Cand.merc. (M.Sc.

) i Finansiering og Regnskab
Cand.merc. (M.Sc.) in Finance and Strategic Management
Master Thesis 2020

Why Include Real Estate and Especially REITs

in Multi-Asset Portfolios?

Authors:
Elias Wiklund Joachim Hansen Flood
101871 102530

Supervisor:
Jens Lunde

Number of pages: 96
Number of characters (with spaces): 173.338

May 15th, 2020

Copenhagen
Abstract
This thesis has explored the characteristics of real estate as a financial asset in regards
to portfolio building, with a particular focus on REITs. This was done to clarify if real
estate should be included in large-cap investors’ portfolios, and when that is the case,
what allocation.

Extensive qualitative analysis has been performed, mainly examining existing

research in both the field of real estate and portfolio theory. The financial implications
of investing in REITs versus buying properties the traditional way (direct) has been
explored, as has the macroeconomic implications for different real estate segments.

The asset allocations of the most prominent large-cap investors have been assessed.
Not surprisingly, all the largest asset managers have allocated some funds to real
estate. An unexpected finding was how dissimilar the asset managers’ portfolios were,
despite having the same goal of growing the capital they are managing for investors.
The disagreement within the asset management industry indicates that further
research on the field has relevance.

Quantitative analyses of the U.S. public stock- and bond markets have been conducted,
where returns, volatility, and correlations were examined. Based on this, historical
tangency portfolios were found, demonstrating what weight of real estate in
combination with other asset classes has yielded the best risk-adjusted returns in the
past. The analysis used data from 1994 until 2020, making it evident that over different
historical periods, different allocations to REITs, stocks, and bonds were optimal.
However, the spread was not too substantial: Using data from the last 20-years, a 14%
allocation to REITs generated the maximum risk-adjusted return, while considering
the last five years the optimal allocation to REITs was 15%.

The thesis has investigated the effects of the ongoing COVID-19 pandemic. Industry
segments relying on non-essential traveling, especially hotels and shopping malls,
have been critically struck by the crisis. Still, as of May 1st, 2020, many real estate
segments have not been affected directly. REITs on the other hand, have been strongly
affected by the crisis, and are down more than the market portfolio across the publicly
traded U.S. markets. Reasons for this, as well as the potential long-term effects on the
market, will also be discussed in the thesis.

1
 Table of Contents
1 Introduction ............................................................................................................ 4
1.1 Research question ............................................................................................. 4
1.2 Motivation ......................................................................................................... 4
1.2.1 Motivation for focusing on REITs ............................................................. 5
1.3 Delimitations .................................................................................................... 6
1.3.1 COVID-19 related delimitations ................................................................ 6
2 Methodology ........................................................................................................... 7
2.1 Review of the data collection ............................................................................ 7
2.2 Models for use in research ................................................................................ 8
3 Theory ................................................................................................................... 10
3.1 Markowitz’ Portfolio Theory........................................................................... 10
3.2 Sharpe .............................................................................................................. 11
3.3 Friedman ......................................................................................................... 12
3.4 CAPM .............................................................................................................. 12
3.5 Multifactor models ......................................................................................... 14
3.5.1 Fama & French Three-Factor Model ....................................................... 14
3.5.2 Fama & French Five-Factor model .......................................................... 16
3.5.3 Carhart’s Four-Factor model ................................................................... 17
3.6 Real Estate Investment Trusts (REITs) ......................................................... 18
3.7 Valuing Real Estate ......................................................................................... 19
3.7.1 Damodaran 2012...................................................................................... 19
4 Data ....................................................................................................................... 21
4.1 Stock Market ................................................................................................... 21
4.2 Bond market ................................................................................................... 22
4.3 Real Estate Market ......................................................................................... 24
5 Analysis ................................................................................................................. 27
5.1 Qualitative analysis ......................................................................................... 27
5.1.1 REITs VS direct real estate investments ................................................. 27
5.1.2 Analysis of existing asset managers......................................................... 32
5.1.3 Macroeconomic factors ............................................................................ 37
5.1.4 The COVID-19 crisis ................................................................................ 41
5.2 Quantitative analysis ...................................................................................... 58
5.2.1 All REIT Index ......................................................................................... 58
5.2.2 Sector Analysis ......................................................................................... 77

