Transcript
Valuation accuracy in vacant office properties A comparison between appraised cap rates and transaction cap rates
SANAZ MIRZAEI
%
Titel research project: Student: Address: Phone: E-mail address: Date proposal: MSc Laboratory: First mentor: Second mentor: External Examinator:
Valuation accuracy in vacant office properties Sanaz Mirzaei [4317882] Buitenhoflaan 17, 2353 MG Leiderdorp 0643418510
[email protected] July 2, 2015 Real Estate Management Drs. Philip Koppels Dr. Ir. Hilde Remøy Dirk Dubbeling (OTB)
ii
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
Preface This master thesis discusses the valuation accuracy in vacant office properties in the Netherlands. It is written for the Real Estate Management graduation lab at the Department of Real Estate and Housing, within the Faculty of Architecture at the TU/Delft. It covers the final official assessment session (P5) of the total of five official assessment sessions for the graduation process started in September 2014. The research proposal is built upon the current problem of the Dutch office market, namely oversupply and the bottlenecks of converting the obsolete office stock. Even though the transformation of surplus stock is generally accepted by various market parties ( e.g. government, investors, redevelopers, financier, academic researchers, etc.), there is no statistically significant evidence pertaining the valuation accuracy of vacant office stock. Therefore, this research aims at providing such indications with a hope to facilitate the adaptation process of the obsolete office stock and overcome the oversupply of the Dutch office market. The result of this research can be found in the rest of this thesis. I thank the following people for their help: my first mentor Philip Koppels and second mentor Hilde Remøy for their kind support and constructive feedback, Aart Hordijk for his valuable input and lastly my mentors at the Amsterdam municipal tax office, Sander Sijm and Thijs Booij. Without their valuable input this thesis would be completely different. Sanaz Mirzaei July 2015
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
iii
Executive summary INTRODUCTION The Dutch office market is experiencing fundamental changes. On one hand, the low dynamics of the office market caused by the financial and real estate crises and the changes in the behavior of office users, reduced the required office space in this market (in 2014 7% lower take-up than 2013). On the other hand, continuous overproduction and oversupply of offices caused by the accelerating expansion of this market from the 1980’s and onwards, resulted in a huge gap between demand and supply in the office market. This all off balanced the Dutch office market that much, that this market never recovered since the 2001 crisis in real estate, even before the financial crisis in 2008 started. As a result there is a relatively high average national vacancy rate of 15.7% which equals 7.8 million square meter of the total office stock. Adaptive re-use strategies as one of the possible strategies for dealing with vacant offices can be seen (partly) as a solution to balance the oversupply of the office market. Adaptive re-use strategies consist of either renovation, within use adaptation of obsolete office buildings, or conversion to new use. PROBLEM STATEMENT Even though the adaptive re-use strategies are mentioned as a possible solution to balance the oversupply of the Dutch office market and decrease the current structural vacancy, they are often obstructed by a lack of financial feasibilities. Many parties (e.g. investors, redevelopers, etc.) believe that valuations of vacant offices are too high in comparison to their actual market value. Previous research has indicated that difficulty and uncertainty in estimating the value of vacant properties has a negative influence on the valuation accuracy. Since a vacant property generates no rental income and may have a relatively low promise for future tenancy, assumptions in appraising this property are more uncertain and therefore prone to some degree of inaccuracy in valuation. Besides limiting the adaptive re-use strategies, other consequences of inaccuracy in valuation are its negative influence on: the solvency rate and judgment about the financial position of property owners, on property cash flow after tax due to the inaccuracy in estimated assessed value, and lastly, on recording a lower volatility in appraisal indices compared to the real estate value (transaction price) which results in not recording the true volatility of the Dutch office market. Therefore, studying the accuracy of appraised market value in comparison to the actual market value (transaction price) of vacant offices is of importance of the Dutch real estate market, Dutch financial institutions as well as the Dutch society. RESEARCH GOAL This thesis focuses on this specific problem by determining the cap rate components as one of the important input variables for determining the property value. This is to understand the differences and similarities in pricing processes of two main market players, namely appraisers and investors to enhance accuracy and consistency in office property valuation in the Netherlands. Subsequently, it aims to increase awareness about valuation accuracy for vacant office properties among various market players (e.g. investors, appraisers, developers, etc.) by studying whether any variation in accuracy depends upon the variation in vacancy rate. Therefore, the main research questions of this study are:
iv
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
1. To what extent do appraised cap rates correspond with transaction cap rates? 2. Can the differences between appraised cap rates and transaction cap rates be explained (partly) by structural vacancy risk? Whereas the main research hypothesis to be tested are as follows: Hypothesis 1 The appraised cap rates of Dutch offices during the rising market are lower (appraised value is higher) than the transaction cap rates (prices), to a larger extent than the declining market (after financial crisis). Hypothesis 2 Appraisers put more weight on easily observable elements in determining office cap rates whereas investors are more concerned with macro-economic variations over time. Hypothesis 3 The differences between appraised cap rates and transaction cap rates of an office property have a positive correlation to its vacancy risks. Theoretical framework Literature on cap rate determinants show a trend from the dominancy of macroeconomic and time-series variations in cap rates, to the development of those where micro-level and propertyspecific factors become more prominent in explaining the driving forces of cap rates. Due to an increase of the investors’ awareness and interest in the quality and characteristics of properties within investment funds, and also the development of ‘Big Data’ and availability of large volumes of information (structured and/or unstructured) regarding property characteristics, this thesis introduces a new dynamic and applicable model specification to determine cap rates as follows:
= + ∗ − + 1 − ∗ + + + − Where; = risk free rate, ⁄ = Loan-to-value ratio, = rate of return on debt, = premium from participation in real estate,
= premium on location attributes,
= premium on property-itself attributes, = premium on property-user attributes,
= a constant expected rate of growth in the NOI.
This model decompounds the risk premium component achieved on investing in the office market, to location, the property-itself and the user (office-user) attributes. Furthermore, it develops these = components by categorizing them into background context, location, property, and office-users determinants of cap rates in a new cap rate regression model specification: Where;
(1)
V = property value, NOI = Net Operating Income in the first year, Ro= overall capitalization rate (%).
=
(2)
= premium on property-itself attributes, = premium on property-user attributes,
i =o anconstant rate Va l u at a c c u r expected a c y i n va c aofngrowth t o f f iin c ethe p rNOI. operties
v
= + ∑ + ∑ + ∑ + ∑ + Where; α = constant or intercept term, = the coefficient of variables indicating context, location, property and office users factors, ε = the error term.
The background context, addresses the overall conditions/circumstances at the time that 1 = (19) investments take place. It consists of variables related to the macro-economic conditions, capital market expectations, office market trends, and sale conditions. This category focuses on elements Where; capover rate (%), = Appraised that mostly change time (time-series). The second category, location, considers the property = net income multiplier. location at both macro- and micro-level and examines both cross-sectional and time-series changes of cap rates. The third category addresses variables pertaining the property itself (property-specific = (20) while the last category focuses on elements that are associated with the office-users characteristics) (tenants). Figure i shows the interrelation among these four categories. Where; = overall capitalization context rate (%), = property value (gross transaction price), = Net Operating Income in the first year.
context
location
property
office-users
Figure i, interrelation among four categories of background context, location, property, and lastly office-users.
Empirical research Data The data for this research is gathered by combining two databases from the Amsterdam Municipal Tax office (in Dutch Gemeentebelastingen Amsterdam), and the TU/Delft property database. In addition to these databases, a complimentary database is created from different sources including: the Central Bureau of Statistics in the Netherlands, De Nederlandsche Bank, and a couple of real estate agencies such as Jones Lang LaSalle, Cushman & Wakefield, and DTZ Zadelhoff. The final database consists of 124 office transactions in Amsterdam from 2004 till 2011. Hypothesis 1 | Valuation accuracy appraised versus transaction cap rates To test this hypothesis, the appraised cap rates and transaction cap rates are compared per year before and after the financial crisis. The transaction cap rates are calculated ex-post, whereas the appraised cap rates are present ex-ante in the Amsterdam Municipal Tax office database. In order to measure the differences between appraised cap rates and transaction cap rates, a total variance
(18)
vi
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
test is used. To gauge the accuracy, both average absolute differences and average directional differences between transaction and appraised cap rates are measured. The absolute differences is an appropriate indicator to see how different, on average, the typical transaction cap rate is from the appraised cap rate. Whereas, the directional differences provides an indication whether there is any tendency for appraised cap rates to consistently understate or overstate the transaction cap rates or whether the errors are randomly spread around zero. Hypothesis 2 | Cap rates determinants To test this hypothesis, the determinants of the appraised cap rates and transaction cap rates are compared. The multiple regression model is used to analyze the cap rate determinants for both appraised and transaction cap rates, using the available variables pertaining the context, location and property elements. Office-users elements could not be entered to the models due to the lack of availability of adequate data. The regression is done once by using appraised cap rates and the other time transaction cap rates as a dependent variable using a hierarchical method (blockwise entry) of entering the independent variables in the regression model, based on their theoretical importance, and based on the context (or time-series), location and property categories. Hypothesis 3 | The myth of vacancy and valuation accuracy To test this hypothesis, the differences between appraised and transaction cap rates are compared with different types of vacancy risk: market conformed vacancy risk, potential structural vacancy risk, and structural vacancy risk. The vacancy risk is calculated as based on the vacancy ratio relative to its direct surroundings. A one-way ANOVA test is used to examine the relation between valuation accuracy and the different types of vacancy risk. Empirical findings Hypothesis 1 | Overstated and smoothed appraised cap rates The results indicate consistently overstated and smoothed appraised cap rates in the Amsterdam office market from 2004 till 2011, compared to the transaction cap rates. The transaction cap rates are on average lower and more volatile compared to appraised cap rates. This indicates that appraised cap rates (WOZ-values), are understating the property values compared to their actual market value (transaction prices). An average deviation of 50% is the result of the differences between appraised cap rates and the transaction cap rates, which corresponds with the same deviation between appraisal value and transaction prices. Figure ii shows the cap rate development transaction, appraised (both WOZ and aggregated cap rates in market reports), from 2004 till 2011 in Amsterdam office market. Figure iii indicates the development of the cap rates (both transaction and WOZ cap rate) in different years in Amsterdam, which clearly emphasizes on the higher volatility and sensitivity of transaction cap rates to different submarkets in Amsterdam, and smoothing in appraised cap rates. Hypothesis 2 | Forward looking investors versus backward looking appraisers The findings deduced from testing this hypothesis show that investors in the Amsterdam office market are more concerned with the cross-sectional and property specific variations in cap rates, whereas the appraiser are more concerned with the time-series variations in cap rates. Figure iv compares the explanatory power of context, location, and property related variables, for transaction
vii
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
Transaction cap rate 0,12
Appraised (WOZ) cap rate Appraised (market) cap rate
0,10
Error bars: 95% CI
0,08
Mean
Figure ii, cap rate development in Amsterdam from 2004 to 2011.
0,06 0,04
Figure iii, cap rate development (on the top transaction cap rates, at the bottom appraised cap rates) in Amsterdam districts from 2004 to 2011
Center New-West North East West WestPoort South Amsterdam districts Amsterdam Submarkets Southeast Center Center New-West New-West North North East East WestWest WestPoort WestPoort South South Southeast Southeast Center New-West North East West WestPoort South Southeast
0,02 0,00
2004
2005
2006
2007
2008
2009
2010
2011
SaleYear
Amsterdam Submarket
0,20
Center New-West North East West WestPoort South Southeast Center New-West North East West WestPoort South Southeast
0,18 0,16 0,14
Mean TP_CapRate
Amsterdam Submarkets
0,12 0,10 0,08 0,06 0,04 0,02 0,00 2004
2005
2006
2007
2008
2009
2010
2011
YearOfSale
2010
2011
Amsterdam Submarkets
0,20
Center New-West North East West WestPoort South Southeast Center New-West North East West WestPoort South Southeast
0,18
2010
2011
Mean WOZ_CapRate
0,16 0,14 0,12 0,10 0,08 0,06 0,04 0,02 0,00 2004
2005
2006
2007
2008
2009
2010
2011
YearOfSale
and appraised cap rates. Whereas Figure v shows the differences of the explanatory power of timeseries (year dummies), location, and property related variables, for transaction and appraised cap rates. Hypothesis 3 | Structural vacancy as a Paradoxical phenomenon The result of the test shows a paradoxical situation when there is structural vacancy risk, both appraisers and investors weighted the vacancy risk in a more similar way when there is a market conformed vacancy ratio. In this case, the appraisers put less emphasis on the vacancy and the differences between appraised cap rates and those of transactions become greater. However, the reliability of the statistical results is very low and not statistically significant, due to the unbalanced
viii
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
Hypothesis 2 | Forward looking investors vs. Hypothesis backward looking 2 | Forward appraisers looking investors vs. backward looking
RTP
RV
RTP
step 1
2%
<
11%
step 1
step 2
30%
>
23%
step 3
19%
>
6%
R2
51%
40%
Figure iv, explanatory power of context, location and property specific variables respectively in transaction and appraised cap rates regression
RV
9%
<
18%
step 2
30%
>
19%
step 3
15%
>
5%
Year dummies
R2
54%
40%
20
Figure v, explanatory power of year dummies, location and property specific variables respectively in transaction and appraised cap rates regression
distribution of the samples in each vacancy risk category. In the total sample data, less than 5% of the transactions contain properties with a potential structural vacancy risk or structural vacancy risk. Conclusion and further research The research findings indicate transaction-valuation differences in cap rates as well as contributing factors to these differences. The transaction cap rates are determined by mostly the location and property related variables, whereas macroeconomic and context variables determine the appraised cap rates vastly in the Amsterdam Office market. The low amount of transactions of vacant office properties indicate that selling an office building, being structurally vacant, seems to be an exceptional situation in the Amsterdam office market. This since there are not many of these type of offices, being sold in Amsterdam. One can interpret this as a possibility of asking a higher value than an investor is willing to pay. Unfortunately, no concrete conclusion can be made due to the lack of market evidence and available data to further investigate on this problem. Further research can be done by using appraised values for the commercial purposes (sale or finance), rather than WOZ values which are appraised for tax purposes in order to improve the degree of comparability between transaction and appraised cap rates. In addition, using transaction cap rates, which are not calculated ex-post, one can reduce the number of assumptions to estimate the transaction cap rates. This can lead to better results of the research. Furthermore, the research findings can be improved in case of multi-tenant and multi-owner properties that are also included in the database in addition to the properties with a single tenant and a single owner that this thesis uses. The degree of comparability between transaction and appraised cap rates can be enhanced when including the transaction of other cities in the Netherlands. This gives opportunity to compare cross-sectional differences among various office markets (e.g., Rotterdam, Utrecht, etc.). Ultimately, these all enhance the chance of analyzing properties with different vacancy rates (including those with potential structural and structural vacancy levels.
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
ix
Table of contents PrefaceII Executive summaryIII
Chapter 1: Introduction
1
Part I: Research Proposal CHAPTER 2: A glance at the Dutch Office market
2 4
1.1 Target audience 1.2 RESEARCH OUTLINE
2.1 Dynamics of DUTCH OFFICE MARKET 2.2 Market disequilibrium 2.3 Replacement market 2.4 Adaptive re-use
Chapter 3: Problem field
1 1
4 5 6 7
10
3.1 Lack of financial feasibility - lack of willingness 3.2 Book Value versus Market Value 3.3 commercial and fiscal accounting perspectives 3.4 book values and type of investors
3.5 Valuation accuracy: APPRAISED MARKET VALUE VERSUS ACTUAL MARKET VALUE 14 3.6 Direct capitalization and its fundamental input variables 15 3.7 Problem Statement 16 3.8 research Goal 17 3.9 Research Questions 18 3.10 Research hypotheses 19 3.11 Research scope 20 3.12 RESEARCH APPROACH 20 3.13 RESEARCH DESIGN 20 3.14 QUALITATIVE RESEARCH METHODS 20 3.15 QUANTITATIVE RESEARCH METHODS 21 3.16 Scientific and societal relevance 22
3.4.1 Institutional investors & book value 3.4.2 Private investors & book value 3.4.3 Corporate real estate & book value
3.16.1 Scientific relevance 3.16.2 Societal relevance
10 10 11 12 12 12 13
22 23
Part II: Theoretical framework 24 Chapter 4: Determinants of cap rate, a background study 26
4.1 Cap rate Determinants in international literature 4.2 cap rate Determinants in Dutch literature 4.3 Cap rate determinants conclusion
26 32 33
x
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
Chapter 5: From fundamental to dynamic cap rates theory 37
5.1 Gordon model 5.2 Discount rate (r)
37 38
5.3 The New Risk premium Model 5.4 The new Dynamic Cap rate model 5.5 The new cap rate regression model specification
41 42 42
5.2.1 Band of investment model 5.2.2 Capital Asset Pricing Model (CAPM) 5.2.3 Summation technique
38 38 39
Chapter 6: Context, location, building and office users as cap rate determinants 45
6.1 context 6.2 location 6.3 property 6.4 office-users
Part III: Empirical research Chapter 7: DATA analysis and Synthesis
7.1 DATA DESCRIPTION
45 48 48 50
51 53
53
7.1.1 GBA database 7.1.2 TU/Delft property database 7.1.3 complimentary database
53 53 54
7.2.2 Transaction cap rates 7.2.2 Method | Total variance test
7.3.1 Inventory of input Variables 60 7.3.2 Calculating the relative vacancy risk 64 7.3.3 Method | Regression analysis, a multiple regression model 65
7.4.1 Cap rates Differences versus vacancy risk 66 7.4.2 Method | One-Way ANOVA test: comparing multiple means 67
7.2 Hypothesis 1 | appraised versus transaction cap rates 7.2.1 Appraised cap rates
54
7.3 Hypothesis 2 | cap rates determinants
59
7.4 Hypothesis 3 | the myth of vacancy and valuation accuracy 66
Chapter 8: Empirical Findings
55 55 59
70
8.1 Descriptive statistics
8.2 Hypothesis 1 | Overstated and smoothed appraised cap rates 75 8.3 Hypothesis 2 | forward looking investors versus backward looking Appraisers 78 8.4 Hypothesis 3 | Structural vacancy as a Paradoxical phenomenon 83
8.1.1 Outliers 8.1.2 Sale transactions per year 8.1.3 Sale transactions per city district 8.1.4 Cap rate development in Amsterdam
70 70 70 71 72
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
xi
Part IV: conclusions Chapter 9: Conclusion
85 87
Chapter 10: recommendation for further research
91
References
93
9.1 Cap rate determinants, the applicability of the new dynamic Cap rate model 87 9.2 valuation accuracy 88 9.3 appraised cap rates versus transaction cap rates 88 9.4 myth of vacancy and accuracy 89 9.5 overall conclusion 89
appendix 1 - cap rate determinants per reviewed article appendix 2 - Transaction cap rate Calculating models appendix 3 - regression model outcomes
97 102 107
1
INTRODUCTION
c h a p t e r 1: i n t r o d u c t i o n
1
Chapter 1: Introduction This thesis studies the determinants of capitalization (cap) rate, a rate of return on real estate investment, which reflects the risk associated with a property investment. Capitalization rate is a real estate valuation measure used in the most common valuation method in the Netherlands, namely the direct capitalization method. Due to limitations associated with the commercial real estate market (lack of transparency, information asymmetry, infrequency of transactions, etc.), investment decision making procedures and investment risk assessing procedures (to ultimately determine the property value), are prone to some uncertainty. These procedures are mostly combined with a qualitative assessment, due to difficulties in identifying the risk sources in real estate investment and the complexity of measuring it, which results in uncertainty and inaccuracy. To define the risk sources more specifically, this thesis aims at determining the cap rate components and quantifying the qualitative aspects of investment decision making and property valuation, by decompounding the determinants of capitalization rate. Because the risk is decompounded, the resulting cap rate model is more structured and flexible, which reduces the uncertainty and qualitative aspects of risk assessment procedures and helps to solve the problems associated with assessing the real estate investment risks. As a result, a new model is created which enhances the accuracy and consistency in real estate valuation, as it creates a dynamic and flexible method to measure the cap rates. This model can be used by many market parties (investors, appraisers, developer, etc.) to more precisely identify the risk inherited in the property investment which ultimately leads to determining the property value more accurately.
1.1 Target audience This newly created model and its results can be used by investors, appraisers and researchers who want to measure the risk associated with real estate in a more systematic and flexible way. Additionally this thesis can also be used as a starting point for further researching the risk sources that are associated with real estate.
1.2 RESEARCH OUTLINE Part I of this research discusses the research proposal, in which an introduction to the current Dutch office market is given in Chapter 2. Chapter 3 continues with further identification of the research problem and discussing the societal and scientific relevance of this research. It discusses the research questions, the main goal of the research, the research methodology and the main research approach, as well as appropriate methods per research question. Part II focuses on the theoretical framework of this research. Chapter 4, focuses on the background studies related to cap rates determinants. Chapter 5 explains the dynamic cap rate model specification. Chapter 6 is an overview of the variables found in the literature which have a significant effect on cap rates. Part III focuses on the empirical research. Chapter 7 introduces the sources of data used in this research. This chapter continues with discussing the data analyses and syntheses performed to test the research hypotheses. In Chapter 8, the main results of the empirical research are explained. Part IV addresses the overall conclusions of this thesis. Chapter 9 answers the research questions, while Chapter 10 provides some recommendations for the further research in this field.
Part I: Research Proposal
CHAPTER 2: A glance at the Dutch Office market CHAPTER 3: problem field
A glance at the Dutch Office market
2
4
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
CHAPTER 2: A glance at the Dutch Office market This chapter provides an introduction into the current situation of the Dutch office market. Section 2.1 discusses the dynamics of the Dutch office market and their interrelations. Section 2.2 explains the Dutch office market disequilibrium and its underlying causes. Section 2.3 explains the replacement market as a consequence of the gap between office supply and demand. Finally, in Section 2.4, the possible solutions to overcome this oversupply are discussed.
2.1 Dynamics of DUTCH OFFICE MARKET The vacancy rate rise of office markets in European cities is a phenomenon that requires more and more attention in order to deal with the risk of structural vacancy in the office stock (Remøy & Van der Voordt, 2011). The Dutch office market is no exception regarding this problem. According to fact sheets of mid-2014 by DTZ Zadelhoff (2014b), the vacancy rate of office buildings in the Netherlands is 15.7% which equals 7.8 million square meter of the total office stock. This is relatively high in comparison to a normal vacancy rate of 3% to 6% (Remøy, Koppels, Van Oel, & De Jonge, 2007). The Dutch office market is experiencing fundamental changes. On one hand, the low dynamics of the office market caused by the financial, plus real estate crises, and the changes in the behavior of office users reduced the required office space in this market (in 2014 7% lower take-up than 2013) (CBRE, 2013; Colliers, 2014; DTZ Zadelhoff, 2014a; Remøy & Van der Voordt, 2011). On the other hand, continuous overproduction and oversupply of office stock caused by accelerating expansion of this market from the 1980’s and onwards results in the high aforementioned vacancy rate, which indicates a relatively huge gap between demand and supply in the office market (Remøy & Van der Voordt, 2011). These facts caused the Dutch office market to become so off balance that this market never recovered since the 2001 crisis in real estate, even before the financial crisis in 2008 hit this market (Remøy & Van der Voordt, 2011). In 2014, fewer office spaces were taken up, creating a greater gap between supply and demand ratios, which contributed to the aforementioned 15.7% vacancy rate (CBRE, 2013; NVM, 2013; Remøy & Van der Voordt, 2011). Understanding the dynamics of this system can explain the current situation of the office market in the Netherlands, thereby identifying the underlying causes of its high vacancy rate. From a macroeconomic perspective, Geltner, Miller, Claytonand Eichholtz (2014) argue that the real estate market is a system that consists of three interrelated markets: the space market, the asset market and the development industry. The space market is the market for the use of space. This market is often referred to as the rental market which determines the rent level and the amount of cash flow a property can generate (Geltner et al., 2014). The asset market is the market for the ownership of properties. This market is also often referred to as the property market which determines the property market value and the flow of financial capital to the real estate market (Geltner et al., 2014). The development industry is the linkage between the property market (asset market) and the rental market (space market). It converts financial capital into physical capital by constructing new
c h a p t e r 2: A g l a n c e at t h e D u t c h O f f i c e m a r k e t
5
Figure 2.1, the real estate system and the interaction among three major components of space market, asset market and development industry, own illustration based on Geltner et al. (2014).
buildings, renovating and/or converting the existing buildings, thereby governing the supply stock in the space market (Geltner et al., 2014). Figure 2.1 illustrates an overview of this system (real estate market) and the interaction among its three components: space market, asset market and development industry.
2.2 Market disequilibrium In an efficient office market, due to the long-run equilibrium between and within the space and asset market, the quantitative and qualitative supply of office space is approximately equal to the demand (Lokhorst, Remoy, & Koppels, 2013). During the economic boom, the market shows signs of scarcity, while during economic recession a higher vacancy rate is dominant (Keeris, 2007). A vacancy rate of 3% to 6% of the total supply in office stock (frictional vacancy) is necessary to enable the mobility in the market (Rabianski, 2002; Remøy et al., 2007). Therefore, in an efficient office market, frictional vacancy is considered a normal status that is a requirement for providing fast solutions for the very dynamic demands for office space (Lokhorst et al., 2013; Rabianski, 2002). Conversely, in an inefficient real estate market, with a significant office surplus, the vacancy rate could increase considerably and excess the normal vacancy rate required to allow easy movement of office users from one place to another (Lokhorst et al., 2013; Rabianski, 2002). An example of such a market is the Dutch office market as explained in § 2.1. Even though the Dutch economy started growing slowly from 2013, there is no evidence of a decrease in the vacancy rate in this market since many companies continue focusing on optimizing
6
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
their location strategy, namely consolidation and decrease in space usage, meaning that there is no demand for expansion in this market (Savills, 2014). Besides that, a large share of vacant office stock is located in mono-functional areas with few or no other amenities, which do not match qualitatively to the demand of office users and is the result of the non-integrated planning policy of the Dutch municipalities with a lack of coordination between them (Rodermond, 2011). These all contribute to the oversupply in the Dutch office market due to a both quantitatively and qualitatively gap between the office supply and office demand, causing a high vacancy rate in this stock (Remøy & Van der Voordt, 2011).
