decrease and risk starts to increase sharply for target debt ratios set over 50%. Furthermore, target debt ratios set over 40% result in an increased probability of not meeting the constraints set by the CFV and WSW. A target debt ratio of 40% with a deviation range of 5%-points and 10%-points is accompanied with a probability of 10.8% and 5.2%, respectively, of not meeting the restrictions. The probability of violating constraints increases significantly for higher target debt ratios.

Like Van der Spek and Hoorenman (2011), the use of debt financing can result in higher returns, but too much debt can be devastating. Concluding, this study shows that Dutch housing associations can set efficient target debt ratios within the constraints set by the CFV and WSW. The efficient capital structure decision for Dutch housing associations is driven by a trade-off between tax benefits, leverage effect due to the relatively stable positive development of asset value, and potential distress cost.

8.2 FURTHER RESEARCH

Further research could try to find if efficient capital structures exist for commercial non-listed companies which are not, partly, supported by governmental institutions. In this model I assume that loans entered by the association are backed by the WSW. Dropping this assumption would lead to a higher interest rate if the association’s leverage increases. According to Tyrell and Bostwick (2005), an increase in the use of debt results in higher interest rates leading diminishing returns and can, possibly, affect the efficient capital structure. Additionally, different types of loans, for example with different maturities or seniority, can be used to create different and more efficient debt portfolios.

Furthermore, a sensitivity analysis can be performed on the dependency of the efficient target debt ratios for different types of housing associations, for example associations with differences in the age build-up within their portfolio of dwellings, or differences in regions like growth regions versus contraction regions. Moreover, to increase the accuracy of the model a simulation could be performed using micro simulation. Micro simulation entails simulating future states of each separate individual dwelling and loan entered (Kramer and Van Welie, 2001).

33

34

R

Baroni, M., Barthélémy F., Mokrane M., 2006. Monte Carlo simulations versus DCF in real estate portfolio valuation. ESSEC Working Papers. ESSEC Business School.

Baroni, M., Barthélémy F., Mokrane M., 2007. Using rents and price dynamics in real estate portfolio valuation. Journal of property management 25(5), 462-486.

Binsbergen, J.H van, Graham J.R., Yang, J., 2010. The cost of debt. Journal of Finance 65(6), 2089-2136. Blok, S., 2013. Brief aan Tweede Kamer. Tweede Kamer, Nederland.

Boelhouwer, P.J., 2007. The future of Dutch housing associations. Journal of Housing and the Built Environment 22(4), 383-391.

Boyd J.W., Ziobrowski, A.J., Ziobrowski, B.J., Cheng, P., 1998. Leverage and real estate investment in mixed-asset portfolios. Journal of real estate portfolio management 4(2), 135-147.

Bradley, M., Jarrel G.A., Kim E.H., 1984. On the existence of an optimal capital structure: Theory and evidence. The Journal of Finance 39(3), 857-878.

Brennan, M., Schwartz E., 1984. Optimal financial policy and firm valuation. Journal of Finance 39(3), 593-607. Brooks, C., 2011. Introductory Econometrics for Finance, Second edition. Cambridge University Press:

Cambridge.

Bunn, D.W., Salo A.A., 1993. Forecasting with scenarios. European Journal of Operational Research 68(3), 291-303.

Cannaday, R.E., Yang T.T., 1996. Optimal leverage strategy: capital structure in real estate investments. Journal of Real Estate Finance & Economics 13(3), 263-271.

Casey, C., 2001. Corporate valuation, capital structure and risk management: A stochastic DCF approach. European journal of operational research 135(2), 311-325.

CFV, 2011. Handreiking voor het toepassen van RJ 645. Centraal fonds Volkshuisvesting, Baarn. Available at: http://www.cfv.nl/media_dirs/8446/media_files_data/handreiking-rj645-31-januari-2012.pdf

CFV, 2012. Sectorbeeld realisaties woningcorporaties verslagjaar 2011. Centraal fonds Volkshuisvesting, Baarn. Available at:

http://www.cfv.nl/media_dirs/8446/media_files_data/sectorbeeld_realisaties_website.pdf