2
 5.2.3 Multi-Factor Regression .......................................................................... 85
6 Conclusions ........................................................................................................... 92
6.1 Alternative ways to invest in real estate ......................................................... 92
6.2 Optimal allocation for REITs in mixed-asset portfolios ................................ 92
6.3 How do the sub-sectors of REITs contribute to portfolios? .......................... 93
6.4 What investor types benefit more or less from real estate? ........................... 93
6.5 What are the effects of COVID-19?................................................................. 94
6.6 Final Conclusion ............................................................................................. 95
7 Appendix ............................................................................................................... 97
7.1 Appendix 1, Tangency portfolios, returns, volatilities and Sharpe ratios of
ALL REIT index, All stocks index, and bonds index ................................................ 97
7.2 Portfolios of REIT sectors with various constraints, 2000-2020 .................98
7.3 Industry Classification ....................................................................................98
8 Bibliography .......................................................................................................... 99

3
1 Introduction
1.1 Research question
Most mutual funds, hedge funds, and other large-cap investors have allocated a part
of their portfolio in real estate. Typical pension funds allocate 6% of their portfolio to
real estate, while typical hedge funds allocate 3.5% (Andonov, Eichholtz, & Kok, 2015).
This thesis will explore the reasons why large-cap investors should choose to invest in
real estate, as well as different strategies investors can apply when doing so. It will also
look into the strategies that investors can apply for building a mixed-asset portfolio
with a real estate component.

Hence, our research question is:

Why should large-cap investors include REITs in their mixed-asset portfolios?

To answer this, we will address the following sub-questions:

 What alternative ways of real estate investment exist, and what are the
strengths and weaknesses of the different alternatives?
 Using modern portfolio theory, how much weight should be allocated to REITs
in a mixed-asset portfolio?
 How can different sub-sectors of REITs contribute to portfolios?
 What investor types will benefit more or less from a larger real estate
allocation?
 What effect can the ongoing COVID-19 crisis be expected to have on the real
estate market, and the chosen portfolio?

1.2 Motivation
Real estate has always been a fundamental asset class. It is fundamental in ways many
industries are not, in that people need real estate for a place to live, a place to work,
and to recreate. Besides, companies need real estate for the production of goods,
services, and logistics. Because real estate is considered essential and permanent, it is
often regarded as less risky than stocks. Since the global financial crisis, many western
countries have lowered their interest rates close to, or even below zero. The weak bond
yields of today’s economy make real estate investments even more interesting. We
realize that real estate investments are perhaps more relevant now than ever and that

4
more asset managers potentially could benefit greatly from having more empirical
analysis and clarity brought to the field. Because of this, we seek to discuss the various
reasons to include real estate in multi-asset portfolios and to determine how this can
be done successfully.

While working on the thesis, a global health crisis, COVID-19 hit the U.S. The number
of infected people and deaths are rising quickly, and we are observing the pandemic
affect more or less every part of our society. People are encouraged to stay in their
homes, many businesses are shut down, workers are being laid off, and the general
uncertainty is high. Not surprisingly in all this chaos, stock prices across more or less
all asset classes are plunging. In this sudden bear market, most asset managers are
seeing their portfolios lose value, and are more than normally looking for safe havens
and hedges. Because of the imminent impact the COVID-19 crisis is causing the
investors and portfolios analyzed in our thesis, we want to investigate the crisis’ effect
on the real estate sector as a whole, and with an extra focus on REIT investments and
mixed-asset portfolios.

1.2.1 Motivation for focusing on REITs

There are several reasons for this thesis’ focus on REITs. Technical reasons aside,
REITs have emerged as an increasingly exciting financial asset. Its purpose to make
diversified real estate investments more accessible for both institutional and retail
investors is a clever way to expand the real estate investment pool and to make the
market more liquid. The first REITs emerged in the USA in the 1960s, but the asset
class has grown a lot in popularity over the last decades. This can be seen in the
development of REIT sizes and the number of REITs. The total market capitalization
of all REITS in the U.S. was USD 8.74bn in the year 1990. In January 2020, the market
cap was more than 100 times bigger, totaling USD 1,300bn (Nareit, 2020a).