2.3 Replacement market As (Remøy, 2007) argued “the office market has become a replacement market, since the stock in use is relatively stable and has no demand for expansion.” The rise of the replacement market and lack of quantitative need for new office buildings is a consequence of overproduction in the Dutch office market after 2000 (Remøy & Van der Voordt, 2013). In fact the construction activity has dropped to a relatively low level historically in the Netherlands. There is rarely any construction of new offices, unless the developers/investors are sure of future tenants (CBRE, 2012). In such a market, new buildings take the place of old and low quality buildings, which is caused by office-user preferences for certain types of buildings and locations. The emergence of the replacement market in which office buildings age at an ever increasing rate , resulted in obsolete office buildings which lead to a higher structural vacancy rate (Remøy, 2010). Structural vacancy, defined as vacancy of office properties for three or more years (Remøy, 2010), is a very critical issue in the Dutch office market. This is explained by the fact that the large amount of vacancies in this market (47%) are office buildings that are vacant for three or more years heading towards the obsolescence of these office buildings (NVM, 2013; Remøy & Van der Voordt, 2011). Nevertheless, obsolescence is not merely related to the age of the buildings (natural depreciation of the buildings) (Bullen & Love, 2011), but also represents aesthetic, functional, social, legal, economic and environmental obsolescence (Baum, 1993). This indicates that the occupancy of buildings is impacted by their potential to meet the current demand of both the office users and the investors. When this is not the case, the buildings are defined as obsolete (Bullen & Love, 2011). Dealing with vacancy in such a market is even more challenging, as many office buildings offered for rent or sale are no longer meeting current requirements of office users (e.g. sustainability), especially, considering the fact that a large amount of the total office stock (66%) consists of buildings over 18 years old (NVM, 2013). In such an unbalanced market, where large amounts of building stock are considered obsolete office buildings that reached the end of their functional and economic lifespan, the need for interventions are crucial to overcome this ever growing structural vacancy in the Dutch office market (Remøy & Van der Voordt, 2011).
c h a p t e r 2: A g l a n c e at t h e D u t c h O f f i c e m a r k e t
7
2.4 Adaptive re-use Adaptive re-use is defined as “a process that changes a disused or ineffective item into a new item that can be used for a different purpose” (Australian Government, 2004). As mentioned in the §2.3, the high vacancy rate is for a the large part attributed to properties that do no longer meet the current demand of office users and are therefore out of the active office property market (CBRE, 2013; NVM, 2013; Remøy & Van der Voordt, 2011). In essence, intervention in terms of transformation or adaptation of the existing buildings, is necessary to re-use these properties in the office market and to decrease the risk of obsolescence (Remøy & Van der Voordt, 2011). Several authors (Bullen & Love, 2011; Langston, Wong, Hui, & Shen, 2008; Remøy & Van der Voordt, 2011; Remøy & Van der Voordt, 2014) mentioned the adaptive re-use as an integral strategy to enhance the financial, social and environmental performance of buildings. In addition, several possible strategies for dealing with structural vacant offices are proposed, such as consolidation, renovation, demolition and new-build, and lastly, conversion to new functions (Remøy & Van der Voordt, 2014). Among these, the conversion of office buildings is seen (partly) as a solution to balance the oversupply of the office market (Remøy, Schalekamp, & Hobma, 2009; Remøy & Van der Voordt, 2011; Wilkinson, Remøy, & Langston, 2014). Remøy and Van der Voordt (2011) consider adaptive re-use in operational continuity of an obsolete office property either by renovation and within use adaptation of obsolete office buildings, or conversion to new use. Recent workplace adaptation strategies of the New Ways of working (NWoW), sustainability and the New Office Concept (NOC) are some examples of within use adaptation of obsolete office buildings. Figure 2.2 shows an example of the within use adaptation of the monumental building of De Rode Olifant which was vacant for 3 years before its adaptation plan (De Rode Olifant, 2011; Spaces, 2014). The possibility of conversion of structurally vacant offices to new use, is the conversion of obsolete office buildings into housing. Remøy and Van der Voordt (2014) argue that this decreases the shortage of the Dutch housing supply while simultaneously ameliorating the performance of obsolete office buildings by introducing a new use. This emphasizes on the fact that, if all office properties are adapted to accommodate modern office uses, the oversupply of office space will persist quantitatively. Therefore, functional transformations, demolition and new-build are inevitable strategies in order to develop an efficient office market (Remøy & Van der Voordt, 2014). Even though the adaptive re-use strategy is mentioned to be a possible solution to balance the oversupply of the Dutch office market and decrease the current structural vacancy (Remøy & Van der Voordt, 2014), they are often obstructed by a lack of financial feasibilities and interest by property owners as well as the developers (Remøy et al., 2009; Remøy & Van der Voordt, 2007; Scheltens, Van der Voordt, & Koppels, 2009).
8
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
Figure 2.2, Red elephant (Rode Olifant), within use adaptation (New Office Concept): this monumental building is located at the business district of the Hague. It was built in 1921 as the oil company Esso headquarters, providing an office area of over 10,000 m2. Between 1943 until the end of II World-War it was occupied by Germans for their own use. In 1993, this building was listed as a monumental building in the Netherlands. From 1990 to 2009, it accommodates the attorney firm of De Brauw & Westbroek, Blackstone. From 2009 to 2012 the building was vacant until Spaces bought the property and fully renovated it to the New Office Concept.This building now has 8,500 m2 of office area (small to mid-size offices), workspots, a business club for 500 persons, 12 meeting rooms, a cafe deli, a restaurant, conversations rooms, a library and 70 parking lots (Figure iv). Spaces made the building representative but also inspiring (e.g. for create brainstorm sessions). Additionally Spaces offer extra services such as storage, printing facilities, in house catering and massages. The same type of facilities are available at all the buildings owned by Spaces.By buying the building, Spaces created a cluster of office buildings that work together with the other locations in Amsterdam (Zuidas, Herengracht and Vijzelstraat). Source: (De Rode Olifant, 2011; Spaces, 2014).
Problem field
3
10
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
Chapter 3: Problem field This chapter discusses the research problem area further. Section 3.1 identifies one of the underlying causes for the lack of financial feasibility as one of the main bottlenecks of the transformation of obsolete offices. Section 3.2 defines two concepts of book value and market value. Section 3.3 elaborates on the various interpretations of book value while Section 3.4 determines the regulatory principle related to book value per type of investor. Section 3.5 sheds light on the ambiguity of the main cause for the lack of financial feasibility by providing a better insight into appraised value and its implication for vacant office properties. Section 3.6 discusses the common valuation method in the Netherlands and its fundamental input variables. In Section 3.7, the problem statement, the main research problem and its consequences are discussed. Section 3.8 focuses on the goal and objectives of this research. In Section 3.9, the main research questions and their subsequent subquestions are formulated. In Section 3.10, the research hypothesis are generated. After this, the scope of this research is determined in Section 3.11. Section 3.12 discusses the overall research methodology used to conduct this research by explaining the main research approach and the most appropriate research strategy to conduct this research. Section 3.13 clarifies the research design to fulfill the main objectives of the research. Section 3.14 discusses the qualitative and Section 3.15 addresses the quantitative methods and technique used to collect and analyze the data. Finally, Section 3.16 addresses the scientific and societal relevance for conducting this research.
3.1 Lack of financial feasibility - lack of willingness As mentioned in Chapter 2, adaptive re-use strategies as a possible solution to balance the oversupply of the Dutch office market, are often obstructed by a lack of willingness to sell by the property owners, as well as a lack of willingness to buy by the developers, due to a lack of financial feasibilities (Remøy & Van der Voordt, 2014). The lack of willingness to sell by the property owners leads to a situation that among all aforementioned possible strategies to deal with structural vacant offices in § 2.3, most property owners choose for the most passive attitude, namely consolidation. This means that they retain the status quo and opt for a wait-and-see tactic hoping for better times since selling is not an option for them. They are unwilling to make a loss on their investment (Remøy & Van der Voordt, 2014; TransformatieTeam, 2011). Such a behavior is explained by the prospect theory, where loss aversion refers to a tendency in whic avoiding losses is preferred over acquiring gains. Psychologically, losses are twice as powerful as gains (Haigh & List, 2005; Kahneman & Tversky, 1979). On the other hand, the willingness to buy and transform by the developer is low, when considering the costs of transformation and the future income of the transformed building (Remøy & Van der Voordt, 2014). The developers thus consider the transformation too risky compared to the expected return and thus avoid taking such a risk.
3.2 Book Value versus Market Value One of the underlying causes for the unwillingness of the property owners, found in the literature (Remøy et al., 2009; Remøy & Van der Voordt, 2007; Scheltens et al., 2009), is the difference between book value and market value of vacant properties. According to the International Valuation Standards Council (IVSC) (2014), Book Value (BV) or Carrying Amount is defined as “The amount at which an asset is recognized in the financial statements of
C h a p t e r 3: P R O B L E M F I E L D
11
an entity after deducting any accumulated depreciation and any accumulated impairment losses”, whereas Market Value (MV) is defined as “The estimated amount for which an asset or liability should exchange on the valuation date between a willing buyer and a willing seller in an arm’s length transaction, after proper marketing and where the parties had each acted knowledgeably, prudently and without compulsion.” The difference between book value and market value of vacant properties can be partly explained by the fact that even though the definition of book value in the financial book keeping is unique (IVSC definition), the interpretation to determine the book value varies with different accounting frameworks. In addition, each type of investor/property owner is obliged to follow the applicable accounting rules for them, which can differ for each type of investor. Therefore, in order to understand the relevant definitions and implications of book value, it is crucial to firstly define the different accounting frameworks and then to determine which type of investor should act under which accounting framework.
3.3 commercial and fiscal accounting perspectives According to Aart Hordijk, Director of the ROZ Real Estate Council (personal interview, October 20, 2014) and the definition of the IVSC (2014), book value is the value at which an asset is carried on a balance sheet. In the Dutch accounting framework, there are two types of balance sheets recognizable: commercial balance sheet and fiscal balance sheet. These two balance sheets fall under the individual annual account category (Van den Berg, 2011). A commercial balance sheet is used for general external reporting purposes (together with the profit and loss account), and provides information on the financial position of an entity. This type of individual annual account, namely the commercial one, is meant for performance measurement of the entity. According to the directives of the Council of Annual Reporting 930.9 (Richtlijnen voor de Jaarverslaggeving), this annual account is used as a base for a wide range of users to take economic decisions (e.g. employees, shareholders, debt investors, the government, etc.). The source and guidelines of commercial account reporting in the Dutch law are embedded in title 9, book 2 of the Dutch Civil Code, which indicates that the commercial account should be prepared based on acceptable norms in society (Van den Berg, 2011). On the other hand, the fiscal (tax) balance sheet determines the amount of tax which an entity is obliged to pay by considering the profit and loss account. This accounting framework is meant merely for the tax authority. The sources of fiscal accounting rules are determined according to sound business practice (goed koopmansgebruik) and relevant specific regulations in the Dutch tax law (Van den Berg, 2011). The Dutch accounting system is influenced by the International Financial Reporting Standards (IFRS). The IFRS aims to provide a global standard for external financial reporting and are established under the International Accounting Standards Board (IASB). The purpose of the IFRS is to provide a true and fair indication of the financial position of an entity (Van den Berg, 2011). The Dutch legislators define clearly in title 9, book 2 of the Dutch Civil Code, when an entity may use the Generally Accepted Accounting Standards in the Netherlands (Dutch GAAP) or IFRS, which depends on the size of the legal entity.
12
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
3.4 book values and type of investors According to Aart Hordijk (personal interview, October 20, 2014), three types of investors can be recognized in the Dutch office market: institutional investors, private investors and corporate real estate (owner-occupied). For each of these investors/property owners, the book value principles are determined in this section. 3.4.1 Institutional investors & book value Institutional investors consist of pension funds, insurance companies, banks, investment institutions and investment funds (Cuppen, 2011). The institutional investors are tax-exempt and therefore they are only obliged to prepare the commercial annual account. This is used as a performance measurement and should follow the International Accounting Standards (IAS) 40 investment properties (Hordijk, personal interview, October 20, 2014). The IAS 40 allows either the fair value model or the cost model as its accounting policy for investment property. Under the fair value model, the investment property is reappraised to fair value. The fair value is defined as “an amount for which an asset could be exchanged between knowledgeable, willing parties in an arm’s length transaction (IAS 40.5). Fair value should reflect the actual market state and circumstances as of the balance sheet date (IAS 40.38). The best evidence of fair value is normally given by current prices on an active market for similar properties at the same location and in the same condition and subject to similar lease and other contracts (IAS 40.45).” (Deloitte, 2014c) On the other hand, under the cost model, an investment property is valued at cost of replacement/ reproduction minus accumulated depreciation and any accumulated impairment losses. However, when an entity adopts the cost model, they must still obtain fair values as IAS 40 specifies that reporting the fair values within the financial statements are mandatory. Therefore, it is more beneficial to use the fair value model than the cost model, when fair values still have to be estimated (Collings, 2012). This means that institutional investors are obliged to adjust their book value annually in order to reflect the actual market value. Therefore, in principle, there is no difference between book value and fair market value estimated by the professional appraisers. As a result, for the institutional investors, book value is in fact the appraised fair market value. 3.4.2 Private investors & book value Private investors are a relatively heterogeneous group of real estate investors, who cannot be specified as one group (Cuppen, 2011). Private investors have two types of balance sheets, namely a commercial balance sheet and a fiscal balance sheet (Hordijk, personal interview, October 20, 2014). The fiscal balance sheet which is used for tax purposes, is the one that integrates depreciation into the book value. This depreciation is more related to the structural and physical condition of a building that is used for taxes under the Dutch GAAP. However, private investors under fiscal Box 3 (incomes from savings and investments), whom have a fixed presumed gain of 4% of the market value under the Box 3 assets minus debt, are taxed at the flat rate of 30% (Deloitte, 2014d). This is in addition to 1.2% over the market value which they always have to pay (Hordijk, personal interview, October 20, 2014).
C h a p t e r 3: P R O B L E M F I E L D
13
For the private investors’ commercial balance sheet, they get advice on the market value from the appraisers/ accountants based on the Dutch GAAP (the Generally Accepted Accounting Standards in the Netherlands). However, the final decision to implement the advised value (appraised market value) to the commercial balance sheet depends on the willingness of the private investors. The economic decision making about the market value is thus not based on the current book value, but rather on the judgment of investors about the re-appraised value. Therefore, it is very hard to judge whether the appraised commercial book value of private investors reflects the actual market value. Clarifying this issue requires a lot of resources and is limited due to the confidentiality of such information. 3.4.3 Corporate real estate & book value Corporate real estate is the real property occupied or held by a business entity for its own operational use and/or to support corporate expansion (Krumm, 2001). Corporate real estate investors have two types of balance sheets, namely a commercial balance sheet and a fiscal balance sheet (Van den Berg, 2011). Non-listed corporate real estate investors have to follow the IAS 16 Property, Plant and Equipment for preparing their fiscal balance sheet. Under the IAS 16 property, plant and equipment is measured at its cost, under either using a cost model or revaluation model. Furthermore, it is depreciated in a way that the depreciable amount is allocated systematically over its useful life (Deloitte, 2014a). Figure 3.1, the type of property owners, their accounting frameworks and regulatory standards
14
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
Listed corporate real estate investors have to follow the IAS 36 Impairment of Assets. The IAS 36 Impairment of Assets seeks to ensure that an entity’s assets are not carried at more than their recoverable amount (Deloitte, 2014b). The interpretation of the IAS 36 is that the book value in the financial statement of a company should be corrected downwards to the fair market value. This is due to the fact that book value is the asset minus liability and the value of the assets may decrease due to depreciation over a period of time by an asset (IVSC, 2014). Therefore, according to Aart Hordijk (personal interview, October 20, 2014), their balance sheet must be devaluated. Figure 3.1 summarizes the relation of the different accounting frameworks, the type of property owners and the applicable regulatory principles per type of investor and accounting framework that is discussed in § 3.3 and § 3.4. Considering the aforementioned sources of determining the book value, it can be argued that the statement about the high book value in comparison to the market value of vacant properties (Remøy et al., 2009; Remøy & Van der Voordt, 2007; Scheltens et al., 2009), is in fact the high appraised value estimated by the professional appraisers. This is because economic decisions regarding an asset (or a portfolio) are not based on the book values, but rather on the appraised values.
3.5 Valuation accuracy: APPRAISED MARKET VALUE VERSUS ACTUAL MARKET VALUE As the real estate market is a very thin market with no frequent transactions as well as lack of public information (transparency), property appraisers are playing an essential task in real estate market as a substitute for selling prices (Fisher, Miles, & Webb, 1999; Hordijk, 2005). As a result, studying the differences between the appraised market value and the actual selling price of a property is an essential task to reflect the true market volatility (Hordijk, 2005). Actual market value is the selling price of a property at which a deal is actually done, namely the transaction price (Geltner et al., 2014). As mentioned in § 3.2 and § 3.4.1, appraised market value is defined as “an amount for which an asset could be exchanged between knowledgeable, willing parties in an arm’s length transaction” (Deloitte, 2014c). Lusht (2012) elaborates on this and mentions that, in fact “market value is ‘the most probable transaction price’ which a specified interest in real property is likely to bring.” Studying the differences between the appraised market value and the transaction price of a property falls under the main topic of valuation accuracy which according to Baum, Crosby, Gallimore, McAllister, and Gray (2000) is defined as “the probability of a valuation being within certain parameters of a sale price”. In the valuation accuracy literature, smoothing and lagging are mentioned as sources of inaccuracy in real estate valuations (Baum et al., 2000; McAllister, Baum, Crosby, Gallimore, & Gray, 2003). Smoothing refers to an under-measurement of actual variance and anchoring on past appraisals, while lagging refers to the fact that appraised values cannot accurately capture the timing of market movement and fall behind prices (Baum et al., 2000). However, uncertainty is not only limited to the lagging effect and capturing an accurate market movement (McAllister et al., 2003), but is also embedded in determining the market value of a vacant office building, since it generates no rental income and may have a relatively low promise for future tenancy (Rodermond, 2011; Schiltz, 2007). French and Gabrielli (2004) argue that the latter
C h a p t e r 3: P R O B L E M F I E L D
15
leads to more uncertainty in the estimation of the market value in comparison to a valuation of a property which has been just let and is not vacant. Schiltz (2007) argues that measuring the risk associated with vacancy on appraised market values creates substantial difficulties for professional appraisers. Schiltz (2007) believes that vacancy is a threat embedded as a risk for all types of real estate, which exist or will happen eventually, whether the property is already let on long term leases or not. Therefore, all the risks associated with vacancy, should be reflected in the property value. Van Gool and Rodermond (2011) support the argument pertaining vacancy risk by Schiltz (2007), and emphasize on its importance, specifically on the problematic situation of structurally vacant offices. This because these are office properties which are vacant for a considerable amount of time and are characterized by having a relatively low promise for future tenancy. Determining the market rent for these offices is thus problematic. In addition, it is not certain how many years of vacancy should be taken into account for value estimation. Above all, the last and the most complicated issue is to determine a correct capitalization rate due to the lack of any market transactions in order to finally calculate the value of the property. These three reasons show clearly that common valuation approaches, sale comparison or income approach, are not the best approaches to estimate the value of vacant properties, since there is a lack or even no available input for these methods. Thus, making it very difficult to determine market value for vacant office properties.
3.6 Direct capitalization and its fundamental input variables Even though the income approach, which determines the value based on the building operating income, is not the most appropriate method to estimate the price of a structurally vacant office property (asmentioned shows that this approach is the most common approach = + ∗ −in§3.5), + 1research − ∗ + + + − in appraising a vacant office property in the Netherlands (Hendershott, 1996; Hordijk & Van de Where; Ridder, 2005; Ten Have, 2000). More precisely, the most frequent and common method within the = risk free rate, income approach used in Dutch practice, according to Rodermond (2011), is the direct capitalization ⁄ = Loan-to-value ratio, method. = rate of return on debt, = premium from participation in real estate,
= premium on location attributes, (1), the property’s Net Operating Income (NOI) of the initial year In thismethod, as shown in equation = premium on property-itself attributes, is divided by the cap rate (also known as Net Initial Yield (NIY) or going-in cap rate), to substantially = premium on property-user attributes, estimate the property value (Geltner et al., 2014; Lusht, 2012): = a constant expected rate of growth in the NOI. From equation (1), it is clear that there are two fundamental determinants of value in direct capitalization, namely cap rate and NOI. As mentioned in § 3.5, vacancy risk is reflected in the
=
(1)
(
Where; V = property value, NOI = Net Operating Income in the first year, Ro= overall capitalization rate (%).
property value (Schiltz, 2007), therefore it should be included in calculating the NOI and/or cap rate. Vacancy is reflected in rental income, as in the case of vacant property, the object is valued = based upon the assumption of ‘market value, as if let on the valuation date’. Then, the latter value will beWhere; corrected ex-post for certain losses on void periods, which are the expected timeframe that = expected required rate of return (discount rate %). a property remains vacant and generates no rental income. However, vacancy risk is also reflected
=
(
(
16
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
in the used cap rates (or in general both gross and net yields; gross yield when value is calculated by gross potential income, and net yield when value is calculated by use of net operating income). This is confirmed with the research done by Schiltz (2007) to study the impact of vacancy on value and gross yields used by property appraisers, in two forms of occupancy rate and the remaining lease term. This study ascertains that appraisers are prone to uncertainty when apprising a vacant property, as the variation in (value and yield) outcomes are larger in comparison to a let property. Schiltz (2007) shows that deviations in both value and yield between appraisers are higher when a property has a shorter remaining lease term or is vacant, compared to a property with a longer remaining lease or the probability of renewing the lease by current tenants. The study by Schiltz (2007) shows that the changes in yields and value are thus proportional to the property occupancy rate. In addition, the study performed by De Roo (2014) shows that there is a huge gap (45%) between the transaction-based cap rates and appraisal-based cap rates (in this study the assessed value cap rates) for non-prime locations with a large number of structural vacant offices in Amsterdam, which has a high amount of obsolete office buildings (40%). All these show that vacancy is a risk inherent for any property and is reflected in the value of real estate in two ways: on the expected cash flow, and on yield (both gross and net). However, this thesis argues that even though there are uncertainties in estimating the NOI, with considerable effort both appraisers and potential buyers arrive to a reasonable accurate NOI. Whereas, variations in the yields and therefore value studied by Schiltz (2007) and De Roo (2014), show that determining an accurate yield is a relatively challenging task, given the fact that the available information on property transactions relating to structurally vacant offices are scarce in the current office market. However, structural vacancy and therefore lack of transaction in the market should also be considered as market evidence and should be incorporated as one of the risk components, when determining cap rates (in case of calculating the value with a NOI rather than a gross potential income).
3.7 Problem Statement Real estate valuation plays an essential role in a nontransparent and thin property market, specifically the office property segment, where properties are not frequently sold. These valuations are the basis for negotiation in transaction processes, advice on property financing (lending) decisions, property taxation, and finally the recorded performance of the overall property market (Baum et al., 2000). Previous research has indicated that a high degree of uncertainty in determining the value of a vacant property has a negative influence on valuation accuracy (Rodermond, 2011; Schiltz, 2007). Besides the fact that appraisal-based indices (both value and cap rate indices) tend to be smoothed and lagged when compared with those of transaction-based indices, many parties (e.g. investors, financiers, redevelopers, etc.) believe that valuations of vacant offices are too high in comparison to their actual market value (Rodermond, 2011).In addition, variations in the yields (gross and net) used to capitalize the property rental income show that determining an accurate yield is a relatively challenging task, given the fact that the available information on property transactions relating to structurally vacant offices are scarce in the current office market (De Roo, 2014; Schiltz, 2007). The inaccurate valuations have a negative impact on adaptive re-use strategies (adaptation and conversion of current office stock) due to a lack of financial feasibilities, even though they are
C h a p t e r 3: P R O B L E M F I E L D
17
mentioned to be a possible solution to balance the oversupply of the Dutch office market and decrease the current structural vacancy (Remøy et al., 2009; Remøy & Van der Voordt, 2007; Scheltens et al., 2009). Another consequence of inaccuracy in estimating property valuation is that it influences the estimation of the solvency rate and judgment about the financial performance of property owners. Solvency rates (short-term and long-term) are ratios which intend to provide information regarding the investors’ ability to meet their financial obligations, or in other words whether cash flow income of investors is sufficient to meet their liabilities (Hillier, Ross, Westerfield, Jaffe, & Jordan, 2013). An overestimated solvency rate provides a false indicator of the investors’ financial position. This creates an increased risk for the financial institutions in the Netherlands and ultimately for the Dutch society. Since valuations are the basis for estimating property taxes, inaccuracy in valuations has an influence on the cash flow of the property owner. The higher the assessed value (the WOZ-value), the more tax property owners have to pay. For the Dutch municipalities, a high WOZ-value could be interesting, since they receive more tax money. On the other hand, it also works against them, since it raises a lot of objections in court against them which exhausts a lot of time and resources from the municipality (De Roo, 2014). In addition, the inaccuracy in valuation of vacant offices illustrates a different volatility in appraisal values throughout time compared to the real market values. As a result of such a difference, in a volatile market, valuations do not show the true volatility of the real estate market (Baum et al., 2000; Hordijk, 2005). This has a relatively high impact on appraisal-based indices such as the IPD index in the Netherlands as a market performance measurement (Schiltz, 2007). These all show that an inaccurate appraised valuation in comparison to the actual market value (transaction price) of vacant offices has many consequences for the Dutch real estate market, Dutch financial institutions, Dutch municipalities as well as the Dutch society.
3.8 research Goal The purpose of this research is to contribute to the body of knowledge of valuation accuracy for structurally vacant office buildings in the Netherlands by determining the cap rate components as one of the important input variables for determining the property value. Since there is some degree of valuation inaccuracy and inconsistency between appraisals and transactions (Baum et al., 2000; De Roo, 2014; Hordijk, 2005; McAllister et al., 2003; Rodermond, 2011; Schiltz, 2007), the main goal of this thesis is to understand the differences and similarities in pricing processes of two main market players, namely appraisers (for their valuation) and investors (for the actual transactions) to enhance accuracy and consistency in office property valuation in the Netherlands. The two main objectives of this research are: to examine the main determinants of office property cap rates respectively for appraisers (valuations) and investors (transactions); subsequently, to gauge whether the differences between appraised cap rates and transaction cap rates are significantly larger for vacant offices in comparison to the non-vacant offices. In other words, whether any variation in accuracy depends upon the variation in vacancy rate. This is all with the hope to increase awareness about valuation accuracy for vacant office properties among various market players (e.g.
18
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
investors, appraisers, developers, etc.) by increasing transparency in the pricing process of these parties by studying the variation in their cap rates. Hopefully, such a transparency can enhance reasonable pricing of properties which facilitate the adaptation process of the obsolete office stock and overcome the oversupply of the Dutch office market.
3.9 Research Questions Based on the problem statement mentioned in § 3.7 and the objectives of this research mentioned in § 3.8, the main question of this research reflects the two folded objectives of the research and consists of two sub-parts: 1. To what extent do appraised cap rates correspond with transaction cap rates? 2. Can the differences between appraised cap rates and transaction cap rates be explained (partly) by structural vacancy risk? The sub research questions related to the first sub-part of the main question are categorized around the three major themes of: determinants of cap rates, valuation accuracy and the variation in pricing mechanism of appraisers and investors. I. What are the main determinants of the capitalization rate? a. How does the background context, such as macro-economic, capital market, real estate market and sale conditions affect variations in cap rates? b. How do macro and micro location characteristics affect variations in cap rates? c. How do property-specific conditions, such as vacancy rate, age, and leasehold conditions, affect variations in cap rates? d. How do office users and their lease terms affect variations in cap rates? II. How large are the differences between the appraised cap rates and transaction cap rates? e. How large are the average absolute differences between appraised cap rates and transaction cap rates? f. Is there any tendency for appraised cap rates to consistency understate or overstate the transaction cap rates? III. What are the differences and/or similarities between appraised cap rate and transaction cap rates? g. What are the determinants of cap rate respectively for appraisers (valuations) and investors (transactions)? h. Do appraisers and investors weigh cap rate components differently and if so how? The sub research questions related to the second sub-part of the main question are categorized around the theme of valuation accuracy of vacant office properties. IV. How is accuracy of the estimated cap rates associated with variations in vacancy rates? i. What are the different types of vacancy? j. How are different vacancy types correlated with differences between appraised cap rate and those from transactions?