From a portfolio analysis stance, there are technical reasons why REITs are better
suited than private market real estate in regards to comparison with other asset
classes. This has to do with the price-making mechanisms of REITs contra private
market investments. These technical reasons are elaborated in the theory chapter and
under delimitations.

5
1.3 Delimitations
In 1971, Harris C. Friedman proved that Markowitz’s mean-variance portfolio theory
is applicable for real estate investments and mixed-asset portfolios as long as return
and risk can be quantified (Friedman, 1971). For publicly traded real estate investment
trusts (REITs), risk and return measures are just as quantifiable as for publicly traded
stocks. Because of this, we largely focus on REITs for portfolio composition. Hence,
the results of the quantitative analysis are only applicable for REITs, and should not
be used as a proxy for the real estate market as a whole. However, private market
investments will be included in the qualitative analysis.

To access the widest array of publicly-traded REITs and public real estate companies
in different sub-industries, this thesis is limited to the U.S. market. Still, the findings
will potentially apply to other similar markets.

The REITs used in the analysis are publicly traded on U.S stock exchanges NYSE,
NASDAQ, and AMEX. It is recognized that when comparing these REITs to the stock
market as a whole, the REITs are part of the market they are compared with. However,
the REITs’ market capitalization is small compared to the market as a whole. All the
companies traded on the NYSE and NASDAQ exchanges have a combined market
capitalization of approximately 40 trillion dollars (Nasdaq, 2019). Ergo, the market
capitalization of the entire REIT segment amounts to about 3.25% of the total stock
market, about the same weight as the largest individual corporations: Microsoft and
Apple Inc., (Yahoo Finance, 2020b). Thus, the REIT returns will not be able to make
a significant impact on the total stock market returns, and comparisons with the stock
market are still viable.

1.3.1 COVID-19 related delimitations

This thesis was produced amidst the COVID-19 crisis. Because of this, the end outcome
of the crisis is not known, and hence not part of the thesis. Because of this, all analysis
of the long term impacts of COVID-19 are based on expert predictions and forecasts,
and should not be seen as absolute facts. All discussion concerning the coronavirus’
impact on the real estate market, as well as other markets, is based on the impact
already seen at the beginning of May 2020, as well as reports, journals and other
writings offering forecasts and expectations about the virus’ effects, also produced
prior to May 15th, 2020. This should be taken into consideration when reading the
analysis and conclusions regarding the COVID-19 crisis and its potential aftermath.

6
2 Methodology
The methodology chapter seeks to answer how the thesis’ research question and sub-
questions will be answered. It is doing so by illuminating the process of examining the
real world for empirical data that can be used for drawing general conclusions. Hence,
the chapter will review the methodological entry to the thesis’ collection of data, and
provide some discussion about the selection of models and theories. It will then
provide a review of the quality of the data collection.

2.1 Review of the data collection

When evaluating the quality of data, two central elements should be taken into
consideration: Validity and reliability. Validity is defined as the extent a concept is
accurately measured in a quantitative study (Heale & Twycross, 2015). Reliability
describes the accuracy of a measurement. Hence, a reliable analysis can be repeated,
by the same or other analysts, and the result will be the same (Bitsch Olsen, 2003). As
extern analysts, without any inside information of the market or real estate industry,
the analysis has been based on public information. In some instances, data that would
have been helpful for the analysis is not available. In these instances, the analysis has
to rely on less reliable data and/or assumptions. Such conditions affect the connection
between the collected data and the thesis’ conclusions negatively, hurting the validity
of the conclusions. This is to some degree unavoidable, but the thesis’ delimitations
help to specify for what areas the conclusions are, and are not, applicable. This
contributes to making the answers to the research questions more reliable and viable,
despite the mentioned challenges.