C h a p t e r 3: P R O B L E M F I E L D
19
Figure 3.2 illustrates the conceptual model of this research and the relation of the main research questions and the main themes.
3.10 Research hypotheses Based on the existing literature that views the inaccuracy of valuation as one of the bottlenecks in the transformation of vacant offices in the Netherlands (Remøy et al., 2009; Remøy & Van der Voordt, 2007; Scheltens et al., 2009), the following hypotheses are generated. Hypothesis 1 The appraised cap rates of Dutch offices during the rising market are lower (appraised value is higher) than the transaction cap rates (prices), to a larger extent than the declining market (after financial crisis). Hypothesis 2 Appraisers put more weight on easily observable elements in determining office cap rates whereas investors are more concerned with macro-economic variations over time. Hypothesis 3 The differences between appraised cap rates and transaction cap rates of an office property have a positive correlation to its vacancy risks. Figure 3.2, conceptual model showing the relation among various themes and subquestions of this research
20
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
3.11 Research scope This research focuses on the Dutch office market, specifically on the asset market (property market) which is one of the three main components of the real estate system as explained in § 2.1. This research focuses within the Dutch office market, on the office properties in Amsterdam. This is due to availability of data and greater number of office transactions in this city. Within the Amsterdam office market, it narrows down to the accuracy of property market value. Within the process of determining the property value, it focuses on the determinants of cap rates as the most important factor in property investment analysis. Furthermore, it zooms in on the valuation accuracy of vacant office properties.
3.12 RESEARCH APPROACH This research uses a mixed methods approach where qualitative and quantitative research strategies are combined (Kumar, 2014). The main rationale for using a mixed methods research is to answer the different types of research questions defined for this research (Bryman, 2012). This approach is the most appropriate method as it achieves the two-folded objectives of this research the best (see § 3.8). The qualitative method in this research is a descriptive and exploratory approach used to develop a theoretical framework for the empirical part of the study. This is to understand the background as well as the different aspects of each sub-part of the research problem (Bryman, 2012; Kumar, 2014). The research hypotheses are constructed based on this qualitative descriptive and exploratory study. Whereas, the quantitative research is used to determine the extent of the research problem by providing statistically significant findings. In addition, the quantitative research is used as a correlational study to establish association (if any) among aspects under the study (Kumar, 2014).
3.13 RESEARCH DESIGN The research as a whole is designed in a two-phase study based on the previously mentioned twofolded objectives of the research in § 3.8. The sequential research design is chosen to fulfill the main objectives of this research for each phase. The sequential order begins with a descriptive and an exploratory qualitative method which is followed by an exploratory and a correlational quantitative study (Kumar, 2014). Figure 3.3 illustrates the different stages designed to conduct this research. The TU/Delft graduation processes (P1 to P5) are also integrated in this scheme.
3.14 QUALITATIVE RESEARCH METHODS Qualitative research is used, firstly to provide a theoretical framework to support formulating and analyzing the research problem regarding valuation accuracy. It is also performed to create a general attitude towards valuation of vacant office properties (in theory and practice) and the difficulties and uncertainties pertained to appraise a (partly) vacant office. Secondly, it is used to enrich the theoretical framework to answer the first sub-part of the research question (first fold of the research objective) in regards to the three main themes of: determinants of cap rates, valuation accuracy and the variation in pricing mechanism of appraisers and investors. This implies a qualitative study to explore determinants of cap rates, and to understand the causes of inaccuracy in valuation by explaining certain behaviors and processes in valuation by different parties. Qualitative research is used for the second sub-part of the research question to identify different types of vacancy relevant for determining office cap rates.
C h a p t e r 3: P R O B L E M F I E L D
21
Figure 3.3, research design and research planning
The required data for the qualitative part is obtained both through primary and secondary sources. The method used to collect primary data was through unstructured interviews with experts in the property appraisal field, namely appraisers and academic researchers. In contrast, content analysis of the existing literature is used as a method to provide data for the qualitative research study. These include sources such as academic publications, official documents from the European Union, the Dutch government or private sources, press releases, and newspapers, all in relation to the main themes of the research questions. Both literature study (content analysis) and unstructured interviews were used for research sub-questions I, III, and IV.
3.15 QUANTITATIVE RESEARCH METHODS Quantitative methods are firstly used to determine what the components of cap rates are that are used in the Dutch office market, by appraisers and investors respectively. Secondly, they are used to determine the extent of the research problem by quantifying the differences between appraised and transaction cap rates and to ascertain the magnitude of variation in appraised cap rates to transaction cap rates due to the variation in vacancy rate. The method used in the quantitative part is secondary data analysis of data collected by other researchers and by public and private organizations in the course of appraised cap rates, transaction cap rates and determinants of cap rates for office properties in the Netherlands. The sample for data analysis is selected based on the following criteria: - office transactions in Amsterdam from 2004 till 2011; - properties with various vacancy rates; - properties which have appraised market values available for the same as the transactions took place; - samples for years before and after the credit crunch.
22
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
Statistical analysis is used for the quantitative research part. The statistical method used for sub-question II is a total variance test, to quantify differences between appraised cap rates and transaction cap rates. For sub-questions III, a multiple regression technique is used to examine the determinants of cap rates in the Dutch office market for both appraisers and investors. Finally, for sub-question IV, One-Way ANOVA test is applied to analyze the mean variation for differences in appraised and transaction cap rates and different office vacancy rates. The research techniques are implemented using SPSS software for statistical analysis. These methods are explained thoroughly in Chapter 7.
3.16 Scientific and societal relevance This section describes the two major relevancies for conducting this research: the scientific relevance that emphasizes on the knowledge gap, and the societal relevance that focuses on the consequences of a less accurate appraised market value of vacant offices for the Dutch society. 3.16.1 Scientific relevance There are several studies related to valuation accuracy in the Netherlands, which are discussed in this section. For the first time in the Netherlands, Hordijk (2005) studied the valuation accuracy concerning commercial real estate based on the ROZ-IPD index over the years 1995-2002. This was a comparison between the achieved selling price and the latest valuation. An average deviation of 7.9% was the result of the differences between appraisal value and the transaction price. This is a relatively high number when compared to the USA (-0.1%) and the UK (5.7%) (Hordijk, 2005). However, the time frame of this study was related to the economic expansion and the result does not represent the depressed market after the 2008 financial crisis in the Netherlands. In addition, the focus of this study was not on the valuation accuracy of vacant offices. Rodermond (2011) has studied the differences between the transaction prices and previous valuation of 19 vacant office properties from 2008 to 2011 in the Netherlands, which resulted in an average difference of -11.72%. However this study does not cover the period before the 2008 crisis. A more recent study which compared valuations and sale price in 12 national markets from 2000 to 2013 including the Netherlands was done by IPD (2014). The weighted average absolute difference between 2004 and 2013 in the Netherlands is 8.1% (IPD, 2014). Even though this analysis covers the time frame of before and after the financial crisis, it does not focus on the commercial real estate market and especially not on the differences between valuations and sale prices of vacant office properties. Besides the literature on valuation accuracy which focused on comparing the property value among appraisers (appraised value) and investors (transactions), a recent study by De Roo (2014) shows the inaccuracy in valuation by examining the differences between appraisal-based cap rates and those of transactions. The results of this study show a huge gap of 45% between the appraised cap rates and transaction cap rates in office areas in Amsterdam with a high vacancy rate. This study illustrates the importance of micro location (submarkets) and structural vacancy in variation of cap rates. However, this study does not quantify the exact correlation of structural vacancy with used cap rates.
C h a p t e r 3: P R O B L E M F I E L D
23
In fact, there are not many studies in the Netherlands to analyze determinant of yields in the Dutch office market except those of Van Norren (2007) and Verhaegh (2005), which show that cap rates are strongly determined by macro-economic variables rather than aspects such as micro location and property-specific elements including the risk of structural vacancy. Based on previous studies (De Roo, 2014; Hordijk, 2005; Rodermond, 2011), it is clear that there is some degree of inaccuracy in Dutch valuation practice. Unfortunately, none of the studies mentioned focus on understanding the pricing mechanism which results in such an inaccuracy in valuations. Therefore, this research emphasizes on the need for a more comprehensive research in this field, to study the differences between the appraisers’ and investors’ pricing mechanism. This research focuses on a larger time frame (before and after the credit crunch), cap rate determinants of both appraisers (valuations) and investors (transaction prices) and specifically focuses on the relation of the valuation accuracy with vacant office properties. 3.16.2 Societal relevance As mentioned in § 3.7, an inaccurate estimation of a property value may lead to a false indicator of the investors’ ability to meet their long-term obligation in respect to the cash flow obtained from the subject property. For instance, because the value of the property assets does not reflect the fair market value there is no accurate indication of the solvency rate of important financial institutions that have large investment portfolios in real estate (e.g. pension funds, insurance companies, banks, etc.). As a result, the financial stability of the important financial institutions is unclear. This worries both the Dutch Central Bank (DNB) as well as the European Union (Van Hoeken & Bruyn, 2014; Vermeulen, 2014). Therefore, the Dutch Central Bank (DNB), together with the AFM decided to conduct research in 2012 on the solvency rate of the Dutch banks. This was to establish a more accurate value of the property assets financed. Unfortunately, this research has not been made public. The property market thus has its influence on the capital market, which in turn has an influence on society (Figure 3.4). A recent example where real estate almost caused the bankruptcy of a bank is that of the SNS bank in the Netherlands. This bank bought the “plagued” Bouwfonds from ABN AMRO which later caused the downfall of the SNS bank (Van Hoeken & Bruyn, 2014). SNS’ property finance portfolio consisted of many high risk loans. To prevent the bank from bankruptcy, the Dutch state nationalized the bank. The direct costs involved with this were 3,7 billion euro, which increased the Dutch budget deficit in 2013 with 0.6% to 3.3%. The national debt rose with 1.6% and other banks had to pay a total amount of 1 billion euro in 2014. It is thus of societal relevance to research the valuation accuracy in relation to the fair property value to determine the actual solvency rate and thereby the financial stability of the financial institutions. Figure 3.4, the impact of property market (real estate crisis) on the capital market (financial crisis) and society (economic crisis)
property market
capital market
society
Part II: Theoretical framework
Chapter 4: Determinants of cap rate, a background study Chapter 5: From fundamental to dynamic cap rates theory Chapter 6: Context, location, property & office-users
Determinants of cap rate, a background study
4
26
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
Chapter 4: Determinants of cap rate, a background study This section reviews the main determinants of cap rates as found in the academic literature. In essence, studies discussing the determinants of the office market capitalization rate, can be divided in two major categories of macro-level and micro-level variation in cap rates (Hoesli & Chaney, 2014; McDonald & Dermisi, 2008). The macro-level variation studies, focus on market-specific variables that capture cap rate variations over time (time-series), such as those that depend on capital market, macro-economic variables and the Metropolitan-Specific Area (MSA). The micro-level variation studies emphasize on property-specific factors which vary from property to property and/or per location (cross-sectional). The rest of this section reviews these studies chronologically based on the country of focus (international and Dutch literature), their methodology, data and ultimately their results.
4.1 Cap rate Determinants in international literature One of the earliest studies on determinants of cap rates is by Froland (1987) in the U.S. According to the author, the movement on the total investment market is more crucial on cap rate variation than simply observing the changes in the real estate market. According to Froland (1987), this is due to the fact that real estate investment is merely one of the various investment opportunities for the investors. Froland (1987) argues that cap rates for property valuation, function the same as price-earnings ratios for the stock exchange market. As a result, they must be approximately equal to the opportunity cost of capital of investors, increased with a risk premium for real estate. To prove this, Froland (1987) examines whether cap rate movement is driven by the capital market and whether its movement is comparable to other yields in the asset trading market. This study analyzes the American Council of Life Insurance (ACLI) dataset from 1970 to 1986, based on transactions done by large institutional investors in the United States. Froland (1987) finds a positive correlation between cap rates and mortgage rates. Cap rates are associated inversely by the earnings/price ratio and ten-year bond rates. In addition, Froland’s study reports a negative effect of economic cycle factors on the cap rate, such as national vacancy rate, the percentage change in real Gross National Product (GNP), and capacity utilization. However, Jud and Winkler (1995) criticize Froland’s work as Froland’s empirical results were not set up based on any theoretical framework; secondly, through using the stepwise regression approach, the mortgage rates, the earnings/price ratio and ten-year bond rates could explain around 86% to 95% of the variation in cap rates. In addition no corrections for autocorrelation were tested. Ambrose and Nourse (1993) used the same ACLI dataset as Froland (1987); however, unlike Froland (1987), their study does not suffer from a lack of ground theory. Ambrose and Nourse (1993) explain the cap rate based on the band of investment model, also known as the Weighted Average Cost of Capital (WACC) in the corporate finance literature (Jud & Winkler, 1995). The WACC approach is based on the theory that the overall cost of capital is the weighted average of debt cost and equity cost (Ambrose & Nourse, 1993; Lusht, 2012). Ambrose and Nourse (1993) analyze transaction-based cap rates of several commercial properties (office, retail, industrial, hotel, etc.) in the U.S. from 1966 to 1988. They use two different statistical techniques to model cap rates: Seemingly Unrelated Regression (SUR) and cross-sectional/timeseries regression. Their empirical results show that using SUR, no significant correlation between cap rates and either the earnings/price ratio or the bond risk premium spread is found. The results of
C h a p t e r 4: d e t e r m i n a n t s o f c a p r at e , a b a c kg r o u n d s t u d y
27
their cross-sectional/time-series regression (panel data), record that cap rates are positively related to the percentage of equity investment, cost of debt, and expected inflation, while being negatively related to earnings/price ratio. In addition, they show that cap rates vary significantly by property type (office, retail, industrial, hotel, etc.). However, cap rate is not strongly tied to a location. Jud and Winkler (1995) elaborated on the study of Ambrose and Nourse (1993) by examining the relation between cap rates and financial variables, in an ad hoc approach. This is achieved by formulating a cap rate model derived from the financial literature, not only based on the WACC model, but also combining it with the Capital Asset Pricing Model (CAPM). The CAPM is a model that describes the relationship between risk and expected return on equity with two factors: time value of money (the risk-free interest rate) and expected market return. In accordance with Ambrose and Nourse (1993), Jud and Winkler (1995) suggests that cap rates are determined by required returns in the debt and equity market. Their findings show that cap rates and capital market returns are significantly related. However, they do not adjust quickly to capital market changes (lagged). In addition, the market relationships change strongly across different locations. According to their empirical findings, Jud and Winkler (1995) ascertain the fact that the property market is not an efficient market and does not fully integrate with the national capital market. Another study explaining macro-level variation in cap rates by analyzing aggregated cap rate data, is from Sivitanides, Southard, Torto, and Wheaton (2001). The authors believe that real estate markets are naturally segmented by metropolitan area and that macro local market conditions play an important role in determining variations in cap rates, both through time and across metropolitan areas. They formulate a cap rate model as a function of the risk-adjusted discount rate and the expected growth rate of income. Their final model was a Time-Series Cross-Section (TSCS) regression model which uses a logarithm of cap rates as a dependent variable, which in turn is related to six regressors: the logarithm of cap rates of the previous year, a Metropolitan-specific real rent index, an annual percent change in the metropolitan- specific real rent index, Government Bond Return Rate as an indicator of the risk-free opportunity cost of capital, the yearly percentage change in CPI as an indicator for inflation expectation, and finally MSA (Metropolitan-specific area) dummies as a fixed effect for each individual market. They investigate the determinants of appraisal-based cap rates using the database of the National Council of Real Estate Investment Fiduciaries (NCREIF). This database contains annual office cap rates for 14 metropolitan areas in the U.S. from 1984 to 2000. The empirical results of their research contains three main aspects. First, cap rates vary persistently across markets due to differences in fixed market characteristics. These characteristics influence the investor’s perceptions of both future income growth and associated risk. Second, changes in market-specific cap rates consist of components which are shaped by historical trends in rental growth and rent levels in the specific market. Such a relationship with historical trends and movements in cap rates, indicates that appraisal-based valuations strongly rely on past evidence in the market, rather than having a forward looking perspective. This is consistent with the arguments regarding anchoring behavior of appraisers by (Baum et al., 2000; Hordijk, 2005; McAllister et al., 2003). Third, cap rates are influenced by national developments in capital market and economy. In addition, Sivitanides et al. (2001) argue that investors are acting irrationally when estimating cap rates as they do not consider the mean reversion of real rents in their expectation about real rental growth. In fact, they use lower (or higher) cap rates when rental cyclical peaks (or troughs), resulted in having overvalued (or undervalued) properties.
28
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
McDonald and Dermisi (2008) continue with the model developed by Jud and Winkler (1995) to formulate their cap rate model. In addition, they enrich their model with variables that may indicate a value change of office properties from the perspective of an investor. McDonald and Dermisi (2008) intended to explain the investors’ behavior to show how they use several variables, such as class and age of buildings, as proxies for forecasting changes in the market value of office properties. Therefore, their study employs not only macro-economic and financial variables that depend on the capital market, but also explain micro-level variations in cap rates by considering individual properties as the unit of observations. The authors apply their model on a database that consists of 132 office building sales in Chicago between 1996 and 2007. Their findings show that a newer building, a class A building, a building which has been renovated, had a negative effect on cap rates, while a decrease in office employment and increase in the local market’s vacancy rate, has a positive relation with cap rates. The latter shows that investors pay close attention to changes in the vacancy rate and its magnitude in that specific market, to forecast the changes in office property prices, as McDonald and Dermisi (2008) suggest in their study. The study by Netzell (2009) is a significant step forward in explaining the micro-level variation in appraisal-based capitalization rates in Sweden. Netzell (2009) examines the rationality of the Swedish property valuations in terms of whether the appraisal cap rates follow the economic theory. This is achieved not only by adding extra micro-level variables, such as micro location dummies (submarkets within the macro office market), but also by comparing the cap rate model to the micro-level explanatory factors with the macro-level and time-series variables. This study explores the micro level variation of both appraisal-based going-in and exit cap rates in three Swedish office markets, namely Stockholm, Gothenburg and Malmö. The author constructs the cap rate model according to the Gordon model (1962) where cap rate equal the difference between property expected required rate of return and rental growth. In Netzell’s model, cap rates are regressed on both property-specific variables and broad time-series variables. The author applies the feasible generalized least squares (FGLS) regression technique to overcome the heteroskedasticity (when the variances of the observations are unequal) and serial correlation in dataset. In addition, the author included location variables to show cross-sectional and time-series variation in cap rates at two levels. The first level includes three Swedish office markets: Stockholm, Gothenburg and Malmö; while the second level includes locational dummies for three segments in each office market: CBD, city center and peripheral property submarkets. The explanatory used as proxies for the property quality included in the regression equation, are the age of the property, ground lease dummy, rent ratio between current rent and market rent and vacancy gap (the difference of current vacancy and long-run vacancy). Moreover, explanatory factors expressing broad time-series variations such as ten-year governmental bond, earnings per share and real earnings on stock are also used in the regression. As mentioned in the previous paragraph, the strength of this study is that the author compares cap rate model regressed only to the micro-level explanatory factors with those that were based on macro-level and time-series variables. This is done by using three different approaches in modeling cap rates where the results of each model is compared with the others to explore different level variation of cap rates. The first is a regression model which is ran using year dummies to examine the existence of any variation over time which is common for all properties. The second is a regression model, which initially ran using fundamental time-series variables to study the effect of macro level variation of cap rates over time. Substantially, this model was tested by means of a two-way
C h a p t e r 4: d e t e r m i n a n t s o f c a p r at e , a b a c kg r o u n d s t u d y
29
ANOVA test for cross-sectional differences in cap rates, which resulted in showing around 85-90% of variation in cap rates in the three office markets. Finally, the third model tested, is a regression model using merely cross-sectional explanatory variables in the model to compare it with the first two models (Netzell, 2009). The results of Netzell’s (2009) empirical study show that a lower market rent and higher long run vacancy rate indicate the quality of a property and are associated with higher cap rates. Going-in and exit cap rates are both negatively associated with higher market rent at the property level. In addition, cap rates are positively related to long run vacancy rates, though exit cap rates are more sensitive to this than going-in cap rates. The age of the property was only significant in one of the Swedish office markets, and not significant for the other cities. A higher cap rate is associated with properties with a ground lease. This is due to the fact that leasehold conditions result in having less control over the property, and therefore are associated with higher risks for the investors. Also, properties located in the peripheries are positively related with cap rates. These all show economically reasonable relations. Going-in cap rates are significantly related to how the actual (current) rent in the first year deviates from the market rent. The same is valid for the deviation of the current vacancy from the long run vacancy rate for the property. In addition, cap rates are positively related to risk-free rate and negatively related to the earnings per share on stock (Netzell, 2009). The overall findings deduced from this research indicate, that first, the variation in cap rate is more related to cross-sectional than time-series variables. Second, it shows that submarket locations (segments) are strongly effecting the determinants of cap rates in Sweden. Furthermore, the direction and magnitude of the explanatory cross-sectional variables are consistent with economic theory (in general with what happens in the market), thereby showing the rationality of Swedish property appraisals. The results for the determinants which indicate broad time-series variations, however, are partly inconsistent with other literature. This may be the result of the short time span of this research, as the time-series variables exhibit different results for different office markets (three cities) and their segments (Netzell, 2009). Chervachidze and Wheaton (2013) argue that a large number of previous cap rate studies considered cap rates are determined by risk-free interest rates and market fundamentals, essentially rent levels and rental growth. This indicates that cap rates are modeled pursuing an adjustment process around equilibrium values. The authors describe this as a ‘Null hypothesis’, meaning a standard literature-based model which uses real estate fundamentals and risk free rates in explaining cap rate variations. However, Chervachidze and Wheaton (2013) believe that the impact of macro-economic capital flows, namely the debt availability on cap rate and property value, is very strong. This study is conducted using the ordinary least squares (OLS) regression model combined with the fixed effects panel technique on appraisal-based data of the NCREIF Property Index in the U.S. since the 1980s. The fixed effects panel technique increases the model coefficients and efficiency. This is achieved by correcting the model for both time-series and cross-sectional variation among metropolitan-specific areas using MSA dummies. According to the authors “this framework is consistent with theoretical expectations that market-specific unobserved characteristics will lead to permanent differences in capitalization rate trends across markets, and the fixed effects method allows us to estimate the effect of these unobservables and test for their statistical significance.”
30
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
Chervachidze and Wheaton (2013) follow a three stage specification progression of cap rate model equations. First, they specified the null hypothesis model which is the most comprehensive model reflecting the standard approach used in the previous literature such as Sivitanides et al. (2001). The authors term this as The “Null” Specification: Market fundamentals and Treasury Rates where cap rates are determined around the equilibrium values. According to Chervachidze and Wheaton (2013), “the equilibrium is estimated at the same time as the adjustment and is determined by two sets of influences: first, the influences of a discount rate that reflects both the opportunity cost of capital and systematic market risk; second, fundamental factors that shape investors’ income growth expectations. This is in line with the literature, which usually uses rental fundamentals and some proxy for interest rate to explain cap rates.” The second model specification is extended by including a risk premium, which according to Chervachidze and Wheaton (2013) is defined as economy wide risk premium over the risk-free rate. This explanatory variable is estimated by subtracting the Moody’s AAA Corporate Bond Index and the 10-year T-Bond. The results of this test show that this variable is positively related to cap rates, due to the fact that investors require to be compensated for a higher risk for a lower property value with the same rental income. Adding the aforementioned variable resulted in a higher goodness of fit tests and adjusted R2 statistics. Chervachidze and Wheaton (2013) emphasis on the significant influence of the risk premia required by investors on property values and criticize the existing literature for omitting it. In the third model, Chervachidze and Wheaton (2013) extended their model specification by adding a debt availability variable. This variable is proxied by the yearly change in total debt outstanding to the GDP. Even though traditional financial economics argue that, in an efficient market, financial structure and the ratio between debt and equity should not impact the property value (Lusht, 2012), recent macro-economic theories view debt availability as a frequent cause of financial crises, while microeconomic argues that debt can enhance property liquidity. This is explained with the fact that when the availability of debt is limited, real estate transactions are more difficult and consequently the property values drop down and, become lower than their fundamental value. As Chervachidze and Wheaton (2013) argue, the easy debt boosts the property transactions which leads to an increase in the property values (even higher than their fundamental prices). This is mentioned to be one of the possible underlying causes of the formation of the price ‘bubbles’ in the property market. The results of the last model show that debt availability is negatively related to cap rates due to the fact that investors associate debt scarcity with illiquidity and higher cap rates. The overall findings deduced from Chervachidze and Wheaton (2013) show that local property fundamentals (rent) can merely explain a small part of the variations in cap rates. However, the macro-economic factors of real T-rate, bond risk premium debt availability affects cap rates strongly. This is because these variables vary at macro level over time and have no cross-sectional variations (cross MSA). Therefore, Chervachidze and Wheaton (2013) argue that the importance of the location, as one of the most important factor for the real estate does not seem to be valid in their study. However, Chervachidze and Wheaton (2013) examine the association of the property values and the
C h a p t e r 4: d e t e r m i n a n t s o f c a p r at e , a b a c kg r o u n d s t u d y
31
overall growth of debt in national economy. This is while, as the authors mention, only 6% of the total current (public and private) debt is devoted to commercial real estate debt. Hoesli and Chaney (2014) complement the earlier research about cap rate determinants not only by adding several property-specific factors which were not tested by previous literature, but also they are the first who compare the determinants of appraisal-based cap rates with transaction-based cap rates in detail. By this, they attempt to enhance the understanding of the underlying reasons behind the differences between valuations and transaction values. They conduct their research on the IAZI database, the largest real estate database in Switzerland, for the timeframe between 1985 and 2010. This study formulates a cap rate model based on the Gordon model (1962) and further develops it as a function of the risk-adjusted discount rate (sum of risk-free interest rate and risk premium), and the expected growth rate of income. Subsequently, they split both risk premium and the expected rate of growth into micro and macro contributions. This is based on the aforementioned two categories of research that have studied either cap rate variations at a macro- or micro-level. In addition, for their empirical model they use variables from previous research in three main streams of: the capital markets, the risk associated with the investment related to local market conditions as well as the individual property, and the investors’ assumption about property value changes in the future, which rely on local market conditions as well as the individual property. As an indicator for the risk-free rate, they considered the yield on government bonds with a maturity of ten years. To include the macro-level risk premium component of cap rates, like previous studies, the connection between stock market is considered. However, to show the micro-level risk premium, variables such as ownership leverage (an existing leasehold or easements), land leverage, sale conditions (auctions, at arm’s length, off-market transactions), tenants risk and diversification, refurbishment risk (age and property condition), illiquidity risk (building volume), and the quality of the location are included in their model. Compared to the previous literature, Hoesli and Chaney (2014) were the first one to include new variables such as auctions, off-market transactions, easements, tenant risk, and construction and building quality at the property level to explore cap rates variations. Finally, as an indicator for rental growth, vacancy rates, rent, GDP and inflation are considered. Their empirical results show that the aforementioned property-specific risks are significantly associated with variations in cap rates. For instance, the ownership leverage (an existing leasehold or easements) is positively correlated with cap rates while construction quality and building condition are negatively associated with variation in cap rates. In addition, factors such as the condition under which a transaction is performed, whether on a free and transparent market, auction or off-market transactions (e.g., between family members or a related legal entity), are used with investors as a proxy for micro-level risk premium and have a strong explanatory power for variations in cap rates. By comparing the determinants of appraisal-based cap rates with transaction-based cap rates, Hoesli and Chaney (2014) confirm that there are differences between appraisers and investors in the way they percept real estate risk which ultimately results in a divergence of the property value (appraised and transaction prices). According to Hoesli and Chaney (2014), the most obvious differences are that risk-free rate, the ratio of rent to average rent, age and volume are more
32
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
important for investors while building condition and macro location are considered more important for appraisers. In addition, investors put more weight on factors such as vacancy rate and risk-free rate, rent relative to average rent whereas these factors are not or less significant for appraisers. Instead, appraisers overweight factors such as building condition and micro location while these are less significant for investors. However, variables such as macro location, price to earnings ratio of stock and GDP are equally significant for appraisers and investors. In addition, in opposite to the findings of Chervachidze and Wheaton (2013), Hoesli and Chaney (2014) suggest that the wellknown real estate dictum of ‘location, location, location’ is confirmed as location has a significant role in the pricing mechanism of properties for both appraisers and investors. However, it should be mentioned that appraisers put more weigh on location than investors. Overall findings of Hoesli and Chaney (2014) suggest that investors and appraisers focus on different factors when estimating cap rates and therefore the property value. The appraisal-based cap rates vary more across property and less over time in comparison to the transaction related cap rates. In fact, appraisers put more weight on elements that are easy to observe (property-specific characters) whereas investors are more concerned with macro-economic variations over time. As Hoesli and Chaney (2014) argue, this is consistent with the appraisal smoothing discussion and clarifies one of the underlying causes of the smoothing effect, as property-specific factors (as the focus of appraisers), hardly vary over time. Another remark by Hoesli and Chaney (2014) regarding investors, is that they focus on a portfolio level and are more concerned with non-diversifiable risks which cannot be eliminated in the portfolio, and less with the risks associated with the individual property when determining cap rates.
4.2 cap rate Determinants in Dutch literature None of the previously discussed literature is determining the cap rates variation in the Netherlands. In fact, the research about determinants of cap rate are very scarce in the Dutch literature. Among all, three master thesis are reviewed to understand the main components of yield (gross and net yields) in the Dutch office market. The first study focuses on determinants of the Gross Initial Yield (GIY) (Verhaegh, 2005), whereas the second study examines the drivers of cap rate (Van Norren, 2007). The last (third) study has a focus on comparing the appraise-based cap rate and those of transactions in the Netherlands (De Roo, 2014). Verhaegh (2005) examines the relevant determinants of the gross initial yield (GIY) at both macro and location level, and micro level (property specific variables) in the Dutch office market, using the ROZ/IPD Property Index, the main benchmark index for Dutch real estate investments of institutional investors in direct property. The database includes 1410 office properties for 2002 and 2003. Due to the limitation of available data, Verhaegh (2005) was not able to include all the relevant determinants of the gross initial yield (GIY) in the analyses. The main five determinants in this study are rent gap or potential measured as the difference between the real rent and market rent, market rent, remaining lease term, operating costs and the age of the property. Data is analyzed with a multiple regression model including the GIY as the dependent variable and the five aforementioned variables as independent variables which resulted in an explanatory power of 40% of variation in GIY. The results show that all the variables, except operating costs and age, are negatively associated with GIY. However, it seems that the age of the property has only limited impact on the GIY. In addition, Verhaegh (2005) adds four main regions in the Netherlands, namely Amsterdam, Rotterdam, Utrecht and The Hague to the model and concludes that GIY varies strongly per region. The overall result of
C h a p t e r 4: d e t e r m i n a n t s o f c a p r at e , a b a c kg r o u n d s t u d y
33
this study indicates that the GIY is more associated with macro-economic and location factors than property-specific variables. However, since this study is performed for a very short time period of 2 years (2002 and 2003), the results should not be generalized as they do not provide strong market evidence. To determine cap rate variations, Van Norren (2007) creates a cap rate model using both macroeconomic factors and local factors (four major cities of Amsterdam, Rotterdam, Utrecht and The Hague) between 1986 to 2006. In total, four separate models are created where the annual cap rate per city is used as the dependent variables separately per model. The main variables Van Norren (2007) determines for these multiple regression models are respectively yearly inflation rate, real risk-free interest rate, GDP (as an indicator of the economic growth), investment volume, the previous cap rate itself, the performance of the stock market, changes in absolute vacancy rate, and rent ratio measured by dividing the current rent level by the long run average rent. However, Van Norren (2007) mentions that due to the lack of available data, merely a few of these variables could be used in order to make a comprehensive cap rate model for the Netherlands. The overall findings of this study show that cap rates are strongly determined by macro-economic factors where their influence on cap rates barley differentiates per location. Among all macro-economic factors, the ten-year government bonds rate has the most explanatory power in variation of cap rates in all the cities. A contrary finding of this research is that vacancy rate did not indicate a strong association with cap rates, except for Utrecht. This is inconsistent with Hoesli and Chaney (2014), McDonald and Dermisi (2008) and Netzell (2009), there vacancy rates are used as a proxy to indicate the changes in the local market and rental growth. De Roo (2014) compares the transaction-based Net Income Multiplier (NIM), the reciprocal of the cap rate, to the assessed value NIM (in Dutch the WOZ-waarde kapitalisatiefactor) in the Amsterdam office market from 2004 to 2013. The findings show that there is a huge gap between the NIM of the transactions and the assessed value. The results show that the differences between these two NIM (of transaction and of assessed value) are not always showing the same relation (always higher or lower). In fact, at the prime office locations in Amsterdam (center and south), the assessed value NIMs are 20% lower (cap rates are higher) than transaction-based NIMs (cap rates). Whereas, for structural vacant offices located in the non-prime locations in Amsterdam (west and south-east), the assessed value NIMs are 45% higher (cap rates are lower) than transaction-based NIMs (cap rates). De Roo (2014) argues that the underlying cause of this difference, especially in the non-prime office locations with a lot of structural vacancy, is that office properties are barley sold as vacant, and when that happens, there is a huge decrease in the property value. This is while one of the base assumptions in determining the assessed value in the Netherlands is that the property should be valued as it is vacant. Therefore, such a decrease in the value in the actual market value is not considered by appraisers as valid market evidence, which results in a huge gap of 45% between assessed value NIMs and those of the transactions. However, it should be mentioned that even though this research points out the differences between transactions (investors) and assessed value (appraisers), it does not fully examine the relevant determinants of cap rates (or NIMs) as the underlying causes of such differentiation.
4.3 Cap rate determinants conclusion It is clear that there are two main approaches towards determining cap rate variations. Those that indicate the dominance of macro-economic and time-series variations in cap rates and the others
34
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
where micro-level and property-specific factors are more powerful in explaining cap rate variations. These differences in macro- and micro-level cap rate variations is partly caused by the type of data (transaction-based cap rates or appraisal-based cap rates) which was examined and/or the country of focus. Mostly where cap rate models are based on appraisal-based data, the results show a higher association of cap rates with the historical trend on cap rates (serial correlation) or overweighting variables that barely vary over time (individual property), which both support the appraisal smoothing theory. Whereas when the transaction-based cap rates are modeled, cap rates are more associated with macro-economic variables, capital market and expected income growth. In general, in most of the U.S. studies, the influence of macro-economic and financial variables is more dominant (Ambrose & Nourse, 1993; Chervachidze & Wheaton, 2013; Froland, 1987; Jud & Winkler, 1995; Sivitanides et al., 2001) than the presence of micro-level variation with a focus on individual property, except for McDonald and Dermisi (2008). Among all, indicators of capital market such as required rate on stock and government bonds, financial variables such as availability of debt, macro-economic factors such as growth in GDP, the differences between rent from its historical trend, vacancy rates in the macro location are used to explain car rate variations. In contrary, more recent studies in Europe, namely in Sweden (Netzell, 2009), Switzerland (Hoesli & Chaney, 2014) show the importance of micro-level factors in determining cap rates. Among all, property age, condition and size, quality and type of users, leasehold, ratio of current rent to median (market) rent, sale conditions and location are mentioned frequently. Finally, the limited studies performed in the Netherlands (Van Norren, 2007; Verhaegh, 2005), show that cap rates are strongly determined by macro-economic factors where their influence on cap rates barely differentiate per location. However, the recent study of De Roo (2014) indicates a gap between appraisal-based NIM and those of transactions, illustrating the importance of micro location (submarkets) and structural vacancy in variation in NIM (also in cap rates). Table 4.1 summarizes the cap rate determinant related studies, their focus (macro or micro level variation in cap rates) and the main limitation. For the complete list of variables found in the studies and their relations to cap rates see Appendix 1.
C h a p t e r 4: d e t e r m i n a n t s o f c a p r at e , a b a c kg r o u n d s t u d y
Table 4.1, the cap rate determinant related studies, their used data, their focus (macro or micro level variation in cap rates) and their main limitation
Focus
35
Name
Year
Country
Used data
Main limitation
Froland
1987
U.S.
transaction-based macro level
did not model cap rates based on any theoretical framework
Ambrose and Nourse
1993
U.S.
transaction-based macro level
cap rates are not strongly tight to a location
Jud and Winkler
1995
U.S.
appraisal-based
macro level
did not considered micro level cap rate variations
Sivitanides, Southard, Torto, and Wheaton
2001
U.S.
appraisal-based
macro level
did not considered micro level cap rate variations
Verhaegh
2005
Netherlands
appraisal-based
macro level
not all the relevant determinants of GIY were used, specifically the micro level determinants
Van Norren
2007
Netherlands
-
macro level
due to the lack of available data, merely a few of these variables could be used in order to make a comprehensive cap rate model
McDonald and Dermisi
2008
U.S.
transaction-based both
It studies only one city in the U.S.
Netzell
2009
Sweden
appraisal-based
both
the expected sign for the time series variables were not consistent with other studies due to the short time span used
Chervachidze and Wheaton
2013
U.S.
appraisal-based
macro level
this study ignores the micro level variations in cap rates
Hoesli and Chaney
2014
Switzerland
both
both
their cap rate model is based on the Gordon model which is rather primitive
De Roo
2014
Netherlands
both
-
does not fully examine the relevant determinants of cap rates
5
From fundamental to dynamic cap rates theory
C h a p t e r 5: F r o m f u n d a m e n ta l t o d y n a m i c c a p r at e s t h e o r y
37
Chapter 5: From fundamental to dynamic cap rates theory This chapter the main theoretical frameworks used for modeling cap rates, as previously = focuses + ∗ on − + 1 − ∗ + + + − discussed in Chapter 4, and further develops them to build a dynamic cap rate model. The chapter beginsWhere; with the use of the basic and static direct capitalization formula (equation 1, mentioned in § = risk free rate, 3.6), and step bystep + ∗ develops − + it1by−integrating ∗ +the Gordon’s + +model, −adynamic discount rate model ⁄ = = Loan-to-value ratio, and itsrelated theory such as the bandof investment model (or WACC), the CAPM and = rate of return onestimations debt, = premium from participation in real estate, summation technique, and a risk premium concept to finally define the dynamic cap rate model Where; free rate, = risk = premium on location attributes, specification.
5.1
Loan-to-value ratio, on property-itself attributes, ⁄==premium of return debt, = =rate premium on on property-user attributes, Gordon model ==apremium participation in real in estate, constantfrom expected rate of growth the NOI.
As mentioned in § on 2.6, the direct capitalization method, estimates the property value by capitalizing = premium location attributes, property-itself attributes, = premium the property’s NetonOperating Income (NOI) of the initial year by the cap rate (equation 1). = premium on property-user attributes,
= = a constant NOI. + expected ∗ − rate of + growth 1 − inthe ∗ + + + − = = + ∗ − + 1 − ∗ + + + −
(1)
Where; Where; = risk free rate, Where; ⁄ = V value, ratio, rate, ==property riskLoan-to-value free NOI Net Operating of return onIncome debt, in the first year, ⁄===rate = Loan-to-value ratio, R capitalization rate (%). in real estate, o= =overall premium from participation = rate of return on debt, Where; premiumfrom on location attributes, = =premium participation in real estate, = level value, WhenVa cash flow is expected to be gained in perpetuity, the discount rate can replace the cap =property premium on property-itself attributes, = premium on location attributes, = Net Operating Income in theattributes, first year, NOI = premium on property-user = rate and be used to capitalize the annual on property-itself o premium R = =overall rate (%).attributes, NOI, as a shortcut to estimate the value of the property capitalization constant expected rate of growth in the NOI. = =a premium on property-user attributes,
(
(
(
(Lusht,2012) which results in equation (2):
= a constant expected rate of growth in the NOI. Where; = expected required rate of return (discount rate %).
= =
(2)
Where; = == expected required rate of return (discount rate %). Where; V = property value, Where; Where; However, is valid when nothe changes in the future income is expected till perpetuity and does not =this Net Operating first year, NOI = requiredIncome rate ofinreturn (discount rate %), value, V = expected property R = overall capitalization rate (%). o takes into account the changes in the future value of a property due to the increase or decrease in = = =expected rate ofIncome growthininthe thefirst NOI.year, NOI Net Operating the operating Ro= overall income. capitalization rate (%).
Where; required rate of return (discount rate %), expected == =− limitation, the Gordon model (1962) considers a growing perpetuity at a constant To overcome thisrate = expected of growth in the NOI.
rate ofg= andexpresses that the property value equals a perpetual future NOI cash flow divided by a Where; = = required ∗ + rate 1 − of ∗return (discount rate), minus a constant expected rate of growth (5) constant expected = expected required rate of return (discount rate %). Where; in theNOI, in equation (3): −shown = as Where; = expected required rate of return (discount rate %). = overall cost of capital, = = ∗ ratio, + 1 − ∗ ⁄== Loan-to-value = cost of debt, = cost of equity. = Where; Where; = overall cost of capital, = expected required rate of return (discount rate %), ⁄ = Loan-to-value ratio, Where; = expected rate of growth in the NOI. == =expected cost of (̅ debt, required rate of return (discount rate %), + − ) cost of equity. == expected rate of growth in the NOI.
(3) (5)
By rearranging equation (1) and (3), the overall capitalization rate can be calculated as follows: Where; = − = expectedreturn on property, == risk free +− (̅ = rate, − ) (4) ̅ = expected market = = ∗ return, + 1 − ∗ (5) = equity beta (( , )⁄ ). Where; = = ∗ + 1 − ∗ (5) = expected return on property, Where; = risk free rate, = overall cost of capital, ̅ = expected market return, Where; ⁄ = Loan-to-value ratio, = equity beta (( cost of capital, = overall , )⁄ ). = cost of debt,
(
( (
(
( (
(
( (
(
( (
Where; 38 = risk free rate, ⁄ = Loan-to-value ratio, = rate of return on debt, = premium from participation in real estate,
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
Considering equation (4), cap rates can be written as a function of discount rate (r) and growth in = premium on location attributes, NOI (g) which indicates factors affecting r and g, may affect cap rates as well. = premium on property-itself attributes,
= premium on property-user attributes,
5.2 Discount (r)rate of growth in the NOI. = a constantrate expected
The discount rate (the required rate of return by investors), considers both time value of the money, and the uncertainty and risk associated with the future cash flow of a specific investment, compared to risks offered by other types of investment in the capital market. In other words, it in fact reflects = the opportunity cost of capital and market risk considering the similar risks of an alternative investment (Geltner et al., 2014; Lusht, 2012).
(1)
Where; V = property value, Due to the nature of the real estate market (heterogeneous, infrequent transactions, and lack of NOI = Net Operating Income in the first year, R o= overall capitalization rate (%). transparency), mentioned in § 2.5, estimating the discount rate is a challenging task in comparison
to other types of investment such as treasury bills or corporate bonds (Lusht, 2012).
=sections discuss the most used methods in order to forecast the discount rate, which has The next been used as a base theory by many authors (Ambrose & Nourse, 1993; Chervachidze & Wheaton, Where; 2013;Hoesli & Chaney, 2014; Jud &(discount Winkler, = expected required rate of return rate1995; %). Netzell, 2009; Sivitanides et al., 2001) mentioned in chapter 4, to model cap rates.
(2)
5.2.1 Band = of investment model As mentioned in § 4.1, the band of investment model, or the WACC approach (Weighted Average Cost of Capital), is used by Ambrose and Nourse (1993), Jud and Winkler (1995) and McDonald Where; = expected required of return rateto %),measure the impact of financing on the return on and Dermisi (2008), to rate model cap(discount rates and = expected rate of growth in the NOI. investment. According to Jud and Winkler (1995), the WACC, as defined in equation (10), used in corporate finance literature, is in fact the required rate of return (discount rate) on projects funded by firms using = −both debt and equity received form capital markets.
(3)
= = ∗ + 1 − ∗
Where; = overall cost of capital, ⁄ = Loan-to-value ratio, = cost of debt, = cost of equity.
(4)
(5) (5)
In the appraisal literature, the band of investment model calculates the overall cost of capital as the = weighted + (̅ − ) sum of the average of expected cash returns to the lender and the investor (Brueggeman & Fisher, 2011; Lusht, 2012). The rationale for the band of investment model explained by (Lusht, Where; 2012)is= expected that since most properties are purchased by the use of both debt and equity, therefore return on property, free rate, parts (debt and equity) should be provided by the annual NOI of a property. = riskto the return these ̅ = expected market return, The debt cost (debt service)⁄is a return to the lender, while the residual income of NOI after debt = equity beta (( , ) ). payment is a return on equity to the investor. As a result, the overall cost of capital is the weighted average of debt cost and equity cost. 5.2.2 Capital Asset Pricing Model (CAPM) The CAPM is another approach to estimate the discount rate by suggesting that the expected return on equity (r) is time-varying (as the investors should be compensated for time value of money), and dependent on expected market return (r ̅_m) as shown in equation (6):
(6)
= overall cost of capital, ⁄ = Loan-to-value ratio, = cost of debt, cost C h a p te r= 5: F rofoequity. m f u n d a m e n ta l t o d y n a m i c c a p r at e s t h e o r y
39
(6)
= + (̅ − )
(
Where; = expected return on property, = risk free rate, ̅ = expected market return, = equity beta (( , )⁄ ).
By considering the fact that the expected return on investment consists of two slices of debt and equity, from equation (5) and (6), the discount rate is written as equation (7):
= ∗ + 1 − ∗ = + (̅ − )
(7) (7)
→ = + ∗ − + 1 − ∗ (̅ − )
Where; = expected return on property, = risk free rate, ̅ = expected market return, = equity beta (( , )⁄ ).
< ̅
RTP , then RTP- RV is a negative amount, therefore expecting VTP The absolute differences is selected to record the magnitude of the relation between the different vacancy risk (market conformed, potential structural, structural vacancy risk) and the differences in cap rates. 7.4.2 Method | One-Way ANOVA test: comparing multiple means In order to measure the relations between differences of appraisal and transaction cap rates, and different vacancy risks, the One-Way ANOVA test is used, since the directional differences between RV and RTP have interval form which is compared with a vacancy risk variable which is a categorical variable (Field, 2013). There are three categories of market conformed, potential structurally vacancy risk, and structurally vacancy risk under the vacancy risk variable, therefore the test is performed by applying post hoc tests. Post hoc tests are a multiple comparison procedures which designed to compare a combination of all different subgroups (at least three groups within a variable) to find
68
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
possible differences without having a prior prediction about the differences in data (Field, 2013). In fact this method compares the mean of directional and absolute differences between RV and RTP with each vacancy level under the vacancy risk variable. The primarily assumption for applying the ANOVA test is provided as data are normally distributed within the groups per variable (Field, 2013). After checking the basic assumption of the normality, the test is ran and the means are compared per each category.
Empirical Findings
8
70
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
Chapter 8: Empirical Findings This chapter discusses the outcome of the empirical research as described in Chapter 7. First the descriptive statistics and the characteristics of the final database are explained. Subsequently, the results of the tested hypotheses are discussed.
8.1 Descriptive statistics The final database consists of 124 office transactions in Amsterdam from 2004 till 2011. Even though the sale-transaction data in the GBA database registers around 500 office transactions of single WOZ objects from 2004 till 2014 in Amsterdam, the final database consist of much less transactions. This is due to the fact that when combining the databases important data, needed to calculate cap rates, was missing. Missing variables include the availability of effective rent levels, WOZ cap rates, and most importantly the lack of BAG ID’s for all the WOZ objects in the database. As a result, the final database contains only data which includes both appraised cap rates and transaction cap rates (expost calculated) and which has a corresponding BAG ID. 8.1.1 Outliers From the 124 office transactions, 3 outliers are recognized and deleted from the sample (see Figure 8.1). In order to recognize the outliers, the z-score of the variable transaction cap rate is calculated. The outliers are defined as cases with a z-score above 2.58 (in absolute terms). The results show three significant outliers which were observations with a z-score above 3.29. These outliers are removed from the sample to avoid their influence on biasing the mean and enlarging the standard deviation in the statistical findings (Field, 2013). Figure 8.1, outliers, scatter plot transaction cap rates in Amsterdam from 2004 to 2011
,80
Transaction Cap Transaction CapRate Rate
,60
,40
,20
,00 2004
2005
2006
2007
2008
2009
2010
2011
Transaction Year Transaction Year
8.1.2 Sale transactions per year The number of sale transactions recorded per year varies significantly. However, the overall pattern shows a gradual increase in the number of transactions towards the financial crisis and a steady decrease in the number of transactions after the financial crisis. Figure 8.2 displays the number of transactions per year. Figure 8.3 shows the average WOZ value versus transaction prices of properties
C h a p t e r 8: E m p i r i c a l F i n d i n g s
Error bars: 95% CI
WOZ value
Transaction price
Mean
25
TransactionSale WOZ_Value
10.000.000 20
Error Bars: 95% CI
7.500.000 Frequency
Mean
Figure 8.2,(right) the number of transactions, and Figure 8.3 (left) average WOZ value versus transaction price per year from 2006 till 2011
71
5.000.000
15
10
2.500.000 5
0
2004
2005
2006
2007
2008
2009
2010
0
2011
2004
2005
2006
2007
2008
2009
2010
2011
YearOfSale
SaleYear Year Transaction
Transaction Year
in Amsterdam from 2004 till 2011. The differences between the WOZ values and transaction prices increased from 2005. The average transaction prices are higher and fluctuated more when compared to the WOZ values of the same year.
Page 1
8.1.3 Sale transactions per city district Within the municipality of Amsterdam, there are eight city districts: Center, New-West, North, East, West, Westpoort, South and South-East (see Figure 8.4). The number of sale transactions recorded per Amsterdam district varies considerably (see Figure 8.5). The majority of the sale transactions occur in Amsterdam Center (45 transactions from 2004 till 2011). Next to the Amsterdam Center, Amsterdam South and West have the most sale transactions, respectively 23 and 19 transactions in the eight year period of 2004 to 2011. The least number of transactions belong to the Amsterdam districts of New-West (5 transactions in total) and Westpoort (4 transaction in total). Table 8.1 shows that the number of sale transactions per year per city district from 2004 till 2011. The districts of New-West, East Westpoort show low dynamics in the office market after the financial crisis. The number of transactions per year is very scarce in these districts. The city Center on the other hand, shows a more stable market where the number of transaction remains almost the same, when compared to before and after the financial crisis. Additionally, in 2007, there were a larger number of transactions (11 in total), in comparisons to the other years. 50
Westpoort North West New-West
Center East
South
Amsterdam Districts
40
Frequency
Figure 8.4,(left) Amsterdam city districts, and Figure 8.3 (right) the number of transactions per districts from 4002 to 2011
30
20
South-East
10
0
Center
New-West
North
East
West
Westpoort
South
South-East
Amsterdam Districts
Amsterdam districts
Page 1
72
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
Total #
Transaction year
Total # of
City districts
transactions
transactions
2004
2005
2006
2007
2008
2009
2010
2011
Centre
5
5
5
11
3
8
4
4
45
New-West
1
-
1
1
2
-
-
-
5
North
3
1
2
-
1
1
1
-
9
East
-
-
2
-
2
1
-
2
7
West
4
1
2
5
3
3
1
-
19
Westpoort
1
-
-
1
-
2
-
-
4
South
1
2
3
5
5
3
-
4
23
South-East
-
-
-
1
4
1
2
1
9
15
9
15
24
20
19
8
11
121
Total # transactions per year
per district
Table 8.1, the number of sale transactions per year per city district from 2004 till 2011 in Amsterdam
8.1.4 Cap rate development in Amsterdam Figure 8.6 shows the cap rate development (both appraised and transaction) from 2004 till 2011. The development of the transaction cap rates record a higher volatility in the Amsterdam office market when compared to the appraised cap rates (both WOZ cap rates and reported market cap rates by C&W). The average appraised cap rates of different years are very close to each other, which supports the appraisal smoothing theory and serial correlation by appraisers. Whereas, transaction cap rates, show a higher volatility and differ significantly per year. This is also clear when comparing the development of transaction cap rates versus aggregated cap rates reported by real estate agencies (see Figure 8.6). Overall the average transaction cap rate for the period 2004 till 2011 is about 6.00% with a standard deviation of 4.40%. While the average appraised cap rate for this period is 9.30% with a significantly lower deviation of 1.20%. Table 8.2 compares the development of transaction and appraised cap rates per year in Amsterdam. On average, the average cap rates before the crisis is lower than the used cap rates after the financial crisis, with an exception of unweighted average cap rates in 2009. This may be caused by the fact that the majority of transactions in 2009 occurs in Amsterdam
Mean Mean
Transaction cap rate 0,12
Appraised (WOZ) cap rate
0,10
Appraised (market) cap rate (source C&W)
0,08
Error bars: 95% CI
0,06 0,04 0,02 0,00
2004
2005
2006
2007
2008
SaleYear Transaction Year
2009
2010
2011
Figure 8.6, cap rate development (both appraised and transaction) from 2004 till 2011 in Amsterdam
C h a p t e r 8: E m p i r i c a l F i n d i n g s
73
Center (8 transactions), which due to the prime location, has a lower associated risk for property investment, thus lower cape rates are recorded.