As the thesis is dependent on publicly available information, another important

element of the thesis’ reliability is the controlling of extern sources. Part of this has
been to minimize the use of internet articles as sources. The reason being, that the
original source can be opaque or difficult to find, making the reliability weak. Many
sources from the internet have been used, but these are transparent sources, where the
author and date are cited, or where the author is deemed a trustworthy organization.
Examples of the first are professors and renowned industry experts such as
Dartmouth’s Kenneth French, MIT’s William Wheaton, and NYU’s Aswath
Damodaran. These professors have not only written books but journals and scientific
articles. These articles are indeed considered highly reliable despite being found
online. Examples of the latter are the SEC, Nareit, and State Street Global Advisors.

7
Even governmental and non-profit organizations can have biases or agendas, but the
requirements and responsibility of these organizations make the authors of this thesis
trust that the numbers and analyses provided by said corporations are reliable.
Another element that helps to minimize the risk of bias, is the use of several sources
supporting each analysis. The notion being that one source’s reliability could be
questionable, but if the same data is confirmed by other unrelated sources, the
reliability of the data is significantly strengthened (Bitsch Olsen, 2003). This thesis
strives to verify as much data as possible with the use of numerous unrelated sources
throughout the analysis.

2.2 Models for use in research

To make conclusions and predictions based on historical observations, one has to rely
on models. This segment will discuss modeling in broad terms and highlight the
potential pitfalls in the modeling process.

A model, how it is used in this thesis, is a simplified representation of something in the

real world. This usage implies that a model is always a representation that is less than
perfect. Given there is something that calls for investigation in the real world, here
called the “real system”, and there is a “problem” related to the real system which calls
for definite “conclusions”, the modeling process can be depicted as in figure one
(Phillips, Ravindran, & Solberg, 1976).

Figure 1: Process of modeling

Source: (Phillips, Ravindran, & Solberg, 1976)

8
The first step is the construction of the model itself; Formulation. This is based upon
an inductive generalization and is a highly important part of the analysis. It requires
decisions as to what aspects of the real system should be incorporated in the model,
what can be left out, and what assumptions should be made. Because of these
assumptions, it follows that the formulation step will include a certain amount of
arbitrariness.

Once the formulation is made, the process gets more scientific. The “deduction” step
involves computation and expressing a sequence of logical statements. This part of the
problem will not be subject to differences in the analysts’ opinion, as long as the
assumptions of the model are well defined. It follows from this that nobody knowing
the formulation and assumptions of the model should be able to question the model
conclusion. In this thesis, the model conclusion includes absolute answers such as
specific portfolio weights and returns of real estate, stocks, and bonds.

The model conclusion cannot, uncritically, provide absolute answers for the real
system. That is why the last process of modeling is “Interpretation”. This part of the
process, just like the formulation, involves human judgment. This involves the
consideration of the results of the model are realistic and plausible. It is important to
review if any factors that have not been taken into consideration in the modulation
could alter the conclusion. This means that aspects of the real system that were
overlooked, ignored, or discarded in the formulation phase of the model may turn out
to be important in the interpretation. Thus, it will improve a model to continuously
reconsider and update the formulation thorough “testing and revision” (Ackoff, 1962).
Readers are encouraged to consider this while interpreting the models presented in
the thesis.

As the connection between the real system and the model is formulation and inductive
generalization, Black swan events like the Covid-19 brings huge ramifications to the
world. With such severe and atypical movements, models will typically do a poor job
of estimating the consequences of the movement. Models that are built on historical
data tend to give model conclusions that are easily transferable to real conclusions
when the parameters continue to move similar to how they have done in the past. The
historical input of this thesis’ model, as well as any other model, does not have any
historical scenarios resembling what is happening because of the COVID19 virus.
Thus, the quantitative model made in this thesis, as well as other models used to

9
forecast the future based on historical data, will likely not be able to produce a credible
forecast for the effects of the COVID-19 crisis. Because of this, the thesis will not use
quantitative modeling to try to forecast the effects of COVID-19 on the real estate
market. Instead, this highly relevant topic will be discussed as part of the qualitative
analysis.