Table 8.2, development of transaction and appraised cap rates per year in Amsterdam from 2004 till 2011
2004 2005 2006 2007 2008 2009 2010 2011
Total /Ave.
Table 8.3 shows the average cap rates (both transaction and appraised) per Amsterdam district from 2004 to 2011. In general, Amsterdam South records the lowest transaction cap rates (ave. of 4.10%), as well as lowest standard deviation (2.50%), which is an excepted result due to the existence of the main business district (the Zuidas) in this district. The Amsterdam district of Westpoort has
15
9
15
24
20
19
8
11
121
8.30
5.50
4.60
5.10
6.80
4.80
6.10
7.70
6.00
5.70
2.10
3.40
4.00
4.35
2.00
2.70
4.60
5.20
11.00
9.00
5.80
6.20
9.20
7.60
9.40
10.75
6.80
Std. Deviation
4.80
4.50
2.20
2.60
5.20
5.70
4.00
4.60
4.40
Mean
9.20
8.80
9.00
8.60
9.20
10.00
10.20
9.60
9.30
8.80
8.00
8.50
8.20
8.80
9.10
9.30
8.80
9.00
9.60
9.60
9.40
9.10
9.60
11.00
11.20
10.30
9.50
0.70
1.10
0.80
1.00
0.80
1.90
1.10
1.10
1.20
Transactions # Mean Lower
Transaction
95% Confidence
Bound
cap rate %
Interval for Mean
Upper Bound
Lower
Appraised cap 95% Confidence
Bound
rate %
Upper
Interval for Mean
Bound
New-West
North
East
West
Westpoort
South
South-East
Total /Ave.
Table 8.3, development of transaction and appraised cap rates per district in Amsterdam from 2004 to 2011
Center
Std. Deviation
45
5
9
7
19
4
23
9
121
4.95
7.40
12.20
6.00
5.40
13.20
4.10
7.30
6.00
3.90
6.10
8.80
1.30
4.00
-3.00
3.00
3.40
5.20
6.00
8.80
15.70
10.80
6.70
30.40
5.20
11.20
6.80
Std. Deviation
3.40
1.10
4.50
5.20
2.70
10.20
2.50
5.10
4.40
Mean
9.00
9.25
10.40
10.30
9.10
10.30
8.60
10.20
9.30
8.60
8.90
9.80
9.80
8.80
9.40
8.20
9.50
9.00
9.50
9.60
10.90
10.8
9.40
11.20
9.10
10.90
9.50
1.40
0.40
0.80
0.50
0.60
0.50
1.10
0.90
1.20
Transactions # Mean Lower
Transaction
95% Confidence
Bound
cap rate %
Interval for Mean
Upper Bound
Lower
Appraised cap 95% Confidence
Bound
rate %
Upper
Interval for Mean
Bound Std. Deviation
74
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
the highest transaction cap rates of 13.20 % with a high standard deviation of 10.20%. This may be explained by the spatial differences in this area, since the areas closer to the train station (Amsterdam Sloterdijk), are the best location in this area, and by getting further than the train station, the quality of the office location is enormously reduced. Table 8.4 shows that the average transaction cap rates per year per city district in Amsterdam. This tables shows that there was an enormous increase in the used cap rates in 2008. Especially in the Center, North and South-East districts, which emphasize on the uncertain and risky investment environment just after the financial crisis in the Amsterdam office market. Whereas, a slight increase in appraised cap rates (see Table 8.5), occurs merely in 2009. Considering the fact that the used WOZ cap rates in this research were those of one year after the sale transactions, the results indicate that appraised cap rates (WOZ cap rates) are in fact lagged two years behind the transaction cap rates. Ave.
Transaction year
transaction
City districts
district
2004
2005
2006
2007
2008
2009
2010
2011
Centre
4.30
6.40
2.90
4.50
7.80
3.95
4.60
2.70
4.95
New-West
9.10
-
7.90
6.70
6.70
-
-
-
7.40
North
14.30
8.00
7.80
-
20.40
8.30
14.90
-
12.20
East
-
-
5.30
-
3.15
2.10
-
11.65
6.00
West
7.70
4.90
4.35
6.40
2.10
3.30
8.80
-
5.40
Westpoort
16.70
-
-
8.80
-
13.65
-
-
13.20
South
4.30
2.60
3.90
4.70
4.00
1.90
-
6.20
4.10
South-East
-
-
-
2.70
11.50
6.40
3.30
4.00
7.30
8.30
5.50
4.60
5.10
6.80
4.80
6.10
7.70
6.00
cap rate %
Total ave. per year
Ave.
Transaction year
appraised
Total ave. per district
2004
2005
2006
2007
2008
2009
2010
2011
Centre
8.90
9.00
8.30
8.60
8.30
9.90
9.80
9.30
9.00
New-West
10.00
-
9.10
9.10
9.10
-
-
-
9.25
North
10.00
10.00
10.00
-
10.00
11.10
12.50
-
10.40
East
-
-
10.00
-
10.00
11.10
-
10.55
10.30
West
8.60
8.30
9.55
8.80
9.10
10.00
9.10
-
9.10
Westpoort
10.00
-
-
10.00
-
10.55
-
-
10.30
South
9.10
8.00
8.30
8.20
8.70
9.20
-
9.00
8.60
South-East
-
-
-
8.30
10.00
11.10
10.55
11.10
10.20
9.20
8.80
9.00
8.60
9.20
10.00
10.20
9.60
9.30
cap rate %
City districts
Total ave. per
Total ave. per year
Table 8.4, average transaction cap rates per year per city district in Amsterdam from 2004 to 2011
Table 8.5, average appraised (WOZ) cap rates per year per city district in Amsterdam from 2004 to 2011
75
Amsterdam Submarkets
Figure 8.7 shows the development of the cap rates (both transaction and WOZ cap rate) in different Center New-West years in Amsterdam, which clearly emphasizes on the higher volatility and sensitivity of transaction North cap East rates to different submarkets in Amsterdam, and smoothing in appraised cap rates.
Amsterdam Subma
0,20
Center New-Wes North East West WestPoor South Southeast Center New-Wes North East West WestPoor South Southeast
0,18 0,16
Mean TP_CapRate
Mean TP_CapRate
Figure 8.7, cap rate development (on the top transaction cap rates, at the bottom appraised cap rates) in Amsterdam districts from 2004 to 2011
West WestPoort South Amsterdam districts Amsterdam Submarkets Southeast Center Center New-West New-West North North East East WestWest WestPoort WestPoort South South Southeast Southeast Center New-West North East West WestPoort South Southeast
0,14 0,12 0,10 0,08 0,06 0,04 0,02
2010
2011
0,00 2004
2005
2006
2007
2008
2009
2010
2011
YearOfSaleYear Transaction Amsterdam Submarkets
0,20
Center New-West North East West WestPoort South Southeast Center New-West North East West WestPoort South Southeast
0,18
2011
0,16
Mean WOZ_CapRate
2010
Mean WOZ_CapRate
9
C h a p t e r 8: E m p i r i c a l F i n d i n g s
0,14 0,12 0,10 0,08 0,06 0,04 0,02 0,00 2004
2005
2006
2007
2008
Transaction Year YearOfSale
2009
2010
2011
8.2 Hypothesis 1 | Overstated and smoothed appraised cap rates The result of the total variance test rejects Hypothesis 1. Hypothesis 1 states that the appraised cap rates are lower than the transaction cap rates, depending on the situation on the market (declining and rising market). The results show that not only appraised cap rates in Amsterdam are not lower than transaction cap rates, but also their relation remains the same for the period before and after the financial crisis. Unlike the pre assumption of the appraised cap rates being lower than the transaction cap rates, the overstated and smoothed appraised cap rates vastly dominate the Amsterdam office market from 2004 till 2011. On average, for the period of 2004 till 2011, appraised cap rates are around 50% higher than the transaction cap rates, considering the average absolute differences between them. However, the average directional differences between the appraised and transaction cap rates indicate a slightly lower difference of around - 40%. Nevertheless, this indicator shows a clear tendency for appraised cap rates to consistently overstate the transaction cap rates, thus understating the property values (see Figure 8.8)
76
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
DifferencesDirectionalCaps Cap rate directional DifferencesAbsoluteCaps
1,20 1,00
differences
differences Error Cap Bars:rate 95%absolute CI
0,80
Error bars: 95% CI
0,60
Mean Mean
0,40 0,20
Figure 8.8, the average directional and absolute differences between transaction and appraised cap rates per year
0,00 -0,20 -0,40 -0,60 -0,80 -1,00
2004
2005
2006
2007
2008
2009
2010
2011
SaleYear Transaction Year
These differences are the highest in 2008 (absolute differences of 51.20%) and 2009 (absolute differences of 74.00%). This can be explained due to the fact that WOZ cap rates are lagged and therefore the differences becomes higher for those years (see Table 8.6) Even though there is a general tendency for appraised cap rates to consistently overstate the transaction cap rates (since the absolute differences between appraised and transaction cap rates remain almost the same throughout time), in the Amsterdam district of North and Westport, the appraised cap rates are lower than transaction cap rates. As a result there is a positive amount of directional differences in these districts (see Table 8.7). The average differences of around 50% (absolute) are very high considering the fact that cap rates are the key input variables for the direct capitalization model, where determining the property value is very sensitive to cap rate changes. There are a couple of explanations for such enormous differences between the transaction and appraised cap rates. The first explanation is that WOZ cap rates are used to determine the property value. The purpose of tax (WOZ-Values), is different than that of appraised cap rates (property sale or finance). This is also obvious when looking at Figure 8.6 in § 8.1.4, where the WOZ cap rates are not only higher than the transaction cap rates, but also on average 30% higher than aggregated appraised cap rates reported in market reports such as C&W. The second explanation is that despite the fact that in both appraisal processes, the value should be defined as the fair market value, and follow the RICS appraisal and valuation standards, there are however many restrictions (fictions) related to the appraisal for the purpose of the tax (WOZ values). The major contrary assumption is that when defining the assessed value (WOZ-Values), the property is considered as if vacant, which is far from reality in many cases. Office properties are barley sold as vacant, and when that happens, there is a huge decrease in the property value.
Page 1
C h a p t e r 8: E m p i r i c a l F i n d i n g s
2004
2005
2006
2007
2008
2009
2010
2011
Total /Ave.
15
9
15
24
20
19
8
11
121
-11.20
-40.00
-49.70
-45.90
- 27.45
- 57.60
- 41.50
- 20.30
-37.80
-36.70
-72.10
-61.90
- 58.70
- 51.80
- 84.00
- 68.70
- 51.10
-45.70
14.40
-7.80
-37.50
- 33.10
- 3.10
- 31.30
- 14.30
105.60
-29.90
Std. Deviation
46.20
41.80
22.00
30.30
52.00
54.65
32.50
459.10
44.00
Mean
41.40
50.20
49.70
47.20
51.20
74.00
46.40
44.00
51.50
30.00
29.80
37.50
35.30
38.45
61.00
26.50
30.00
46.80
52.90
70.60
61.90
59.10
64.00
86.90
66.30
58.00
56.30
20.60
26.50
22.00
28.10
27.20
25.60
23.80
20.70
26.50
Transactions # Mean 95%
Lower
Directional
Confidence
Bound
differences %
Interval for
Upper
Mean
Bound
95%
Lower
Absolute
Confidence
Bound
differences %
Interval for
Upper
Mean
Bound
New-West
North
East
West
Westpoort
South
South-East
Table 8.7, average directional and absolute differences between transaction and appraised cap rates per districts
Centre
Std. Deviation
Total /Ave.
Table 8.6, average directional and absolute differences between transaction and appraised cap rates per year
77
45
5
9
7
19
4
23
9
121
-50.00
-20.20
18.10
-42.40
-40.00
31.60
-52.20
-27.90
-37.80
-60.60
-30.60
-16.40
-85.00
-55.10
-132.10
-64.70
-67.10
-45.70
-39.60
-9.90
52.60
0.10
-24.80
195.20
-39.70
113.10
-29.90
Std. Deviation
34.90
8.30
44.80
46.00
31.40
102.80
28.90
51.00
44.00
Mean
55.60
20.20
38.00
58.10
43.70
79.50
56.20
49.30
51.50
48.10
9.90
16.70
42.90
31.50
-14.50
47.60
28.30
46.80
63.10
30.60
59.10
73.40
56.00
173.40
64.70
70.30
56.30
24.80
8.30
27.60
16.50
25.50
59.00
19.70
273.10
26.50
Transactions # Mean 95%
Lower
Directional
Confidence
Bound
differences %
Interval for
Upper
Mean
Bound
95%
Lower
Absolute
Confidence
Bound
differences %
Interval for
Upper
Mean
Bound
Std. Deviation
78
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
8.3 Hypothesis 2 | forward looking investors versus backward looking Appraisers Based on the regression outputs, Hypothesis 2 is not a valid statement, since it shows that investors in the Amsterdam office market are more concerned with the cross-sectional and property specific variations in cap rates, whereas the appraiser is more concerned with the time-series variations in cap rates. However, the results show a clear difference between appraisers and investors, when determining the cap rates. The regression output of the transaction cap rate model shows an R2 value of .51, which indicates that the included variables (in total 13 variables) in the regression account for 51% of the variation in transaction cap rates (see Appendix 3). The expected impact of the independent variables on cap rats shown in Table 8.8 (see § 7.3.3.2 for the complete list of the used variables in the regression), are in accordance with the previous research mentioned in Part II (see also § 7.3.1). The context variables used in the regression model are office employment index (as a proxy for the macroeconomic conditions) and office investment volume as a ratio of the total investment in the commercial real estate in the Dutch real estate market (a proxy of the capital market). An increase in the office employment caused a decrease in the used cap rates by the investors, which is explained by the demand increase for the office space, thereby decreasing the vacancy and ultimately reducing the risk associated with the investment. Even though the explanatory power of this variable is less than 1%, the impact of this variable is significant at 0.05 level. The office investment volume ratio is negatively and significantly (sig. at 0.05 level) related to the transaction cap rates. An increase of one unit in the office investment ratio results in a cap rate decrease of around 0.05 units (5%). This indicates that the availability of capital to invest in the office segment, has a significant influence on the perception of the risk by the investors in this segment. Even though the magnitude of the used context variables is significant, all in all they have a very low explanatory power of less than 1% of the variations in the transaction cap rate model. This while the explanatory power of location variables (Amsterdam districts) is 30%, which indicates that 30% of the variation in transaction cap rates depends on where the property is located in the Amsterdam office market. The fixed effect of city districts matches with the market expectation for these locations. Properties located in Amsterdam South (DD7), decreases the transaction cap rate with 3%, whereas a property situated in Amsterdam Westpoort (DD6) increases transaction cap rate with about 9%. The size of a building is positively related to the transaction cap rate. An increase of one unit results in a cap rate increase of around 0.006 units. Even though, this is not a very substantial explanatory power, the magnitude of this variable is very significant (sig. at 0.01 level). This is explained by the fact that the bigger the size of a property, the higher the risk associated with the investment, since it reduces the liquidity of property investment. The rent ratio is positively and significantly (sig. at 0.01 level) related to the transaction cap rate which explains the forward looking attitude of the investors, who interpret the higher rent level
C h a p t e r 8: E m p i r i c a l F i n d i n g s
Table 8.8, transaction cap rates, regression output
79
Coefficients transaction cap rate as dependent variable Unstandardized Coefficients B
Std. Error
(Constant)
,389
,156
OfficeJobs
-,002
,001
OfficeInvest -,053 DD2
,045
DD3
Standardized Coefficients Beta
Unstandardized Coefficients t
Sig.
B
Std. Error
2,493
,014
,119
,073
-,175
-2,159
,033
,022
-,173
-2,370
,020
,018
,203
2,552
,012
,038
,018
,076
,012
,456
6,175
,000
,074
DD4
,030
,013
,159
2,215
,029
DD6
,088
,018
,359
4,994
,000
DD7
-,034
,009
-,307
-3,706
DD8
,040
,018
,240
LnLFAm2
,006
,002
,197
RentR
,031
,008
LnAge
,007
LnAfstDBZ_ IC_Station
-,021
Asbest
,024
Standardized Coefficients Beta
t
Sig.
1,627
,107
,173
2,132
,035
,013
,442
5,844
,000
,025
,014
,133
1,772
,079
,088
,018
,358
4,942
,000
,000
-,035
,009
-,314
-3,698
,000
2,184
,031
,035
,019
,207
1,841
,069
2,721
,008
,007
,002
,209
2,801
,006
,296
3,999
,000
,028
,008
,272
3,367
,001
,002
,292
3,490
,001
,007
,002
,273
3,247
,002
,010
-,247
-2,175
,032
-,020
,010
-,238
-2,024
,046
,012
,145
1,989
,049
,023
,013
,139
1,809
,073
DY2004
,025
,013
,191
1,918
,058
DY2005
,009
,014
,053
,613
,541
DY2006
-,011
,012
-,082
-,911
,364
DY2007
-,010
,011
-,094
-,960
,340
DY2009
-,008
,012
-,069
-,720
,473
DY2010
-,010
,014
-,056
-,691
,491
DY2011
,008
,013
,051
,612
,542
relative to the market rent as a temporary situation and thus, they applied a higher cap rate (and vice versa). The property age as it was expected, is positively and significantly associated with the cap rate. However, it has a low explanatory power of cap rate variation, since it merely contributes less than 1% to the cap rate model. The distance to the station as a proxy for accessibility is negatively and significantly (sig. at 0.05 level) related to cap rates. The resulted sign for this variable is not in accordance with the expectation of this research. This may be explained by the fact that there are other measures to gauge the accessibility of a property, such as the distance to other public transports (bus, metro, etc.) or accessibility by car, which due the sample size was not included in the regression. As a result the explanatory power of this variable is not only minor (less than 1%), but also is not consistent with previous research expectations. The existence of asbestos in the building, as a proxy for the environmental condition, is positively and significantly (sig. at 0.05 level) associated with transaction cap rate variations. Properties with
80
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
asbestos, required a higher cap rate (about 2%), in comparison to the ones without. This indicates the magnitude of environmental conditions influences on the perception of the risk. When year dummies are used in the transaction cap rate regression model, the value of R2 improved to .54, showing an explanatory power of independent variables of 54% (see Appendix 3). Capturing the time-series variations in cap rates, by considering the fix effect of each year on transaction cap rates, improves the fitness of the model and better explains the time-series variations in cap rates than the context variables. To be precise, the time varying variables explain about 9% of changes in transaction cap rates, which is a higher than the explanatory power of context variables (2%) in the first transaction cap rate regression model. This is while the locational dummies (Amsterdam districts) account for 30% of the variation in cap rates (the same as the first model), and property specific variables account for 15% of this variation, which is slightly lower than the first model. Nevertheless, the signs of the beta coefficients and their significance for all independent variables remain the same in both model specifications (see Table 8.8). The regression output of the appraised cap rate, with the same variable used for the transaction cap rate, shows a lower value of R2 (.40), in comparison to the transaction cap rate model. This indicates that the included variables in the regression account for 40% of the variation in appraised cap rates (see Appendix 3). Context variables (office employment and office investment ratio) used in the appraised cap rate regression model, explain 11% of the variations in cap rates. This is a higher amount in comparison to the explanatory power of these variables in the transaction cap rate model. This is while the location variables and property specific variables count respectively for 23% and 6% of the appraised cap rate variations, which are lowered when compared to their explanatory power in the transaction cap rate model. When year dummies are used in the appraised cap rate regression model, the value of R2 improved to .42, showing an explanatory power of independent variables of 42% (see Appendix 3). By considering the fixed effect of each year on appraised cap rates, the fitness of the model is improved. The time-series variables explain about 18 % of the changes in appraised cap rates, which is higher than the explanatory power of context variables (11%) in the appraised cap rate regression model with context variables. This is while the locational dummies (Amsterdam districts) account for 19% of the variation in cap rates, and property specific variables account for 5% of this variation, which is slightly lower than the first model (appraised cap rate with context variables). Nevertheless, the signs of the beta coefficients and their significance for all independent variables remain the same in both model specifications (see Table 8.9). By comparing the regression output of the transaction cap rates and appraised cap rates, it is clear that appraisers are more concerned with context (macroeconomic and financial risks), and timeseries variations (year dummies) in cap rates than investors. The time-series variables explain 18 % of the changes in appraised cap rates, which is two times higher than the explanatory power of them in the transaction cap rate variations (9%). On the other hand, investors are pertained more with the magnitude and significance of the Amsterdam submarkets (city districts) on cap rates, thus cross sectional variation in cap rates. Even though the appraisers expect the same influence of Amsterdam submarkets on the cap rates, the explanatory power of these submarkets are not substantial. For instance, while the fixed effect of
C h a p t e r 8: E m p i r i c a l F i n d i n g s
Table 8.9, appraised cap rates, regression output
81
Coefficients appraised cap rate as dependent variable Unstandardized Coefficients
Standardized Coefficients
B
Std. Error
Beta
(Constant)
,067
,049
OfficeJobs
,001
,000
,158
Unstandardized Coefficients t
Sig.
B
Std. Error
1,359
,177
,122
,023
1,746
,084
Standardized Coefficients Beta
t
Sig.
5,218
,000
OfficeInvest
-,016
,007
-,183
-2,254
,026
DD2
,007
,006
,113
1,274
,205
,008
,006
,132
1,437
,154
DD3
,016
,004
,347
4,232
,000
,016
,004
,334
3,917
,000
DD4
,012
,004
,234
2,936
,004
,014
,004
,261
3,092
,003
DD6
,013
,006
,188
2,356
,020
,012
,006
,177
2,169
,032
DD7
-,004
,003
-,138
-1,493
,138
-,004
,003
-,133
-1,385
,169
DD8
,014
,006
,302
2,472
,015
,015
,006
,324
2,564
,012
LnLFAm2
,000
,001
,035
,437
,663
,000
,001
,043
,517
,607
RentR
-,006
,002
-,198
-2,407
,018
-,004
,003
-,144
-1,583
,117
LnAge
,001
,001
,170
1,830
,070
,001
,001
,192
2,022
,046
LnAfstDBZ_ IC_Station
-,004
,003
-,174
-1,377
,171
-,005
,003
-,203
-1,531
,129
Asbest
,000
,004
-,003
-,035
,972
,001
,004
,016
,181
,856
DY2004
,002
,004
,057
,511
,610
DY2005
,000
,005
,001
,007
,994
DY2006
-,002
,004
-,044
-,429
,669
DY2007
-,001
,003
-,029
-,263
,793
DY2009
,009
,004
,252
2,330
,022
DY2010
,009
,005
,177
1,924
,057
DY2011
,004
,004
,099
1,043
,299
the South-East district shows an increase of 4% in the cap rates (β = 0.040), appraisers are less concerned with the impact of this location on the applied cap rate (β = 0.006). The weighting method of the investors and appraisers is even more different when looking at the magnitude and explanatory power of the property specific variables. The property specific variables explain only 5% of the variation in appraised cap rates, which is three times lower than the explanatory power of them in the transaction cap rate variations (15%). For example, considering the impact of rent ratio on cap rates, investors are more forward looking and interpret the positive rent ratio as a temporary market event and expect a decrease in the rent values and their returns to their market fundamental. Whereas, appraisers are more backward looking and literally extrapolate the increase in the rent with the decrease in cap rates. Complimentary to this fact, when comparing the current appraised cap rates to the ones of the previous year (a one year lagged appraised cap rate), the coefficient of the lagged cap rates is around 0.80, which shows a substantial effect on the estimation of the appraised cap rates. This emphasizes on the existence of serial correlations and anchoring on the past in appraised based indices (appraisal smoothing). Table 8.10 and 8.11 compare the coefficients of the estimated model for transaction and appraised cap rates.
82
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
Coefficients transaction vs. Appraised cap rates with context variables Transaction Cap Rates B
Std. Error
(Constant)
,389
,156
OfficeJobs
-,002
,001
OfficeInvest
-,053
,022
DD2
,045
DD3
,076
DD4 DD6
Appraised Cap Rates
t
Sig.
B
Std. Error
2,493
,014
,067
,049
-,175
-2,159
,033
,001
,000
-,173
-2,370
,020
-,016
,007
,018
,203
2,552
,012
,007
,012
,456
6,175
,000
,016
,030
,013
,159
2,215
,029
,088
,018
,359
4,994
,000
DD7
-,034
,009
-,307
-3,706
DD8
,040
,018
,240
2,184
LnLFAm2
,006
,002
,197
RentR
,031
,008
,296
LnAge
,007
,002
LnAfstDBZ_IC_Station -,021 Asbest
,024
Beta
Beta
t
Sig.
1,359
,177
,158
1,746
,084
-,183
-2,254
,026
,006
,113
1,274
,205
,004
,347
4,232
,000
,012
,004
,234
2,936
,004
,013
,006
,188
2,356
,020
,000
-,004
,003
-,138
-1,493
,138
,031
,014
,006
,302
2,472
,015
2,721
,008
,000
,001
,035
,437
,663
3,999
,000
-,006
,002
-,198
-2,407
,018
,292
3,490
,001
,001
,001
,170
1,830
,070
,010
-,247
-2,175
,032
-,004
,003
-,174
-1,377
,171
,012
,145
1,989
,049
,000
,004
-,003
-,035
,972
Coefficients transaction vs. Appraised cap rates with year dummies Transaction Cap Rates B
Std. Error
Beta
(Constant)
,119
,073
DD2
,038
,018
,173
DD3
,074
,013
DD4
,025
,014
DD6
,088
DD7
-,035
DD8
Appraised Cap Rates
t
Sig.
B
Std. Error
Beta
t
Sig.
1,627
,107
,122
,023
2,132
,035
,008
,006
,132
5,218
,000
1,437
,154
,442
5,844
,000
,016
,004
,133
1,772
,079
,014
,004
,334
3,917
,000
,261
3,092
,003
,018
,358
4,942
,000
,012
,009
-,314
-3,698
,000
-,004
,006
,177
2,169
,032
,003
-,133
-1,385
,169
,035
,019
,207
1,841
,069
,015
,006
,324
2,564
,012
LnLFAm2
,007
,002
,209
2,801
RentR
,028
,008
,272
3,367
,006
,000
,001
,043
,517
,607
,001
-,004
,003
-,144
-1,583
,117
LnAge
,007
,002
,273
3,247
,002
,001
,001
,192
2,022
,046
LnAfstDBZ_IC_Station -,020
,010
-,238
-2,024
,046
-,005
,003
-,203
-1,531
,129
Asbest
,023
,013
,139
1,809
,073
,001
,004
,016
,181
,856
DY2004
,025
,013
,191
1,918
,058
,002
,004
,057
,511
,610
DY2005
,009
,014
,053
,613
,541
,000
,005
,001
,007
,994
DY2006
-,011
,012
-,082
-,911
,364
-,002
,004
-,044
-,429
,669
DY2007
-,010
,011
-,094
-,960
,340
-,001
,003
-,029
-,263
,793
DY2009
-,008
,012
-,069
-,720
,473
,009
,004
,252
2,330
,022
DY2010
-,010
,014
-,056
-,691
,491
,009
,005
,177
1,924
,057
DY2011
,008
,013
,051
,612
,542
,004
,004
,099
1,043
,299
Table 8.10, comparison between transaction and appraised cap rates, regression output with macroeconomic and financial variables
Table 8.11, comparison between transaction and appraised cap rates, regression output with year dummies
C h a p t e r 8: E m p i r i c a l F i n d i n g s
83
Essentially, when determining cap rates, appraisers put more weight on the cap rate variations over time, as well as what happened in the past, whereas investors are more concerned with the crosssectional variation (different submarkets in Amsterdam), and are pertained with the mean reversion process of the financial and growth variables (e.g. rent ratio). However, the results are not consistent with the previous research in the other countries, where transaction cap rates vary strongly by macro and time-series variables. Location, location and location seems to be a valid statement and the most influential factor in determining cap rates for the investors in the Amsterdam office market.