3 Theory
The theory chapter presents, explains, and discusses the core theories used in the
thesis’ further analysis. Hence, the chapter lays the theoretical foundation that the rest
of the thesis is built on. Most of the theories discussed are from the field of finance and
portfolio theory, but literature about real estate valuation and REITs are also explored.

3.1 Markowitz’ Portfolio Theory

For the portfolio theory, the thesis assumes that investors seek to maximize their
return for any given level of risk. Hence, their investment decisions are based on
modern portfolio theory (MPT). This theory was introduced by Harry Markowitz in
1952. By creating a method for quantifying the risk and return of any combination of
assets, Markowitz allows investors to construct the highest yielding portfolio for any
risk preference, or the or the least risky portfolio for any return preference. This is also
known as the mean-variance portfolio (MVP).

Markowitz’s portfolio theory builds on the basic perception that a diversified portfolio
is less risky than an undiversified one, because the value fluctuations of the assets in
the portfolio to some extent will weigh each other out, thus reducing the combined
variance. However, Markowitz takes the theory one step further by quantifying the
diversification benefit of portfolios by analyzing the variance and covariance of the
assets in the portfolio (Markowitz, 1952). All possible combinations of assets that
generate maximum return per risk create portfolios located on “the efficient frontier”,
and are regarded as efficient portfolios. The efficient frontier is illustrated in figure
two. As the efficient frontier represents the optimal portfolios in the tradeoff between
risk and return, it is impossible to find a portfolio that is above the frontier. The
portfolios under the frontier are not compensated sufficiently for their risk and are
therefore inefficient portfolios.

10
Figure 2: Efficient Frontier

Source: (Markowitz, 1952)

Markowitz’s portfolio theory is widely accepted but has a drawback in its assumption
of a normal distribution of returns. Standard deviation is the only measure of risk in
the model and is measured as movements in both directions. Thus, with a skewed
distribution, tail losses will not be considered.

3.2 Sharpe
The risk-return relationship described by Markowitz was later used as a performance
measurement for mutual funds by William Sharpe (Sharpe W. F., 1966). Sharpe
named this the “reward-to-variability-ratio” but it is widely known as the “Sharpe
ratio”. The Sharpe ratio provides a risk-adjusted performance measure by dividing the
excess return of a risky security by the standard deviation of said security.

𝑝𝑜𝑟𝑡𝑓𝑜𝑙𝑖𝑜 𝑟𝑒𝑡𝑢𝑟𝑛 − 𝑟𝑖𝑠𝑘 𝑓𝑟𝑒𝑒 𝑟𝑎𝑡𝑒

𝑆ℎ𝑎𝑟𝑝𝑒 𝑅𝑎𝑡𝑖𝑜 =
𝑝𝑜𝑟𝑡𝑓𝑜𝑙𝑖𝑜 𝑣𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦

11
Thus, the ratio measures how well investors are compensated for the risk they are
taking. It is one of the most commonly used ratios for comparing portfolios.
Furthermore, the Sharpe ratio is used for building portfolios with the Markowitz
method, as the combination of assets maximizing the Sharpe ratio creates the tangency
portfolio, the point where the capital allocation line (CAL) meets the efficient frontier.

A drawback of the Sharpe ratio is that it, just like the MPT, assumes normally
distributed returns. Furthermore, the Sharpe ratio produces a numerical value of the
return to risk relationship, which can be difficult to interpret without comparison with
other assets.

3.3 Friedman
Friedman (1971) shows that the models of Markowitz’s modern portfolio theory, used
for constructing stock-portfolios, also can be applied for real estate and mixed-asset
portfolios. The concept of selecting real estate assets are identical to the selection of
stocks, as long as the risk and return can be quantified.

3.4 CAPM
Markowitz's modern portfolio theory laid the foundation for the development of one
of the first and most influential factor models, the Capital Asset Pricing Model
(CAPM). The CAPM model is based on the work of William Sharpe (1964), John
Lintner (1965), and Jan Mossin (1966) who each separately published papers that
contributed to the formulation of the model. The model is still widely used to find the
cost of capital when valuing assets.

The model differentiates the risk of an asset, between firm-specific risk and market
risk. Market risk, called systematic risk by Sharpe, is defined as the risk that it is not
possible to eliminate by diversification. The model assumes the investor is fully
diversified and therefore has eliminated all firm-specific risk and deems only
systematic risk as a relevant risk measure.

The CAPM model describes the linear relationship between systematic risk and
expected return. It assumes that investors expect to be compensated with a higher

12
expected return for taking on more risk. Systematic risk is usually denoted as beta in
the CAPM model.

The CAPM prices assets based on two risks that investors face in the model: time value
of money and sensitivity to markets risk. The formula compensates investors for the
time value of money by adding the risk-free rate as a minimum expected rate of return.
Exposure to systematic risk is compensated by adding a risk premium on the excess
return on the market portfolio. This risk premium is denoted as beta and is a measure
of sensitivity to volatility in the market portfolio. Market beta is given by the formula:

𝑐𝑜𝑣(𝑟𝑖 , 𝑟𝑀 )
𝛽𝑖 =
𝑣𝑎𝑟(𝑟𝑀 )

The CAPM formula is defined as:

𝐸[𝑟𝑖 ] = 𝑟𝑓 + 𝛽𝑖 (𝐸[𝑟𝑀 ] − 𝑟𝑓 )

Where:

𝐸[𝑟𝑖 ] = 𝐸𝑥𝑝𝑒𝑐𝑡𝑒𝑑 𝑟𝑒𝑡𝑢𝑟𝑛 𝑜𝑓 𝑎𝑠𝑠𝑒𝑡 (𝑖) 𝑟𝑓 = 𝑅𝑖𝑠𝑘 𝑓𝑟𝑒𝑒 𝑟𝑎𝑡𝑒

𝐸[𝑟𝑀 ] = 𝐸𝑥𝑝𝑒𝑐𝑡𝑒𝑑 𝑟𝑒𝑡𝑢𝑟𝑛 𝑜𝑓 𝑚𝑎𝑟𝑘𝑒𝑡 𝑝𝑜𝑟𝑡𝑓𝑜𝑙𝑖𝑜 𝛽𝑖 = 𝐵𝑒𝑡𝑎 𝑜𝑓 𝑎𝑠𝑠𝑒𝑡 (𝑖)

The market portfolio has a beta of 1 and is the average beta of the market. A higher
than 1 indicates a higher sensitivity towards volatility in the market portfolio than
average. An asset with a beta of 1,5 will increase 150 basis points when the market
increases 100 basis points. In the CAPM model a higher beta indicates a riskier asset
and requires a higher return.

The CAPM model is built on several underlying assumptions.

1. There are no transaction costs or taxes

2. All assets are marketable and can be divided into infinitely small parts.
3. All investors are price takers and do not affect the price by their own trading.
4. Investors act rationally, making decisions solely on risk-return assessment.
5. Investors can, without limitation, engage in short-sale.
6. Investors can lend and borrow unlimitedly at the risk-free rate.
7. All investors have homogenous time frames and expectations towards returns,
variances, and covariances.

13
The assumptions imply that all investors agree on the risk-free rate and the efficient
frontier of risky assets (Munk, 2018, s. 293). All investors, therefore, invest in the
market portfolio and the risk-free asset.

The CAPM model has been criticized for being untestable because the true market
portfolio is practically unobservable. Roll (1977) states that stock exchange indices are
only a partial measure of the true global market portfolio and that it would need to
consist of every security in the world to be true. The CAPM model has also been
criticized by several researchers, among them Fama and French (1992), for being too
simplistic to describe the cross-section of expected returns.

3.5 Multifactor models

3.5.1 Fama & French Three-Factor Model
Fama and French introduced in 1992 a multi-factor model that similarly to CAPM aims
to predict the expected return of stocks. Fama and French expanded the model with
two additional factors intended for explaining patterns in average returns that were
not explained by the CAPM model.

By the late 1980s several researchers had found market patterns that contradicted the
CAPM model. Banz (1981) identified a negative relationship between the size of a
company and average returns. Meaning that firms with large market capitalization
perform on average worse than firms with a smaller market capitalization. Small firms
had average returns that were too high according to the CAPM model, given their beta
estimates.