Figure 8.9 shows the comparison of the explanatory power of context, location, and property related variables, respectively for transaction and appraised cap rates, when added to the regression model. Whereas, Figure 8.10 compares the explanatory power of year dummies (as proxies for time series variations) between transaction and appraised cap rates. Hypothesis 2 | Forward looking investors vs. backward Hypothesis looking 2 | appraisers Forward looking investors vs. b Figure 8.9, (left) explanatory power of context, location and property specific variables respectively in transaction and appraised cap rates regression Figure 8.10, (right) explanatory power of year dummies, location and property specific variables respectively in transaction and appraised cap rates regression
RTP
RV
RTP
step 1
2%
<
11%
step 1
step 2
30%
>
23%
step 3
19%
>
6%
R2
51%
40%
RV
9%
<
18%
step 2
30%
>
19%
step 3
15%
>
5%
Year dummies
R2
54%
40%
20
8.4 Hypothesis 3 | Structural vacancy as a Paradoxical phenomenon The result of the test shows a paradoxical phenomenon, which could not confirm the hypothesis 3. When the property is categorized as structural vacant, the differences between the appraised cap rates and those of transactions decreases (both directional and absolute differences), in comparison to the situation when a property has a market conformed vacancy level. This shows that when there is structural vacancy risk, both appraisers and investors weighted the vacancy risk roughly the same, but when it was market conformed vacancy ratio the appraisers put less emphasis on the vacancy and the differences becomes greater. Figure 8.11 illustrates the linear trend of the mean in directional differences and Figure 8.12 shows the linear trend of the mean in absolute differences in cap rates compared to the subgroups of vacancy risk variable. As mentioned in §8.3, this can be explained by the fact that, the base assumptions in determining the assessed value in the Netherlands is that the property should be valued as if vacant. An extra risk will only incorporate the calculation of the WOZ-value, when there is a high structural vacancy in a neighborhood. This can explain that when there is a structural vacancy level (since the appraisers should not follow the friction of appraising as if vacant), the differences between the appraised and transaction cap rates are reduced.
84
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
In general, the result of this test is not statistically significant, due to the unbalanced distribution of the samples in each vacancy risk category. Neither the Post hoc test records any significant mean variances of cap rates differences among the different vacancy risk categories of market conformed, potential structural, and structural vacancy risks. Table 8.12 summarizes the mean differences in directional and absolute cap rate differences compared to different categories of vacancy risk.
,70
Mean of DifferencesAbsoluteCaps
Mean of DifferencesDirectionalCaps
,00
-,20
-,40
,60
,50
,40
-,60
,30 Market conformed
Potential structural
structural
Market conformed
Vacancy Risk VacancyRisk
Potential structural
Figure 8.11, the linear trend of the mean in cap rate directional differences compared to vacancy risks
Figure 8.12, the linear trend of the mean in cap rate absolute differences compared to vacancy risks
Directional differences vs. vacancy risks 95% Confidence Interval for Mean Vacancy Risks Market conformed
N
Mean
Std. Deviation
Std. Error
Lower Bound
Upper Bound
Minimum
Maximum
115
-37,8%
44,2%
4,1%
-46,0%
-29,7%
-96,8%
155,2%
-100,3%
-28,8%
-81,1%
-55,6%
Page 1
Potential structural
3
-64,5%
14,4%
8,3%
Structural
3
-9,6%
49,7%
28,7%
-133,0%
113,9%
-66,3%
26,5%
121
-37,8%
44,0%
4,0%
-45,7%
-29,9%
-96,8%
155,2%
Minimum
Maximum
Total
Absolute differences vs. vacancy risks 95% Confidence Interval for Mean Vacancy Risks
N
Mean
Std. Deviation
Std. Error
Lower Bound
Upper Bound
Market conformed
115
51,6%
26,6%
2,5%
46,7%
56,6%
2,7%
155,2%
Potential structural
3
64,5%
14,4%
8,3%
28,8%
100,3%
55,6%
81,1%
Structural Total
structural
Vacancy Risk VacancyRisk
3
34,6%
28,5%
16,5%
-36,2%
105,4%
11,0%
66,3%
121
51,5%
26,5%
2,4%
46,8%
56,3%
2,7%
155,2%
Table 8.12, the mean differences in directional and absolute differences compared to different categories of vacancy risk
Page 1
Part IV: conclusions
Chapter 9: Conclusion Chapter 10: recommendation for further research
9
Conclusion
C h a p t e r 9: C o n c l u s i o n
87
Chapter 9: Conclusion This research examines the main cap rate determinants for the Amsterdam office market, for appraisers and investors respectively. Furthermore, it measures the differences between the appraised and transaction cap rates in order to gauge whether their differences are caused by variations in vacancy rate, or whether they are due to differences in the pricing mechanism between appraisers and investors. This chapter addresses the main findings of this research by answering the two main research questions: 1. To what extent do appraised cap rates correspond with transaction cap rates? 2. Can the differences between appraised cap rates and transaction cap rates be explained (partly) by structural vacancy risk? These questions are answered based on four major themes: cap rate determinants, valuation accuracy, the variation in pricing mechanism of appraisers and investors, and valuation accuracy of vacant office properties.
9.1 Cap rate determinants, the applicability of the new dynamic Cap rate model Literature on cap rate determinants show a trend from the dominancy of macro economic and time-series variations in cap rates, towards the state where micro-level and property-specific factors become more prominent in explaining the driving forces of cap rates. Due to an increase of the investors’ awareness and interest in the quality and characteristics of properties within investment funds, and also the development of ‘Big Data’ and availability of large volumes of information (structured and/or unstructured) regarding the property characteristics, this thesis introduces a more dynamic and applicable model specification to determine cap rates. This model decompounds the risk premium component achieved on investing in office market, to location, the property-itself and the user (office-user) attributes. Furthermore, it develops these components by categorizing them into background context, location, property, and office-users determinants of cap rates. This model includes not only macro-economic, time varies and cross sectional variation in cap rates, but also transforms a partly qualitative investment decision-making procedure and assessment of office properties to a more objective and quantitative method, by including variables pertaining the property itself (property-specific characteristics) and elements that are associated with the officeusers (tenants). This provides a better specification of the risk premium achieved for office real estate which leads to more careful measuring of the risk associated with an investment, especially in the problematic situation of the Dutch office market. In addition, this research defines and measures the specific vacancy risk, as one of the major problems in the Dutch market, by considering the vacancy rate at the property level relative to its direct surroundings. This enables assessing the value of the vacant properties based on the different vacancy risk associated with the property introduced in this thesis, namely market conformed, potential structural and structural vacancy risk. This provides a comprehensive tool for investors as well as appraisers to determine and estimate cap rates, in a more objective and quantitative process, specifically regarding the estimation of the vacancy risk. This is contrary to the current qualitative assessment of the risk, to determine the
88
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
property value which is prone to some degree of uncertainty, and leads to an inaccuracy in valuation.
9.2 valuation accuracy The findings from this research emphasize on consistently overstated and smoothed appraised cap rates in the Amsterdam office market from 2004 till 2011, compared to the transaction cap rates. The transaction cap rates are on average lower and more volatile compared to appraised cap rates. This indicates that appraised cap rates (WOZ-values), are understating the property values compared to their actual market value (transaction prices). An average deviation of 50% is the result of the differences between appraised cap rates and the transaction cap rates, which corresponds with the same deviation between appraisal value and transaction prices. These differences can be explained by the fact that the used data are the appraisal data for the tax purposes, and not for the commercial purposes (sale or finance). Even though this amount should correspond with the fair market value of the property, there are many restrictions (fictions) related to the tax appraisal such as the fiction of a property being appraised as if vacant, which is far from reality in many cases. Finally, there is also various evidence of a mismatch between the developments of transaction cap rates versus aggregated cap rates reported by real estate agencies. Even though the overall development of these cap rates are the same, transaction cap rates are more volatile while the aggregated cap rates reported by real estate agencies are smoothed. This emphasizes on the fact that appraisal indices do not completely record the true volatility of the office market through time.
9.3 appraised cap rates versus transaction cap rates By comparing the regression output of the transaction cap rates and appraised cap rates, it is clear that appraisers are more concerned with context (macroeconomic and financial risks), and timeseries variation (year dummies) in cap rates. Whereas investors are more concerned with crosssectional and micro-level and property related variations in cap rates. The empirical results of this research based on context, location and property indicate that location is a key factor in determining cap rates in the Amsterdam office market. Cap rates vary persistently across markets due to differences in fixed market characteristics which influence the perception of investors in predicting the rental growth and associated risk. Nevertheless, the magnitude and significance of the Amsterdam submarkets (city districts) on cap rates are higher for the investors, compared to appraisers. Regarding the property related variables, cap rates consist of components which are shaped by historical trends in rental growth in the Amsterdam submarkets. These relationships with historical trends indicate that appraisers strongly rely on past evidence in the market, rather than having a forward looking perspective. This phenomenon is consistent with the arguments regarding anchoring behavior of appraisers. On the other hand, investors consider the mean reversion of real rents in their rental growth expectation, since they use higher cap rates when compared to rental cyclical peaks. These all indicate that appraisers are prominently more backward looking, and more concerned with changes in macroeconomics and time varying aspects, whereas investors are more forward looking, and pay close attention to the submarket within the Amsterdam office market.
C h a p t e r 9: C o n c l u s i o n
89
9.4 myth of vacancy and accuracy The differences between appraised cap rates and transaction cap rates have a negative correlation to its vacancy risks, namely market conformed, potential structural, and structural vacancy risk. The empirical results show that when there is structural vacancy level, both appraisers and investors weighted the vacancy risk similarly. Whereas, when there is a market conformed vacancy ratio, the appraisers put less emphasis on the vacancy risk and the differences becomes greater. The reliability of the statistical results are very low. Since less than 5% of the transactions studied, contain properties with a potential structural vacancy risk or structural vacancy risk.
9.5 overall conclusion The research findings indicate a transaction-valuation difference in cap rates as well as contributing factors to these differences. The transaction cap rates are determined by mostly the location and property related variables, whereas macroeconomic and context variables determine the appraised cap rates vastly in the Amsterdam Office market. It also shows that selling an office building, being structurally vacant, seems to be an exceptional situation in the Amsterdam office market. Since there are not that many of these being sold in Amsterdam. One can interpret this as a possibility of asking a higher value than an investor is willing to pay. Unfortunately, no concrete conclusion can be made due to the lack of market evidence and available data to further investigate this phenomenon.
10
recommendation for further research
C h a p t e r 10: r e c o m m e n d at i o n f o r f u r t h e r r e s e a r c h
91
Chapter 10: recommendation for further research Based upon the results of this study this chapter focuses on a couple of suggestions and recommendations for future research. In order to improve the degree of comparability between transaction and appraised cap rates, further research can be done while using appraised value for the commercial purposes (sale or finance), rather the WOZ values which are appraised for tax purposes. The use of transaction cap rates which are not calculated ex-post, can reduce the number of assumptions to estimate the transaction cap rates. This can lead to better research results, since it enhances the objectivity and reduces biases. In addition, this study focuses merely on the WOZ-objects which are the properties with a single tenant and a single owner. The research results can be improved in case that the multitenant and multi-owner properties are also included in the database. Furthermore, the degree of comparability between transaction and appraised cap rates can be improved when including the transaction of other cities in the Netherlands. This gives an opportunity to compare cross-sectional differences among various office markets (e.g., Rotterdam, Utrecht, etc.) in the Netherlands. Finally, these all enhance the chance of analyzing properties with different vacancy rates (including those with potential structural and structural vacancy levels).
References
References
93
References Ambrose, B. W., & Nourse, H. O. (1993). Factors influencing capitalization rates. Journal of Real Estate Research, 8(2), 221-237. Australian Government. (2004). Adaptive Reuse. Canberra: Commonwealth of Australia Retrieved from http://www.environment.gov.au/system/files/resources/3845f27a-ad2c-4d40-882718c643c7adcd/files/adaptive-reuse.pdf. Baum, A. (1993). Quality, depreciation, and property performance. Journal of Real Estate Research, 8(4), 541565. Baum, A., Crosby, N., Gallimore, P., McAllister, P., & Gray, A. (2000). The influence of valuers and valuations on the workings of the commercial property investment market. Reading and Nottingham Trent Universities. Boots, R. (2014). The in-transparency of the Amsterdam office market, the underlying incentive and effective rental price development. (Master), Delft University of Technology, Delft. Brueggeman, W. B., & Fisher, J. D. (2011). Real Estate Finance and Investment. New York, USA. Bryman, A. (2012). Social research methods: Oxford university press. Bullen, P., & Love, P. (2011). Factors influencing the adaptive re-use of buildings. Journal of Engineering, Design and Technology, 9(1), 32-46. CBRE. (2012). Rotterdam Central Business District: an insider’s view: CBRE. CBRE. (2013). Making a difference: opportunities in the Dutch office market: CBRE. Chervachidze, S., & Wheaton, W. (2013). What Determined the Great Cap Rate Compression of 2000–2007, and the Dramatic Reversal During the 2008–2009 Financial Crisis? The Journal of Real Estate Finance and Economics, 46(2), 208-231. Colliers. (2014). Office Market Report H1 2014 | The Netherlands: Colliers International. Collings, S. (2012). IAS 40 Investment Properties. Retrieved December 25, 2014, from http://www.globalaccountantweb.com/ias-40-investment-property/ Cullen, I. (1994). The accuracy of valuations revisited. Paper presented at the The Cutting Edge Conference 1994. Cuppen, J. (2011). Private en institutionele beleggers samen in één privaat vastgoedfonds? (MRE), Amsterdam School of Real Estate. Damodaran, A. (2012). Investment valuation: Tools and techniques for determining the value of any asset: John Wiley & Sons. De Rode Olifant. (2011). Geschiedenis. Retrieved May 11, 2014, from http://www.derodeolifant.eu/ De Roo, J. (2014). De WOZ waarde is conservatief, een onderzoek naar de WOZ waarde van kantoren in Amsterdam. (MSRE), Amsterdam School of Real Estate. Deloitte. (2014a). IAS 16 - Property, Plant and Equipment. International Accounting Standards. Retrieved December 25, 2014, from http://www.iasplus.com/en/standards/ias/ias16 Deloitte. (2014b). IAS 36 - Impairment of Assets. International Accounting Standards. Retrieved December 25, 2014, from http://www.iasplus.com/en/standards/ias/ias36 Deloitte. (2014c). IAS 40 - Investment Property. International Accounting Standards. Retrieved December 26, 2014, from http://www.iasplus.com/en/standards/ias/ias40 Deloitte. (2014d). International tax: Netherlands highlights 2014. Retrieved November 19, 2014, from http://www2.deloitte.com/content/dam/Deloitte/global/Documents/Tax/dttl-taxnetherlandshighlights-2014.pdf DTZ Zadelhoff. (2014a). Kansrijke kerngebieden: Rotterdam Central District. Rotterdam: DTZ Zadelhoff. DTZ Zadelhoff. (2014b). Nederland compleet: Factsheets kantoren- en bedrijfsruimtemarkt | medio 2014. Amsterdam: DTZ Zadelhoff. Englund, P., Gunnelin, Å., Hoesli, M., & Söderberg, B. (2004). Implicit forward rents as predictors of future rents. Real Estate Economics, 32(2), 183-215. Fama, E. F., & French, K. R. (2004). The capital asset pricing model: Theory and evidence. Journal of Economic Perspectives, 18, 25-46. FGH Bank. (2015). The power of a different perspective FGH Real Estate Report 2015. Utrecht: FGH Bank. Field, A. (2013). Discovering statistics using IBM SPSS statistics: Sage. Fisher, J. D., Miles, M. E., & Webb, R. B. (1999). How reliable are commercial appraisals? Another look. Real Estate Finance, 16, 9-15. Francke, M. K. (2014). Hedonic Price Models. (Course reader in the topic Real estate valuation and investments). Amsterdam: University of Amsterdam (UvA) Business School.
94
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
French, N., & Gabrielli, L. (2004). The uncertainty of valuation. Journal of Property Investment & Finance, 22(6), 484-500. Froland, C. (1987). What determines cap rates on real estate? The Journal of Portfolio Management, 13(4), 7782. Geltner, D., Miller, N. G., Clayton, J., & Eichholtz, P. (2014). Commercial real estate analysis and investments (3rd ed.). USA: OnCourse Learning. Gordon, M. J. (1962). The investment, financing, and valuation of the corporation: RD Irwin. Haigh, M. S., & List, J. A. (2005). Do professional traders exhibit myopic loss aversion? An experimental analysis. The Journal of Finance, 60(1), 523-534. Harding, B. (2012). The profitable influence of lease incentives for new office developments. (Master), Delft University of Technology, Delft. Hendershott, P. H. (1996). Valuing properties when comparable sales do not exist and the market is in disequilibrium. Journal of Property Research, 13(1), 57-66. Hillier, D., Ross, S. A., Westerfield, R. W., Jaffe, J., & Jordan, B. D. (2013). Corporate finance: McGraw-Hill. Hoesli, M., & Chaney, A. (2014). Transaction-based and appraisal-based capitalization rate determinants. Real Estate Review. Hoesli, M., Jani, E., & Bender, A. (2006). Monte Carlo simulations for real estate valuation. Journal of Property Investment & Finance, 24(2), 102-122. Hordijk, A. (2005). Valuation and Construction Issues in Real Estate Indices: Europe Real Estate Publishers B.V. Hordijk, A., & Van de Ridder, W. (2005). Valuation model uniformity and consistency in real estate indices. Journal of Property Investment & Finance, 23(2), 165-181. doi: doi:10.1108/14635780510584355 IPD. (2014). IPD Real Estate Index Analyses: Valuation and Sale Price Comparison Report: IPD. IVSC. (2014). International valuation glossary. Retrieved October 13, 2014, from http://www.ivsc.org/glossary Jud, D. G., & Winkler, D. T. (1995). The capitalization rate of commercial properties and market returns. Journal of Real Estate Research, 10(5), 509-518. Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Econometrica: Journal of the Econometric Society, 263-291. Keeris, W. (2007). Gelaagdheid in leegstand. In Van der Voordt T. (Ed.), Transformatie van kantoorgebouwen. Thema’s, actoren, instrumenten en projecten (pp. 204-211). Rotterdam: 010 Publishers. Krumm, P. J. (2001). History of real estate management from a corporate perspective. Facilities, 19(7/8), 276286. Kumar, R. (2014). Research Methodology: A Step-by-Step Guide for Beginners (4th ed.): SAGE Publications Ltd. Langston, C., Wong, F. K., Hui, E. C., & Shen, L.-Y. (2008). Strategic assessment of building adaptive reuse opportunities in Hong Kong. Building and Environment, 43(10), 1709-1718. Lizieri, C., & Venmore-Rowland, P. (1991). Valuation accuracy: a contribution to the debate. Journal of Property Research, 8(2), 115-122. Lokhorst, J., Remoy, H., & Koppels, P. (2013). Verborgen leegstand: mismatch tussen vraag en aanbod. Real Estate Research Quarterly, 12 (3), 2013. Lusht, K. M. (2012). Real estate valuation: principles and applications. USA: KML Publishing. McAllister, P. (1995). Valuation accuracy: a contribution to the debate. Journal of Property Research, 12(3), 203216. McAllister, P., Baum, A., Crosby, N., Gallimore, P., & Gray, A. (2003). Appraiser behaviour and appraisal smoothing: some qualitative and quantitative evidence. Journal of Property Research, 20(3), 261-280. McDonald, J. F., & Dermisi, S. (2008). Capitalization rates, discount rates, and net operating income: The case of downtown Chicago office buildings. Journal of Real Estate Portfolio Management, 14(4), 363-374. Netzell, O. (2009). A study of micro-level variation in appraisal-based capitalisation rates. Journal of Property Research, 26(3), 235-263. NVM. (2013). The Dutch property market in focus - facts and figures 2013. Nieuwegein: NVM. Rabianski, J. S. (2002). Vacancy in market analysis and valuation. Appraisal Journal, 70(2), 191-199. Remøy, H. (2007). De markt voor transformatie van kantoren tot woningen. In Van der Voordt T. (Ed.), Transformatie van kantoorgebouwen: thema's, actoren, instrumenten en projecten (pp. 194 - 203). Rotterdam: 010 Publishers. Remøy, H. (2010). Out of office a study of the cause of office vacancy and transformation as a means to cope and prevent. (Diss Delft 2010), IOS, Amsterdam. Retrieved from
References
95
http://resolver.tudelft.nl/uuid:9c24b779-1c61-4a88-921a-04d3e12a8e46 Remøy, H., Koppels, P., Van Oel, C., & De Jonge, H. (2007). Characteristics of vacant offices: A Delphi-approach. Paper presented at the ENHR International Conference. Remøy, H., Schalekamp, M., & Hobma, F. (2009). Transformatie van kantoorterreinen, een stappenplan. Real Estate Research Quarterly 8(4), 22-27. Remøy, H., & Van der Voordt, T. (2007). A new life: conversion of vacant office buildings into housing. Facilities, 25(3/4), 88-103. Remøy, H., & Van der Voordt, T. (2011). Redesign: Upgrading the building stock to meet (new) user demands. Paper presented at the Management and Innovation for a Sustainable Built Environment: MISBE 2011, CIB International Conference, Amsterdam, 20-23 June 2011. Remøy, H., & Van der Voordt, T. (2013). Adaptability: how to accommodate changing user preferences. Paper presented at the Proceedings of the 20th European Real Estate Society 20th Annual Conference, Vienna (Austria), 3-6 July, 2013; authors version. Remøy, H., & Van der Voordt, T. (2014). Adaptive reuse of office buildings into housing: opportunities and risks. Building Research & Information, 42(3), 381-390. Rodermond, W. (2011). Het taxeren van leegstaande kantoorruimte. (MRE), Amsterdam School of Real Estate, Amsterdam. Savills. (2014). Netherlands Market in Minutes: increase in occupier and investor demand October 2014: Savills. Scheltens, A., Van der Voordt, T., & Koppels, P. (2009). Key issues in successfull transformations of industrial heritage. Proceedings of SASBE, 15-19. Schiltz, S. (2007). Valuation of Vacant Properties: IVBN, Vereniging van Institutionele Beleggers in Vastgoed, Nederland. Sivitanides, P., Southard, J., Torto, R. G., & Wheaton, W. C. (2001). The determinants of appraisal-based capitalization rates. Real Estate Finance, 18(2), 27-38. Spaces. (2014). This is your Space. Retrieved May 17, 2014, from http://www.spaces.nl/wpcontent/uploads/2013/02/Brochure-2013-NL1.pdf Ten Have, G. G. (2000). Taxatieleer onroerende zaken: EPN, Educatieve Partners Nederland. TransformatieTeam. (2011). Transformatie kantoren gaat niet vanzelf. Rotterdam: TransformatieTeam en SBR. Van den Berg, M. (2011). Fiscal and commercial accounting rules on financial instruments. (Master), Tilburg School of Economics & Management, Tilburg. Van Gool, P. (2011). Moet een belegger wel huurincentives geven? ASRE paper. Amsterdam School of Real Estate. Amsterdam. Van Gool, P., & Rodermond, W. (2011). Waarde lege kantoren verschilt nauwelijks van de verkoopprijs. Vastgoedmarkt, 38, 60 - 61. Van Hoeken, F., & Bruyn, N. (2014). Vastgoed, ethiek en taxaties. BOSS magazine, 49, 80-90. Van Norren, H. (2007). Wat zijn de determinanten van het aanvangsrendement en is dit aanvangsrendement te voorspellen voor de nabije toekomst? (Master), University of Amsterdam. Verhaegh, M. (2005). Determinanten van de BAR op kantoren. (MSRE), Amsterdam School of Real Estate. Vermeulen, S. (2014, 31 January). DNB wil blackrock-rapport over vastgoed banken onder de pet houden. QOUTE. Wilkinson, S. J., Remøy, H., & Langston, C. (2014). Sustainable Building Adaptation: Innovations in Decisionmaking: John Wiley & Sons.
Appendices
li methodology
1987 ACLI dataset from 1970 to 1986 for apartments, retail, office, and industrial properties
publication year dataset
1993 ACLI dataset for commercial and industrial proper 1966 to 1988
Appendices
97
transaction-based or appraisal-based data
transaction-base
transaction-base
appendix 1 - cap rate determinants per reviewed articleSeemingly Uncorrelated Regression and cross-sect multiple regression model (stepwise regression approach) research technique
methodology
literature
series (panel data) regression
● positive effect ● negative effect ● tested and significant ○ tested but not significant
macro level variables author/authors country publication year inflation expectation
mortgage rate dataset transaction-basedratio or appraisal-based earnings/price data
research technique
GDP
● positive effect ● negative effect ● tested and significant ○ tested but not significant
macro level variables
GNP inflation expectation
Article 1 Article 2 time-series cross-sectional Froland indicator/description Ambrose and Nourse indicator/description U.S. U.S. the spread between long-term and shortthe spread between long-term and short1987 1993 ● term government bond rates term government bond rates ACLI dataset from 1970 to 1986 for apartments, retail, office, ACLI dataset for commercial and industrial properteis from ● and industrial properties 1966 to 1988 the S&P 500
●
transaction-base
term government bond rates
●
● the S&P 500
national vacancy rate
debt to GDP ratio
●
the percentage change in real Gross National Product
capacity utilization
debt spread vacancy rate
●
transaction-base
indicator/description time-series cross-sectional indicator/description time-series cross-sectional the percentage change in real Gross the spread between long-term and short○ ● ●
the spread between long-term and shortNational Product term government bond rates
earnings/price vacancy rate ratio
GNP
●
Seemingly Uncorrelated Regression and cross-sectional/time series (panel data) regression
multiple regression model (stepwise regression approach)
mortgageutilization rate capacity
debt-to-equity GDP cost of debt
time-series c
●
●
average loan to value (LTV)
●
property mortgage costs
●
○ ●
national vacancy rate ●
debt-to-equity lagged debt spread cost of debt
average loan to value (LTV)
●
property mortgage costs
●
debt to GDP ratio
equity spread debt spread
lagged equity spread lagged debt spread expected NOI growth rate equity spread market rent lagged equity spread expected NOI growth rate
rent ratio market rent rent ratio
risk-free rate
determinants of cap rate
determinants of cap rate
risk-free cap raterate cap rate
north- south- east
location dummies
north- south- east
location dummies
rating for the macro location
○
rating for the macro location
market risk/risk premium market risk/risk premium change in employment change infinancial financial employment micro level micro levelvariables variables class A building
class A building
condition of the property
condition ofquality the property construction age of the building construction quality newof dummy age the building renovation dummy new dummy occupancy rate
building size dummy renovation parking dummy occupancy rate price per square meter building size date of sale parking dummy easement dummy price per square meter ground lease dummy date of sale rent ratio dummy easement
ground lease dummy vacancy gap
rent land ratio leverage auction dummy
vacancy gap
other off market dummy
land leverage auction dummy tenant diversification tenant risk illiquidity risk volume 1
other off market dummy illiquidity risk volume 2
rating for the micro location operating costs tenant diversification remaining lease term
tenant risk illiquidity risk volume 1 illiquidity risk volume 2
indicatorindicator
time-series cross-sectional indicator time-series cross-sectional
indicator time-series cross-sectional
time-series c
li methodology
98
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
publicat dataset
transact data
methodology
literature
research
effect e effect nd significant ut not significant
author/authors country publication year dataset
Article 3
Jud and Winkler U.S. 1995 National Real Estate Index (NREI) panel data for office, warehouse/distribution, retail, and apartment properties
Article 4 Sivitanides, Southard, Torto, and Wheaton U.S. 2001 NCREIF Property Index for several property type
transaction-based or appraisal-based data
appraisal-based
appraisal-based
research technique
multiple regression model
Time-Series Cross-Section (TSCS) regression model
macro level variables
indicator
time-series cross-sectional indicator Annual percent change in consumer price index (CPI)
inflation expectation
● positive effect ● negative effect ● tested and significant ○ tested but not significant
macro le
inflation
mortgag
earnings GDP
time-series cross-sectional
GNP
●
mortgage rate
capacity
earnings/price ratio
vacancy
GDP
debt-tocost of d
GNP
debt to G
capacity utilization vacancy rate
debt spr
debt-to-equity cost of debt
lagged d
debt to GDP ratio debt spread
the return on BAA-rated debt minus the risk free rent
●
lagged debt spread
the return on BAA-rated debt minus the risk free rent, lagged respectively 1 and 2 periods
●●
lagged equity spread
lagged e
expected market r
● ●●
expected NOI growth rate market rent
● in two forms: Metropolitan-specific real rent index and also Annual percent change in the metropolitan- specific real rent index
rent ratio
Government Bond Return Rate
risk-free rate
determinants of cap rate
cap rate location dummies
●
●
rent rati ●
●
Logarithm of the annual capitalization rate calculated as the ratio of the NOI over the ending value
●
MSA dummies - fixed effect for each market
●
cap rate
●
rating for the macro location market risk/risk premium change in financial employment micro level variables class A building condition of the property construction quality age of the building new dummy renovation dummy occupancy rate building size parking dummy price per square meter date of sale easement dummy ground lease dummy rent ratio vacancy gap land leverage auction dummy other off market dummy
tenant diversification tenant risk illiquidity risk volume 1 illiquidity risk volume 2
indicator
time-series cross-sectional indicator
risk-free
time-series cross-sectional
determinants of cap rate
equity spread
the total return on the S&P 500 Index minus risk free rent the total return on the S&P 500 Index minus risk free rent lagged respectively 1 and 2 periods
equity sp
location
rating fo
market r
change i micro le
class A b
conditio construc age of th
new dum
renovati occupan building parking d price per date of s
easemen
ground l
rent rati vacancy
land leve auction
other of
rating for the micro location operating costs remaining lease term
tenant d
tenant r illiquidit
illiquidit
li methodology
1987 ACLI dataset from 1970 to 1986 for apartments, retail, office, and industrial properties
publication year dataset
1993 ACLI dataset for commercial and industrial proper 1966 to 1988
methodology
● positive effect ● negative effect ● tested and significant ○ tested but not significant
literature
Appendices transaction-base
transaction-base
research technique
multiple regression model (stepwise regression approach)
Seemingly Uncorrelated Regression and cross-sect series (panel data) regression
macro level variables author/authors country publication year inflation expectation
mortgage rate dataset transaction-based or appraisal-based
earnings/price ratio data research technique
GDP macro level variables
● positive effect ● negative effect ● tested and significant ○ tested but not significant
99
transaction-based or appraisal-based data
GNP inflation expectation
Article 5 indicator/description Verhaegh The Netherlands 2005 the spread between long-term and shortgovernment ROZ / IPDterm Property Index bond rates
Article 5
time-series indicator/description Van cross-sectional Norren ●
The Netherlands 2007 -
time-series c
the spread between long-term and shortterm government bond rates
●
● the S&P 500 appraisal-based
●
multiple regression model with gross initial yield (GIY) as the dependent variable
● multiple regression model
indicator time-series cross-sectional indicator the percentage change in real Gross ○ National Product
mortgageutilization rate capacity
time-series cross-sectional ●
● national vacancy rate
earnings/price vacancy rate ratio
●
debt-to-equity GDP cost of debt GNP
average loan to ●value (LTV)
●
property mortgage costs
●
debt to GDP ratio capacity utilization in an absolute term
vacancy rate debt spread
●
●
rent ratio
●
●
risk-free rate
●
debt-to-equity cost of debt lagged debt spread debt to GDP ratio
equity spread debt spread lagged debt spread lagged equity spread equity spread expected NOI growth rate market rent lagged equity spread
expected NOI growth rate rent ratio market rent
●
risk-free rate
determinants of cap rate
determinants of cap rate
cap rate cap rate
●
north- south- east
location dummies location dummies
●
rating for the macro location
●
rating for the macro location
market risk/risk premium market risk/risk premium change in employment change infinancial financial employment micro level variables micro level variables class A building
indicator
indicator
time-series cross-sectional indicator
time-series cross-sectional indicator time-series cross-sectional
class A building
condition of the property
construction condition ofquality the property age of the building construction quality newof dummy age the building
●
renovation dummy
new dummy occupancy rate
building size dummy renovation parking dummy occupancy rate price per square meter building size date of sale parking dummy easement dummy price per square meter ground lease dummy date of sale rent ratio
easement dummy
●
ground lease dummy vacancy gap
rent land ratio leverage auction dummy
vacancy gap
other off market dummy
land leverage auction dummy tenant diversification tenant risk illiquidity risk volume 1
other off market dummy illiquidity risk volume 2 rating for the micro location operating costs tenant diversification remaining lease term
tenant risk illiquidity risk volume 1 illiquidity risk volume 2
● ●
time-series c
li methodology
100
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
publicat dataset
transact data
methodology
literature
research
effect e effect and significant ut not significant
Article 6
Article 67
McDonald and Dermisi U.S. 2008 132 office sales in downtown Chicago from 1996 to 2007
Netzell Sweden 2009 3022 discounted cash flow market valuations of office properties in Stockholm, Gothenburg and Malmö during 19982004
transaction-based or appraisal-based data
transaction-based
appraisal-based
research technique
multiple regression model
ANOVA and Feasible Generalized Least Square (FGLS) estimator
macro level variables
indicator
author/authors country publication year
dataset
time-series cross-sectional indicator
● positive effect ● negative effect ● tested and significant ○ tested but not significant
macro le
inflation
mortgag
earnings GDP
time-series cross-sectional
GNP
inflation expectation mortgage rate
capacity
○
earnings/price ratio
vacancy
GDP
debt-tocost of d
GNP
debt to G
capacity utilization
vacancy rate
∆ Vacancy rate (the change in the downtown Chicago vacancy rate)
●
●
vacancy gap - the deviation of current vacancy and long-run vacancy per segment and year
●
debt spr
●
debt-to-equity cost of debt
lagged d
debt to GDP ratio
equity sp
debt spread lagged debt spread
lagged e
equity spread
expected market r
lagged equity spread expected NOI growth rate market rent
rent rati the deviation of current rent from the market rent per segment and year
risk-free rate
●
3-month Treasury bill rate for the quarter in which the building was sold
The government bond real earnings on stock index
●
●
cap rate
●
determinants of cap rate
cap rate location dummies 3 different cities
rating for the macro location market risk/risk premium change in financial employment micro level variables
class A building
expected rate of return for the entire capital market (S&P 500) - the risk-free interest rate the growth in office employment
indicator A or B (e.g. the buildings with the most upto-date facilities is A, otherwise is B)
condition of the property construction quality age of the building
The government bond rate earnings per share
●
●
●
●
●
time-series cross-sectional indicator
time-series cross-sectional
(Age x renovated) occupancy rate with in the building
rating fo
market r
change i micro le
class A b
●
conditio construc age of th
different effect on different submarkets
○/●
new dum
● ○ ○
parking in building
location
●
new dummy renovation dummy occupancy rate building size parking dummy price per square meter date of sale
risk-free
determinants of cap rate
rent ratio
○ ○ ○
easement dummy ground lease dummy
●
rent ratio
the difference between the rent ratio for an individual property and the rent ratio per segment and year
●
vacancy gap
the difference between the vacancy gap for an individual property and the vacancy gap per segment and year
●
renovati occupan building parking d price per date of s
easemen
ground l
rent rati
land leverage auction dummy
vacancy other off market dummy
land leve auction
tenant diversification tenant risk illiquidity risk volume 1
other of
illiquidity risk volume 2 rating for the micro location operating costs remaining lease term
three segments including CBD, central and other
●
tenant d
tenant r illiquidit
illiquidit
li methodology
1987 ACLI dataset from 1970 to 1986 for apartments, retail, office, and industrial properties
publication year dataset
1993 ACLI dataset for commercial and industrial proper 1966 to 1988
methodology
● positive effect ● negative effect ● tested and significant ○ tested but not significant
literature
Appendices transaction-base
transaction-base
research technique
multiple regression model (stepwise regression approach)
Seemingly Uncorrelated Regression and cross-sect series (panel data) regression
macro level variables author/authors country publication year inflation expectation
mortgage rate dataset transaction-based or appraisal-based earnings/price ratio data
research technique
GDP
● positive effect ● negative effect ● tested and significant ○ tested but not significant
101
transaction-based or appraisal-based data
macro level variables
GNP inflation expectation
Article 8 indicator/description Chervachidze and Wheaton U.S. the spread between long-term and short2013 term government bond rates NCREIF Property Index
Article 9
time-series cross-sectional indicator/description Hoesli and Chaney ● ●
time-series c
Switzerland the spread between long-term and short2014 term government bond rates IAZI database for the period between 1985 to 2010
●
the S&P 500 ●
appraisal-based
The fixed effects panel technique with ordinary least squares (OLS) estimators
multiple regression considering the logarithm of cap rates as dependent variable
indicator time-series cross-sectional indicator the percentage change in real Gross ○ National Product
mortgageutilization rate capacity
● national vacancy rate
earnings/price vacancy rate ratio
●
●
both transaction-based and appraisal-based
time-series cross-sectional
Shiller P/E-ratio of the S&P 500 defined as the current price to the average inflation-adjusted earnings from the past ten years.
●
As a proxy for growth rate defined as loan to value (LTV) average ● Growth in nominal GDP
debt-to-equity GDP cost of debt
●
property mortgage costs
GNP
●
debt to GDP ratio capacity utilization Vacancy rate of the municipality in which the property is located
debt spread vacancy rate
●
●
debt-to-equity lagged debt spread cost of debt debt to GDP ratio
equity spread debt spread
the growth in the debt as a fraction of GDP the spread between the Moody’s AAA yield and the 10 year Treasury bond
● ●
lagged equity spread lagged debt spread expected NOI growth rate equity spread market rent lagged equity spread
expected NOI growth rate rent ratio market rent ●
Real T-Bond yield calculated as nominal yield minus inflation rate
●
cap rate
Logarithm of the annual capitalization rate NCREIF database calculated from Net Operating Income and asset values.
●
location dummies
MSA dummies - fixed effect for each market
risk-free rate
determinants of cap rate
risk-free rate cap rate
determinants of cap rate
Rent relative to median rent
Log(Real Rent ratio) calculated as a ratio of real rent data from CBRE rent database for a given MSA in a given quarter to the historical average of real rent for this MSA
rent ratio
location dummies rating for the macro location rating for the macro location
●
●
●
ten-year government bonds
●
north- south- east ●
used to capture variation in α (fixed effect per each market) from Bad to Very Good as an indication for risk premium and growth
●
● ●
market risk/risk premium market risk/risk premium change in employment change infinancial financial employment micro level micro levelvariables variables
indicatorindicator
time-series cross-sectional indicator time-series cross-sectional
indicator time-series cross-sectional
class A building
class A building
condition of the property
condition ofquality the property construction age of the building construction quality newof dummy age the building renovation dummy new dummy occupancy rate building size dummy renovation parking dummy occupancy rate price per square meter building size date of sale parking dummy easement dummy price per square meter ground lease dummy date of sale
1: Bad to 4: Very Good
●
1: Bad to 4: Very Good
● ●
Ln (Age) new building consider when it is not older than two years when the property has been refurbished
a non-possessory right to use and/or enter onto the real property of another
● ● ●
● ●
Rent relative to median rent
rent ratio easement dummy
●
ground lease dummy vacancy gap
rent land ratio leverage auction dummy
vacancy gap
other off market dummy
land leverage auction dummy tenant diversification tenant risk illiquidity risk volume 1
other off market dummy illiquidity risk volume 2
rating for the micro location operating costs tenant diversification remaining lease term
tenant risk illiquidity risk volume 1 illiquidity risk volume 2
measured as Ln (volume/lot size)
●
a force sale
●
whenever the transaction was neither an auction nor done at arm's length, i.e. when the sale was e.g. in relation with a related legal entity or to a family member
●
Percentage of rents from commercial tenants (Mixed used complex) average tenant quality (residential) Ln (volume) Centered square of lVol to capture potential nonlinearity the location within the macro location Bad to Very Good
● ● ● ● ●
time-series c
= risk free rate, ⁄ = Loan-to-value ratio, = rate of return on debt, 102 = premium from participation in real estate,
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
= premium on location attributes,
appendix 2 - Transaction cap rate Calculating models = premium on property-itself attributes,
four = premium property-user attributes, In total modelonbased on two main methods of pure cap rates (basic capitalization methods), = a constant expected rate of growth in the NOI. and cap rates with corrections (conventional method), are used to calculated the transaction cap rates. The first model (Model 1), is explained thoroughly in Chapter 7, § 7.2.2. Figure i shows the calculating cap rates. mentioned in Chapter 7, § 7.2.2, this model uses the ERV (18) =model + ∑ofpure + ∑ As + ∑ + ∑ + (Effective Rental Value) based on firstly, the availability of ERVs of a property, or it calculates it based on theWhere; average effective rents of rental transactions in the radius of 500 m, 1000 m around the α = constant or intercept term, property, or coefficient average effective percontext, Amsterdam the selling = the of variablesrents indicating location,district propertyat and office userstime, factors,when no actual rental ε = the error term. transactions were present for specific properties. The second model (Model 2), uses the same pure cap rate model where all the risk associated with the investment are reflected in the cap rate and 1 imposed = as secondary correction. However the differences between Model 1 and (19) are not Model 2 is related to differences in the input for the ERVs. Model 2 uses merely the average ERVs per Where; Amsterdam district the(%),selling time to calculate the TRI (Total Rental Income). Both cap rates for = Appraised capatrate netModel income multiplier. Model= 1 and 2 are based on the following equation:
=
(20) (20)
Where; = overall capitalization rate (%), = property value (gross transaction price), = Net Operating Income in the first year.
The third model (Model 3) and the fourth model (Model 4) are based on the conventional model that calculates the cap rates by considering corrections on the actual transaction price as shown in equation i (see also Figure ii): DIRECT INCOME CAPITALIZATION
Figure i, the calculating model of pure cap rates (all risks included in cap rates)
Property type: Office
Gross rental value SaleID#
ObjectID#
size (m² LFA)
Effective Rent Value (ERV) (€/m² LFA)
Market Rent Value (€/m² LFA)
60
##
1835
87
87
Total Rental Income (TRI)
€
159.645
Net rental value 10,0%
Operating Expenses Ratio (OER) of TRI
Net Operating Income (NOI)
€
143.681
Capitalization (Cap) rate (NIY)
1 Cap rate (NIY)
11,8% 2
Net Transaction Price to Gross Transaction Price Value including KK
€
1.217.135 6,3%
€
1.145.000
Transfer tax and legal fees (KK %)
Value excluding purchase costs
Where; = overall capitalization rate (%), = property value (gross transaction price), A p p e n d i c=eNet s Operating Income in the first year.
=
103
(i)
(i
Where; = overall capitalization rate (%), = property value (gross transaction price), = Net Operating Income in the first year.
In fact the realized transaction price is transformed by the following equation:
() = ∑
Gross transaction price (TP)
(i
Where, Corrected Gross Transaction Price = (Net transaction price + Correction for KK%) + ToM = time for a property to be let, Correction for difference between actual rent (ERV) and market rent + Corrections for n = the number of WOZ-objects in one neighborhood.
vacancy and its related costs + Correction for overdue maintenance
Correction for KK%: 1 8 = 3,1,4 = to=the 2.6 This correction, which is transferring the realized transaction price gross transaction price by 3 3 correcting for transfer tax and legal fees, is explained thoroughly in § 7.2.2.
Correction the difference between actual rent (ERV) and market rent + Corrections for vacancy ( =for ∗ ) ∗ ∗ 1.16 and its related costs These corrections are categorized under 3 conditions: current building has no vacancy, is partially Where; vacant, or overall is structurally vacant. = cost of vacancy (€), The second one (partially vacant) is not valid for the used database, 2 = total lettable floor area ), those with a single tenant and a single owner (the tax units used by since all the transaction sales (m are = effective rental value, the municipality the(year). WOZ-values). Therefore, in this part the two conditions of whether = the lengthto ofcalculate the void period the property is completely vacant or fully occupied are discussed as follows:
Figure ii, main steps of calculating transaction cap rates for model 3 and 4
(i
104
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
- No vacancy In the case of a non-vacant object, three correction entries are made. The first one calculates the difference between present value of actual rent (ERV) and market rent (average ERVs in the 500m radius), of the remaining income till the end of the lease contract. The second one calculates the present value of income loss due to the vacancy, by considering first, the length of the void period in the direct neighborhood of the property and second the year in which the vacancy occurs. The third one corrects for additional vacancy costs which contain costs for marketing and letting commission. Based on the market evidence, the costs for marketing and letting commission are calculated as 16% of the total vacancy costs. Since in this thesis the total rental income is calculated using the effective rent, the incentives are already incorporated in the rent, and no incentive correction is needed. - Structurally vacant In the case of a structurally vacant object, two correction entries regarding vacancy are made. The first one calculates the present value of income loss due to the vacancy, by considering the length of the void period in the property’s direct neighborhood. The second one corrects for additional vacancy costs of marketing and letting commission (16%). To avoid complexity and subjectivity (bias) of using DCF to calculate the NPV of the remaining lease term (due to the lack of data regarding the investors discounted rate), a shortcut formula is used with an assumption of no growth in property rent and value (the concept of cash flow in perpetuity). Therefore, the correction for both the difference between actual rent (ERV) and market rent, and income loss due to the vacancy are not discounted to the NPV and assumed to occur in the first year. Where;
= overall capitalization rate (%), = property value (gross transaction price), However, length ofIncome the void = the Net Operating in theperiods first year.are calculated precisely as explained in the next paragraphs.
Calculating Void period In order to measure = the potential income loss due to the vacancy, the average void period in the direct neighborhood of the WOZ-objects are calculated at the selling time. The void period is the expected period that a property remains vacant and generates no rental income. To measure that, Where; of the average Time on Market (ToM) from the time that a property becomes vacant till the length = overall capitalization rate (%), it is let over specific period of 5 years, is calculated. This is estimated based on equation =again, property valuea(gross transaction price), = Net Operating Income in the first year. (ii), using the registered tax vacant codes of the WOZ-objects per neighborhood, where no rental income for the WOZ-objects were registered to be taxed.
() = ∑
(ii)
(i)
(ii)
Where, ToM = time for a property to be let, n = the number of WOZ-objects in one neighborhood.
Object#
2002
2003
2004
2005
2006
1
X
X
0
0
0
2
X
X
X
0
X
3
X
0
0
0
0
( X=Occupied, =0=vacant
∗ ) ∗ ∗ 1.16
Where; = overall cost of vacancy (€), = total lettable floor area (m2), = effective rental value,
2007
2008
2009
2010
2011
X
X
X
X
X
X
X
2012
2013
2014
X
X
X
X
X
0
0
0
1 8 X X X 0 = X 3,1,4 = X = 2.6 3 3 X
X
X
Table i, an example to explain how the void periods are calculated
(iii)
() =
Where; = overall capitalization rate (%), A pp e n d i c e s property value (gross transaction price), ∑= = Net Operating Income in the first year.
(ii)
105
ere, M = time for a property to be let, For example when a property is sold in2006, the average void period for that specific neighborhood the number of WOZ-objects in one neighborhood. () = ∑ (in the example in Table i, consists of three properties), is expected to be as follows:
(i
Where, ToM = time 1 for a property to8be let, n = the= number WOZ-objects one neighborhood. of 3,1,4 = =in 2.6
3
3
1 8 Calculating Cost of vacancy = 3,1,4 = = (iii) 2.6 = ( ∗ ) ∗ ∗ 1.16 3 3 After calculating the void periods, the income loss due to vacancy is calculated according to equation (iii):
ere; = overall cost of vacancy (€), = total lettable floor area (m2), = effective rental value, = the length of the void period (year).
= ( ∗ ) ∗ ∗ 1.16
(iii)
Where; = overall cost of vacancy (€), = total lettable floor area (m2), = effective rental value, = the length of the void period (year).
Correction for overdue maintenance This correction is done based on the available information from the GBA (the filled questionnaire by new buyers) over whether at the time of sale there was an overdue maintenance, and in case of such an overdue, the costs were specified in euro of these foreseen maintenances. The differences between Model 3 and Model 4 is the same as the differences between Model 1 and Model 2, regarding the input for the ERVs. Model 3 considers the sequence of available ERV in the property, the radius of 500 m, 1000 m, or if none of these ERVs are available it uses the average ERVs per Amsterdam district. Whereas Model 4 simply uses the average ERVs per Amsterdam district in the calculation model. The calculation model for the cap rates with correction is shown in Figure iii. Figure iv compares the outcome of these models (Model 1 to Model 4), also with the average appraised cap rates for both WOZ-values and market reports (based on C&W). For Model 1 and Model 2 the data from 2004 to 2011 is included in the model, while Model 3 and Model 4 included the cap rates from 2006 to 2011. The reason for omitting the years 2004 and 2005 for Models 3 and 4, is due to the calculation of the void periods, which the length of the average Time on Market (ToM) from the time that a property becomes vacant till it is let again, over a specific period of 5 years, is calculated. Since the tax vacancy database includes data merely since 2002, the first year that the transaction cap rates for Model 3 and 4 could be calculated, was 2006. For the purpose of this thesis, the outcome of Model 1 is selected due to the fact that the result are firstly representing the pure cap rates where all risk associated with the investments is reflected in the calculated cap rates. Secondly, the use of effective rents based on the sequence of the property itself, radius of 500 m, 1000 m and lastly if none of the previous ERVs are available, the average ERVs of the Amsterdam districts are considered. This is due to the fact that firstly these outcomes are more close to what real estate agenciesreport (with some exceptions in 2009). Secondly it is because of the significant impact of the location on the rental value and the provided incentives in the Amsterdam office market. Using the average ERVs per district seem to be far from the dynamics
(i
106
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
of the submarkets in Amsterdam. Finally, the most crucial reason to choose this model (Model 1), is to decrease the subjective assumptions and inputs to calculate the cap rates. This model has the DIRECT INCOME CAPITALIZATION least manipulation in the data, therefore providing the most objectively calculated cap rates, and Property type: Office thus reduces the biases. Figure iii, the calculating model of cap rates with corrections
Gross rental value SaleID#
ObjectID#
size (m² LFA)
Effective Rent Value (ERV) (€/m² LFA)
Market Rent Value (€/m² LFA)
60
##
1835
87
87
Total Rental Income (TRI)
€
159.645
Net rental value 10,0%
Operating Expenses Ratio (OER) of TRI
Net Operating Income (NOI)
€
143.681
Capitalization (Cap) rate (NIY)
1 Cap rate (NIY)
11,4% 2
Gross market value Before corrections ERV -/- Market Rent Average Void Period (AVP) in year Marketing and letting commission Cost of Vacancy (CoV) Overdue maintenance
€
€ €
0,2551 16,0% 42.517 -
Total corrections
€
1.259.652
Value including KK
€
1.217.135 6,3%
€
1.145.000
Net Transaction Price to Gross Transaction Price
Transfer tax and legal fees (KK %)
Value excluding purchase costs
Cap rates Figure iv, a comparison among the outcomes different models (Model 1 to Model 4)
12,00% 10,00% 8,00% 6,00% 4,00% 2,00% 0,00% 2004 Model 1
2005
2006 Model 2
2007 Model 3
2008 Model 4
2009
2010 WOZ
2011 C&W
Appendices
107
appendix 3 - regression model outcomes Transaction cap rate regression with context variables Descriptive Statistics Mean
Std. Deviation
N
TP_CapRate
,060
,044
121
OfficeJobs
105,492
3,344
121
OfficeInvest
,463
,143
121
DD2
,041
,200
121
DD3
,074
,263
121
DD4
,058
,234
121
DD6
,033
,180
121
DD7
,190
,394
121
DD8
,074
,263
121
LnLFAm2
5,929
1,397
121
RentR
,926
,421
121
LnAge
2,769
1,730
121
LnAfstDBZ_IC_Station
7,463
,516
121
Asbest
,074
,263
121
Variables Removed
Method
Variables Entered/Removeda Model
Variables Entered
1
OfficeInvest, OfficeJobsb
Enter
2
DD6, DD7, DD2, DD4, DD3, DD8b
Enter
3
LnLFAm2, Asbest, RentR, LnAge, LnAfstDBZ_IC_Stationb
Enter
a. Dependent Variable: TP_CapRate b. All requested variables entered.
Model Summaryd DurbinWatson
Change Statistics Model
R
1
,135a
R Square
Adjusted R Square
Std. Error of the Estimate
R Square Change
F Change
df1
df2
Sig. F Change
,018
,002
,044
,018
1,099
2
118
,337
2
,564
b
,318
,270
,038
,300
8,216
6
112
,000
3
,716c
,513
,454
,033
,195
8,567
5
107
,000
1,972
a. Predictors: (Constant), OfficeInvest, OfficeJobs b. Predictors: (Constant), OfficeInvest, OfficeJobs, DD6, DD7, DD2, DD4, DD3, DD8 c. Predictors: (Constant), OfficeInvest, OfficeJobs, DD6, DD7, DD2, DD4, DD3, DD8, LnLFAm2, Asbest, RentR, LnAge, LnAfstDBZ_IC_Station d. Dependent Variable: TP_CapRate
108
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
ANOVAa Model 1
2
3
Sum of Squares
df
Mean Square
F
Sig.