Stattman (1980) and Rosenberg, Reid, and Lanstein (1985) discover evidence of
another anomaly that is contradicting CAPM. Their research show that there is a
positive relationship between the average returns of US stocks and the firm’s book-to-
market ratio. Both anomalies in expected return that researchers have found confirm
that there are other characteristics than just sensitivity to the market portfolio, which
explains the expected return of stocks and that CAPM is not a sufficient model.

Fama & French’s three-factor (FF3) model uses two additional factors for size and
value in addition to the market factor of CAPM, to better explain stock performance.
The two new factors are constructed by creating two portfolios that mimic the

14
performance of size and value stocks. The portfolios are constructed by sorting stocks
into two Size groups and three Value groups (independent 2x3 sorts). The Size group
is divided by the median market cap of the market (Fama & French used NYSE stocks).
The Value group is broken up by the 30th and 70th percentile of B/M for stocks in the
market. The Size factor, called small-minus-big (SMB), is the average of the three
small stock portfolio returns minus the average of the three big stock portfolio returns.
The Value factor, high-minus-low (HML), is the average return of the two high B/M-
portfolios minus the two low B/M-portfolios.

The two new factors are used similarly to how the stocks beta coefficient is used to
predict the expected returns in the CAPM model. Stocks have different sensitivity
towards SMB and HML factors, and the coefficient to these variables will affect the
expected return of individual stocks in the FF3 model. The three-factor model time
series regression is formulated as (Fama & French, 1996):

𝑅𝑖,𝑡 − 𝑅𝑓,𝑡 = 𝑎𝑖 + 𝛽𝑖 (𝑅𝑀𝑡 − 𝑅𝑓,𝑡 ) + 𝑠𝑖 𝑆𝑀𝐵𝑡 + ℎ𝑖 𝐻𝑀𝐿𝑡 + 𝑒𝑖,𝑡

𝑅𝑖𝑡 = return of a stock or portfolio at time t 𝑆𝑀𝐵 = size premium

𝑅𝑓,𝑡 = risk-free rate of return at time t 𝑠𝑖 = size factor exposure
𝑅𝑀𝑡 = return of market portfolio at time t 𝐻𝑀𝐿 = value premium
𝛽𝑖 = beta of asset (i)(market factor exposure) ℎ𝑖 = value factor exposure
𝑎𝑖 = intercept (zero if factors capture all variation) 𝑒𝑖,𝑡 = residuals (assumed=0)

Testing of the model by Fama and French found that the FF3-model had an average
𝑅 2 of 0,93 on regressions on 25 size-value portfolios (Fama & French, 1996). This
means that on average 93% of the variation in the portfolio was explained by the three
factors. They also find that the addition of the two factors captures most of the patterns
in average return that CAPM did not explain.

Fama and French (1993) suggest the model find any application that requires
estimates of expected stock returns. This includes estimating the cost of capital and
measuring abnormal returns. In 1996, Fama and French published evidence of the
three-factor model's superior explanation of returns compared to the CAPM model.

15
Their research shows that their model produces three to five times fewer pricing errors
than the CAPM model (Fama & French, 1996).

The FF3 model has been criticized for being an incomplete model for expected returns,
and several researchers have proposed adding additional factors to the model. The
most noteworthy works are by Mark Carhart (1997) and Fama & French themselves
who expand their model in response to criticism in 2015.

3.5.2 Fama & French Five-Factor model

Several researchers have published evidence of shortcomings in the three-factor
model, and that there are several prominent patterns in average returns that the model
does not explain. Novy-Marx (2013) Titman, Wei, and Xie (2004) and others have
found evidence that the three factors miss much of the variation in average returns
related to profitability and investment.

The published evidence motivated Fama and French themselves to update and add
additional factors to their model. Novy-Marx (2013) showed that firms with a high
profitability-to-book-equity ratio earn significantly higher average returns than
unprofitable firms. Titman et al., (2004) found that firms that invest aggressively
subsequently achieve negative benchmark-adjusted returns. In response to this, Fama
and French include two new factors for profitability and investment, and thereby
creating a five-factor model. (Fama & French, 2015).