Regression
,004
2
,002
1,099
,337b
Residual
,229
118
,002
Total
,233
120
Regression
,074
8
,009
6,537
,000c
Residual
,159
112
,001
Total
,233
120
Regression
,120
13
,009
8,677
,000d
Residual
,113
107
,001
Total
,233
120
a. Dependent Variable: TP_CapRate b. Predictors: (Constant), OfficeInvest, OfficeJobs c. Predictors: (Constant), OfficeInvest, OfficeJobs, DD6, DD7, DD2, DD4, DD3, DD8 d. Predictors: (Constant), OfficeInvest, OfficeJobs, DD6, DD7, DD2, DD4, DD3, DD8, LnLFAm2, Asbest, RentR, LnAge, LnAfstDBZ_IC_Station Coefficientsa
Model 1
2
3
Unstandardized Coefficients
Standardized Coefficients Beta
B
Std. Error
(Constant)
,173
,132
OfficeJobs
-,001
,001
OfficeInvest
-,040
,029
(Constant)
,094
,119
OfficeJobs
,000
,001
-,020
OfficeInvest
-,033
,025
DD2
,027
,018
DD3
,071
DD4
,011
DD6 DD7
Collinearity Statistics t
Sig.
1,308
,193
Tolerance
VIF
-,068
-,731
-,131
-1,406
,466
,963
1,038
,162
,963
1,038
,789
,432
-,243
,809
,870
1,150
-,107 ,121
-1,320
,190
,931
1,074
1,513
,133
,948
1,055
,014 ,015
,424
5,247
,000
,930
1,075
,058
,723
,471
,938
1,066
,080 -,009
,019
,325
4,106
,000
,969
1,032
,009
-,078
-,948
,345
,896
1,116
DD8 (Constant)
,023
,014
,140
1,695
,093
,896
1,117
,389
,156
2,493
,014
OfficeJobs
-,002
,001
-,175
-2,159
,033
,690
1,450
OfficeInvest
-,053
,022
-,173
-2,370
,020
,855
1,169
DD2
,045
,018
,203
2,552
,012
,716
1,397
DD3
,076
,012
,456
6,175
,000
,836
1,196
DD4
,030
,013
,159
2,215
,029
,886
1,128
DD6
,088
,018
,359
4,994
,000
,879
1,137
DD7
-,034
,009
-,307
-3,706
,000
,663
1,509
DD8
,040
,018
,240
2,184
,031
,378
2,648
LnLFAm2
,006
,002
,197
2,721
,008
,870
1,150
RentR
,031
,008
,296
3,999
,000
,830
1,205
LnAge
,007
,002
,292
3,490
,001
,652
1,534
LnAfstDBZ_ IC_Station
-,021
,010
-,247
-2,175
,032
,354
2,828
Asbest
,024
,012
,145
1,989
,049
,851
1,175
a. Dependent Variable: TP_CapRate
Appendices
109
Excluded Variablesa Collinearity Statistics Model 1
2
Beta In
t
Sig.
Partial Correlation
Tolerance
VIF
Minimum Tolerance
DD2
,083b
,900
,370
,083
,980
1,020
,946
DD3
,400
4,679
,000
,397
,968
1,033
,933
DD4
,012b
,136
,892
,013
,987
1,013
,952
DD6
,296
3,376
,001
,298
,995
1,005
,958
DD7
-,201b
-2,236
,027
-,202
,993
1,007
,957
DD8
b
,103
1,106
,271
,102
,952
1,050
,917
LnLFAm2
,206b
2,269
,025
,205
,973
1,027
,946
RentR
b
,208
2,260
,026
,205
,950
1,053
,917
LnAge
,118b
1,279
,203
,117
,968
1,033
,933
LnAfstDBZ_ IC_Station
b
,162
1,777
,078
,162
,985
1,015
,949
Asbest
,120b
1,315
,191
,121
,992
1,008
,958
LnLFAm2
c
,226
2,858
,005
,262
,918
1,089
,862
RentR
,296c
3,667
,000
,329
,841
1,189
,823
LnAge
,204
2,342
,021
,217
,774
1,293
,749
LnAfstDBZ_ IC_Station
-,113c
-,951
,344
-,090
,434
2,306
,434
Asbest
,178c
2,158
,033
,201
,866
1,155
,816
b
b
c
a. Dependent Variable: TP_CapRate b. Predictors in the Model: (Constant), OfficeInvest, OfficeJobs c. Predictors in the Model: (Constant), OfficeInvest, OfficeJobs, DD6, DD7, DD2, DD4, DD3, DD8
Transaction cap rate regression with year dummies Descriptive Statistics Mean
Std. Deviation
N
TP_CapRate
,060
,044
121
DY2004
,124
,331
121
DY2005
,074
,263
121
DY2006
,124
,331
121
DY2007
,198
,400
121
DY2009
,157
,365
121
DY2010
,066
,250
121
DY2011
,091
,289
121
DD2
,041
,200
121
DD3
,074
,263
121
DD4
,058
,234
121
DD6
,033
,180
121
DD7
,190
,394
121
DD8
,074
,263
121
LnLFAm2
5,929
1,397
121
RentR
,926
,421
121
LnAge
2,769
1,730
121
LnAfstDBZ_IC_Station
7,463
,516
121
Asbest
,074
,263
121
110
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
Variables Entered/Removeda Model
Variables Entered
Variables Removed
Method
1
DY2011, DY2010, DY2005, DY2006, DY2004, DY2009, DY2007b
Enter
2
DD2, DD6, DD7, DD3, DD4, DD8b
Enter
3
LnLFAm2, Asbest, LnAge, RentR, LnAfstDBZ_IC_Stationb
Enter
a. Dependent Variable: TP_CapRate b. All requested variables entered.
Model Summaryd Change Statistics Model
R
1
DurbinWatson
R Square
Adjusted R Square
Std. Error of the Estimate
R Square Change
F Change df1
df2
Sig. F Change
,294a
,087
,030
,043390
,087
1,533
7
113
,163
2
,626
b
,392
,318
,036389
,305
8,944
6
107
,000
3
,737c
,543
,463
,032294
,152
6,772
5
102
,000
1,946
a. Predictors: (Constant), DY2011, DY2010, DY2005, DY2006, DY2004, DY2009, DY2007 b. Predictors: (Constant), DY2011, DY2010, DY2005, DY2006, DY2004, DY2009, DY2007, DD2, DD6, DD7, DD3, DD4, DD8 c. Predictors: (Constant), DY2011, DY2010, DY2005, DY2006, DY2004, DY2009, DY2007, DD2, DD6, DD7, DD3, DD4, DD8, LnLFAm2, Asbest, LnAge, RentR, LnAfstDBZ_IC_Station d. Dependent Variable: TP_CapRate
ANOVAa Model 1
2
3
Sum of Squares
df
Mean Square
F
Sig.
Regression
,020
7
,003
1,533
,163b
Residual
,213
113
,002
Total
,233
120
Regression
,091
13
,007
5,302
,000c
Residual
,142
107
,001
Total
,233
120
Regression
,127
18
,007
6,743
,000d
Residual
,106
102
,001
Total
,233
120
a. Dependent Variable: TP_CapRate b. Predictors: (Constant), DY2011, DY2010, DY2005, DY2006, DY2004, DY2009, DY2007 c. Predictors: (Constant), DY2011, DY2010, DY2005, DY2006, DY2004, DY2009, DY2007, DD2, DD6, DD7, DD3, DD4, DD8 d. Predictors: (Constant), DY2011, DY2010, DY2005, DY2006, DY2004, DY2009, DY2007, DD2, DD6, DD7, DD3, DD4, DD8, LnLFAm2, Asbest, LnAge, RentR, LnAfstDBZ_IC_Station
Appendices
111
Coefficientsa Unstandardized Coefficients
Standardized Coefficients
B
Std. Error
Beta
(Constant)
,068
,010
DY2004
,016
,015
,117
DY2005
-,012
,017
DY2006
-,022
,015
DY2007
-,017
DY2009
-,020
DY2010 DY2011
Model 1
2
3
Collinearity Statistics t
Sig.
Tolerance
VIF
6,998
,000
1,050
,296
,652
1,533
-,074 -,164
-,715
,476
,745
1,342
-1,473
,143
,652
1,533
,013 ,014
-,151
-1,261
,210
,567
1,764
-,165
-1,428
,156
,608
1,644
-,007 ,009
,018
-,040
-,394
,694
,765
1,307
,016
,058
,542
,589
,710
1,409
(Constant)
,062
,010
DY2004
,001
,013
,008
6,400
,000
,083
,934
,578
1,729
DY2005
-,012
,015
DY2006
-,025
,013
-,069
-,760
,449
,694
1,442
-,185
-1,924
,057
,616
1,623
DY2007
-,013
DY2009
-,026
,012
-,118
-1,125
,263
,516
1,937
,012
-,213
-2,103
,038
,552
1,812
DY2010 DY2011
-,014
,016
-,081
-,915
,362
,725
1,380
,017
,014
,113
1,243
,217
,683
1,465
DD2
,019
,017
,088
1,125
,263
,925
1,081
DD3
,071
,013
,425
5,331
,000
,896
1,116
DD4
,004
,015
,022
,272
,787
,867
1,153
DD6
,086
,019
,350
4,486
,000
,936
1,068
DD7
-,013
,009
-,120
-1,473
,144
,850
1,176
DD8
,017
,014
,099
1,205
,231
,847
1,181
(Constant)
,119
,073
1,627
,107
DY2004
,025
,013
,191
1,918
,058
,454
2,205
DY2005
,009
,014
,053
,613
,541
,609
1,642
DY2006
-,011
,012
-,082
-,911
,364
,551
1,813
DY2007
-,010
,011
-,094
-,960
,340
,470
2,126
DY2009
-,008
,012
-,069
-,720
,473
,487
2,053
DY2010
-,010
,014
-,056
-,691
,491
,672
1,488
DY2011
,008
,013
,051
,612
,542
,638
1,568
DD2
,038
,018
,173
2,132
,035
,678
1,475
DD3
,074
,013
,442
5,844
,000
,781
1,280
DD4
,025
,014
,133
1,772
,079
,798
1,254
DD6
,088
,018
,358
4,942
,000
,852
1,173
DD7
-,035
,009
-,314
-3,698
,000
,621
1,611
DD8
,035
,019
,207
1,841
,069
,355
2,814
LnLFAm2
,007
,002
,209
2,801
,006
,806
1,240
RentR
,028
,008
,272
3,367
,001
,687
1,455
LnAge
,007
,002
,273
3,247
,002
,631
1,584
LnAfstDBZ_ IC_Station
-,020
,010
-,238
-2,024
,046
,322
3,102
Asbest
,023
,013
,139
1,809
,073
,763
1,311
a. Dependent Variable: TP_CapRate
112
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
Excluded Variablesa Collinearity Statistics Model 1
2
Beta In
t
Sig.
Partial Correlation
Tolerance
VIF
Minimum Tolerance
DD2
,053b
,581
,563
,055
,965
1,036
,562
DD3
,403b
4,729
,000
,408
,936
1,069
,565
DD4
-,006
-,060
,952
-,006
,928
1,078
,557
DD6
,329b
3,772
,000
,336
,952
1,051
,564
DD7
-,218
-2,412
,018
-,222
,949
1,054
,566
DD8
,070b
,737
,463
,069
,902
1,109
,547
LnLFAm2
b
,211
2,284
,024
,211
,912
1,097
,566
RentR
,194b
1,941
,055
,180
,790
1,266
,552
LnAge
,087
,929
,355
,087
,925
1,082
,567
LnAfstDBZ_ IC_Station
,125b
1,329
,186
,125
,902
1,108
,567
Asbest
,117b
1,239
,218
,116
,906
1,103
,559
LnLFAm2
c
,217
2,742
,007
,257
,856
1,168
,515
RentR
,251c
2,869
,005
,268
,693
1,443
,496
LnAge
,172
1,978
,050
,189
,736
1,359
,516
LnAfstDBZ_ IC_Station
-,132c
-1,088
,279
-,105
,384
2,606
,384
Asbest
,174c
2,070
,041
,197
,782
1,279
,513
b
b
b
c
a. Dependent Variable: TP_CapRate b. Predictors in the Model: (Constant), DY2011, DY2010, DY2005, DY2006, DY2004, DY2009, DY2007 c. Predictors in the Model: (Constant), DY2011, DY2010, DY2005, DY2006, DY2004, DY2009, DY2007, DD2, DD6, DD7, DD3, DD4, DD8
WOZ cap rate regression with context variables Descriptive Statistics Mean
Std. Deviation
N
WOZ_CapRate
,093
,012
121
OfficeJobs
105,492
3,344
121
OfficeInvest
,463
,143
121
DD2
,041
,200
121
DD3
,074
,263
121
DD4
,058
,234
121
DD6
,033
,180
121
DD7
,190
,394
121
DD8
,074
,263
121
LnLFAm2
5,929
1,397
121
RentR
,926
,421
121
LnAge
2,769
1,730
121
LnAfstDBZ_IC_Station
7,463
,516
121
Asbest
,074
,263
121
Appendices
113
Variables Entered/Removeda Model
Variables Entered
1
OfficeInvest, OfficeJobsb
Variables Removed
Method Enter
2
DD6, DD7, DD2, DD4, DD3, DD8b
Enter
3
LnLFAm2, Asbest, RentR, LnAge, LnAfstDBZ_IC_Stationb
Enter
a. Dependent Variable: WOZ_CapRate b. All requested variables entered. Model Summaryd Change Statistics
DurbinWatson
Model
R
R Square
Adjusted R Square
Std. Error of the Estimate
R Square Change
F Change
df1
df2
Sig. F Change
1
,333a
,111
,096
,011835
,111
7,383
2
118
,001
2
,583b
,340
,293
,010468
,229
6,469
6
112
,000
3
,631
,398
,325
,010227
,058
2,072
5
107
,075
c
2,069
a. Predictors: (Constant), OfficeInvest, OfficeJobs b. Predictors: (Constant), OfficeInvest, OfficeJobs, DD6, DD7, DD2, DD4, DD3, DD8 c. Predictors: (Constant), OfficeInvest, OfficeJobs, DD6, DD7, DD2, DD4, DD3, DD8, LnLFAm2, Asbest, RentR, LnAge, LnAfstDBZ_IC_Station d. Dependent Variable: WOZ_CapRate ANOVAa Model 1
2
3
Sum of Squares
df
Mean Square
F
Sig.
Regression
,002
2
,001
7,383
,001b
Residual
,017
118
,000
Total
,019
120
Regression
,006
8
,001
7,211
,000c
Residual
,012
112
,000
Total
,019
120
Regression
,007
13
,001
5,447
,000d
Residual
,011
107
,000
Total
,019
120
a. Dependent Variable: WOZ_CapRate b. Predictors: (Constant), OfficeInvest, OfficeJobs c. Predictors: (Constant), OfficeInvest, OfficeJobs, DD6, DD7, DD2, DD4, DD3, DD8 d. Predictors: (Constant), OfficeInvest, OfficeJobs, DD6, DD7, DD2, DD4, DD3, DD8, LnLFAm2, Asbest, RentR, LnAge, LnAfstDBZ_IC_Station Coefficientsa
Model 1
2
Unstandardized Coefficients
Standardized Coefficients Beta
B
Std. Error
(Constant)
,018
,036
OfficeJobs
,001
,000
OfficeInvest
-,019
,008
(Constant)
,015
,033
Collinearity Statistics t
Sig.
,501
,618
Tolerance
VIF
,213
2,409
-,219
-2,475
,018
,963
1,038
,015
,963
1,038
,453
,651
114
3
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
OfficeJobs
,001
,000
,214
2,604
,010
,870
1,150
OfficeInvest
-,019
,007
-,215
-2,699
,008
,931
1,074
DD2
,005
,005
,075
,949
,345
,948
1,055
DD3
,015
,004
,315
3,952
,000
,930
1,075
DD4
,012
,004
,222
2,807
,006
,938
1,066
DD6
,011
,005
,158
2,027
,045
,969
1,032
DD7
-,005
,003
-,146
-1,799
,075
,896
1,116
DD8
,009
,004
,196
2,421
,017
,896
1,117
(Constant)
,067
,049
1,359
,177
OfficeJobs
,001
,000
,158
1,746
,084
,690
1,450
OfficeInvest
-,016
,007
-,183
-2,254
,026
,855
1,169
DD2
,007
,006
,113
1,274
,205
,716
1,397
DD3
,016
,004
,347
4,232
,000
,836
1,196
DD4
,012
,004
,234
2,936
,004
,886
1,128
DD6
,013
,006
,188
2,356
,020
,879
1,137
DD7
-,004
,003
-,138
-1,493
,138
,663
1,509
DD8
,014
,006
,302
2,472
,015
,378
2,648
LnLFAm2
,000
,001
,035
,437
,663
,870
1,150
RentR
-,006
,002
-,198
-2,407
,018
,830
1,205
LnAge
,001
,001
,170
1,830
,070
,652
1,534
LnAfstDBZ_ IC_Station
-,004
,003
-,174
-1,377
,171
,354
2,828
Asbest
,000
,004
-,003
-,035
,972
,851
1,175
Beta In
t
Sig.
Partial Correlation
Tolerance
VIF
Minimum Tolerance
DD2
,043b
,492
,624
,045
,980
1,020
,946
DD3
b
,300
3,562
,001
,313
,968
1,033
,933
DD4
,196b
2,280
,024
,206
,987
1,013
,952
DD6
,132
1,521
,131
,139
,995
1,005
,958
DD7
-,266b
-3,170
,002
-,281
,993
1,007
,957
DD8
,171
1,949
,054
,177
,952
1,050
,917
LnLFAm2
-,001b
-,009
,993
-,001
,973
1,027
,946
RentR
-,279
-3,260
,001
-,289
,950
1,053
,917
LnAge
-,018b
-,199
,842
-,018
,968
1,033
,933
LnAfstDBZ_ IC_Station
,138
1,587
,115
,145
,985
1,015
,949
Asbest
-,022b
-,254
,800
-,023
,992
1,008
,958
LnLFAm2
,033c
,406
,686
,038
,918
1,089
,862
RentR
-,204
-2,491
,014
-,230
,841
1,189
,823
LnAge
,107c
1,233
,220
,116
,774
1,293
,749
LnAfstDBZ_ IC_Station
-,121
-1,039
,301
-,098
,434
2,306
,434
Asbest
,004c
,049
,961
,005
,866
1,155
,816
a. Dependent Variable: WOZ_CapRate Excluded Variablesa Collinearity Statistics Model 1
2
b
b
b
b
c
c
a. Dependent Variable: WOZ_CapRate b. Predictors in the Model: (Constant), OfficeInvest, OfficeJobs c. Predictors in the Model: (Constant), OfficeInvest, OfficeJobs, DD6, DD7, DD2, DD4, DD3, DD8
Appendices
115
WOZ cap rate regression with year dummies Descriptive Statistics Mean
Std. Deviation
N
WOZ_CapRate
,09
,01
121
DY2004
,12
,33
121
DY2005
,07
,26
121
DY2006
,12
,33
121
DY2007
,20
,40
121
DY2009
,16
,37
121
DY2010
,07
,25
121
DY2011
,09
,29
121
DD2
,04
,20
121
DD3
,07
,26
121
DD4
,06
,23
121
DD6
,03
,18
121
DD7
,19
,39
121
DD8
,07
,26
121
LnLFAm2
5,93
1,40
121
RentR
,93
,42
121
LnAge
2,77
1,73
121
LnAfstDBZ_IC_Station
7,46
,52
121
Asbest
,07
,26
121
Model
Variables Entered
Variables Removed
Method
1
DY2011, DY2010, DY2005, DY2006, DY2004, DY2009, DY2007b
Enter
2
DD2, DD6, DD7, DD3, DD4, DD8b
Enter
3
LnLFAm2, Asbest, LnAge, RentR, LnAfstDBZ_IC_Stationb
Enter
Variables Entered/Removeda
a. Dependent Variable: WOZ_CapRate b. All requested variables entered. Model Summaryd Change Statistics Model
R
1
DurbinWatson
R Square
Adjusted R Square
Std. Error of the Estimate
R Square Change
F Change
df1
df2
Sig. F Change
,423a
,179
,128
,011623
,179
3,521
7
113
,002
2
,610
b
,372
,296
,010445
,193
5,485
6
107
,000
3
,648c
,420
,317
,010285
,048
1,673
5
102
,148
2,155
a. Predictors: (Constant), DY2011, DY2010, DY2005, DY2006, DY2004, DY2009, DY2007 b. Predictors: (Constant), DY2011, DY2010, DY2005, DY2006, DY2004, DY2009, DY2007, DD2, DD6, DD7, DD3, DD4, DD8 c. Predictors: (Constant), DY2011, DY2010, DY2005, DY2006, DY2004, DY2009, DY2007, DD2, DD6, DD7, DD3, DD4, DD8, LnLFAm2, Asbest, LnAge, RentR, LnAfstDBZ_IC_Station d. Dependent Variable: WOZ_CapRate
116
Va l u at i o n a c c u r a c y i n va c a n t o f f i c e p r o p e r t i e s
ANOVAa Model 1
2
3
Sum of Squares
df
Mean Square
F
Sig.
Regression
,003
7
,000
3,521
,002b
Residual
,015
113
,000
Total
,019
120
Regression
,007
13
,001
4,879
,000c
Residual
,012
107
,000
Total
,019
120
Regression
,008
18
,000
4,099
,000d
Residual
,011
102
,000
Total
,019
120
a. Dependent Variable: WOZ_CapRate b. Predictors: (Constant), DY2011, DY2010, DY2005, DY2006, DY2004, DY2009, DY2007 c. Predictors: (Constant), DY2011, DY2010, DY2005, DY2006, DY2004, DY2009, DY2007, DD2, DD6, DD7, DD3, DD4, DD8 d. Predictors: (Constant), DY2011, DY2010, DY2005, DY2006, DY2004, DY2009, DY2007, DD2, DD6, DD7, DD3, DD4, DD8, LnLFAm2, Asbest, LnAge, RentR, LnAfstDBZ_IC_Station
Coefficientsa
Model 1
2
3
Unstandardized Coefficients
Standardized Coefficients Beta
B
Std. Error
(Constant)
,092
,003
DY2004
,000
,004
DY2005
-,004
,005
DY2006
-,002
DY2007
-,006
DY2009 DY2010
Collinearity Statistics t
Sig.
35,418
,000
Tolerance
VIF
,004
,038
-,081
-,820
,970
,652
1,533
,414
,745
1,342
,004
-,062
,004
-,187
-,584
,561
,652
1,533
-1,648
,102
,567
1,764
,008
,004
,010
,005
,249
2,276
,025
,608
1,644
,204
2,098
,038
,765
1,307
DY2011
,004
,004
(Constant)
,089
,003
,089
,885
,378
,710
1,409
31,852
,000
DY2004
3,291E-05
,004
DY2005
-,001
,004
,001
,009
,993
,578
1,729
-,024
-,261
,795
,694
1,442
DY2006
-,002
DY2007
-,003
,004
-,052
-,534
,594
,616
1,623
,003
-,085
-,801
,425
,516
1,937
DY2009 DY2010
,010
,004
,280
2,716
,008
,552
1,812
,009
,004
,190
2,107
,037
,725
1,380
DY2011
,006
,004
,129
1,387
,168
,683
1,465
DD2
,005
,005
,081
1,019
,310
,925
1,081
DD3
,014
,004
,292
3,605
,000
,896
1,116
DD4
,012
,004
,228
2,769
,007
,867
1,153
DD6
,010
,005
,144
1,817
,072
,936
1,068
DD7
-,004
,003
-,119
-1,435
,154
,850
1,176
DD8
,010
,004
,204
,847
1,181
(Constant)
,122
,023
DY2004
,002
,004
DY2005
3,319E-05
,005
DY2006
-,002
DY2007
-,001
2,450
,016
5,218
,000
,057
,511
,610
,454
2,205
,001
,007
,994
,609
1,642
,004
-,044
-,429
,669
,551
1,813
,003
-,029
-,263
,793
,470
2,126
Appendices
117
DY2009
,009
,004
,252
2,330
,022
,487
2,053
DY2010
,009
,005
,177
1,924
,057
,672
1,488
DY2011
,004
,004
,099
1,043
,299
,638
1,568
DD2
,008
,006
,132
1,437
,154
,678
1,475
DD3
,016
,004
,334
3,917
,000
,781
1,280
DD4
,014
,004
,261
3,092
,003
,798
1,254
DD6
,012
,006
,177
2,169
,032
,852
1,173
DD7
-,004
,003
-,133
-1,385
,169
,621
1,611
DD8
,015
,006
,324
2,564
,012
,355
2,814
LnLFAm2
,000
,001
,043
,517
,607
,806
1,240
RentR
-,004
,003
-,144
-1,583
,117
,687
1,455
LnAge
,001
,001
,192
2,022
,046
,631
1,584
LnAfstDBZ_ IC_Station
-,005
,003
-,203
-1,531
,129
,322
3,102
Asbest
,001
,004
,016
,181
,856
,763
1,311
a. Dependent Variable: WOZ_CapRate
Excluded Variablesa Collinearity Statistics Model 1
2
Beta In
t
Sig.
Partial Correlation
Tolerance
VIF
Minimum Tolerance
DD2
,041b
,467
,641
,044
,965
1,036
,562
DD3
b
,263
3,097
,002
,281
,936
1,069
,565
DD4
,203b
2,344
,021
,216
,928
1,078
,557
DD6
,123
1,419
,159
,133
,952
1,051
,564
DD7
-,237b
-2,794
,006
-,255
,949
1,054
,566
DD8
,171
1,925
,057
,179
,902
1,109
,547
LnLFAm2
,022b
,242
,809
,023
,912
1,097
,566
RentR
-,222
-2,361
,020
-,218
,790
1,266
,552
LnAge
,021b
,238
,812
,022
,925
1,082
,567
LnAfstDBZ_ IC_Station
,144
1,618
,109
,151
,902
1,108
,567
Asbest
-,009b
-,095
,925
-,009
,906
1,103
,559
LnLFAm2
,050c
,598
,551
,058
,856
1,168
,515
RentR
-,161
-1,763
,081
-,169
,693
1,443
,496
LnAge
,140c
1,577
,118
,151
,736
1,359
,516
LnAfstDBZ_ IC_Station
-,125
-1,012
,314
-,098
,384
2,606
,384
Asbest
,013c
,153
,879
,015
,782
1,279
,513
b
b
b
b
c
c
a. Dependent Variable: WOZ_CapRate b. Predictors in the Model: (Constant), DY2011, DY2010, DY2005, DY2006, DY2004, DY2009, DY2007 c. Predictors in the Model: (Constant), DY2011, DY2010, DY2005, DY2006, DY2004, DY2009, DY2007, DD2, DD6, DD7, DD3, DD4, DD8