The profitability factor is constructed as the difference between the returns on

diversified portfolios of stocks with robust and weak profitability and is called Robust-
Minus-Weak (RMW). Profitability is measured by the ratio of a firm’s profits to its
assets. Profits are the annual revenues minus cost of goods sold, interest, selling,
general and administrative expenses, all is then divided by book equity.

The investment portfolio is calculated as the difference between returns on diversified

portfolios of stock with high and low investments and is called Conservative-Minus -
Aggressive (CMA). A firm’s investment level is determined by its growth in total assets
in the previous fiscal year.

In Fama and French’s paper introducing the five-factor model, the factor portfolios are
constructed in three different ways. The portfolios that are used in this paper are
sorted as 2 × 3. Where the size factor is divided into two, and the remaining are sorted
in three. In practice, stocks on the NYSE are first divided into two, along with the

16
median market value into groups of small and big firms. For each size, group stocks
are again sorted into three groups by the 30th and 70th percentile, for high, neutral, and
low according to the specific ratios. The HML, RMW, and CMA portfolios are the
average return on the two high portfolios minus the two small portfolios, and the
neutral stocks are excluded. RMW, for example, is calculated as:

𝑅𝑀𝑊 = 1/2(𝑆𝑚𝑎𝑙𝑙 𝑅𝑜𝑏𝑢𝑠𝑡 + 𝐵𝑖𝑔 𝑅𝑜𝑏𝑢𝑠𝑡) − 1/2(𝑆𝑚𝑎𝑙𝑙 𝑊𝑒𝑎𝑘 + 𝐵𝑖𝑔 𝑊𝑒𝑎𝑘)

Since the other three factors are sorted on size, the SMB is divided into three portfolios,
each sorted on the three other factors. The profitability size factor is calculated as:

𝑆𝑀𝐵(𝑅𝑀𝑊) = 1/3(𝑆𝑚𝑎𝑙𝑙 𝑅𝑜𝑏𝑢𝑠𝑡 + 𝑆𝑚𝑎𝑙𝑙 𝑁𝑒𝑢𝑡𝑟𝑎𝑙 + 𝑆𝑚𝑎𝑙𝑙 𝑊𝑒𝑎𝑘) − 1/3(𝐵𝑖𝑔 𝑅𝑜𝑏𝑢𝑠𝑡

+ 𝐵𝑖𝑔 𝑁𝑒𝑢𝑡𝑟𝑎𝑙 + 𝐵𝑖𝑔 𝑊𝑒𝑎𝑘)

The full SMB is the average return of the three SMB portfolios:

𝑆𝑀𝐵 = 1/3(𝑆𝑀𝐵(𝐻𝑀𝐿) + 𝑆𝑀𝐵(𝑅𝑀𝑊) + 𝑆𝑀𝐵(𝐶𝑀𝐴) )

The five-factor model time series regression equation is:

𝑅𝑖,𝑡 −𝑅𝑓,𝑡 = 𝑎𝑖 +𝛽𝑖 (𝑅𝑀𝑡 − 𝑅𝑓,𝑡 )+𝑠𝑖 𝑆𝑀𝐵𝑡 +ℎ𝑖 𝐻𝑀𝐿𝑡 + 𝑟𝑖 𝑅𝑀𝑊 + 𝑐𝑖 𝐶𝑀𝐴 + 𝑒𝑖,𝑡

Where 𝑟𝑖 and 𝑐𝑖 are the factor exposure to the profitability and investment portfolios.
Fama and French found the five-factor model to explain 71%-94% of the cross-section
variance of expected return for the size, value, profitability, and investment portfolios.

3.5.3 Carhart’s Four-Factor model

Fama and French’s five-factor model has been criticized for leaving out a momentum
factor (Blitz, 2018). The momentum factor of Carhart is widely accepted within
academia and is one of the most notable expansions of the original three-factor model.
According to ROBECO (2018) many researchers will end up adding a momentum
factor to the five-factor model, which is why we in this paper also will include a
momentum factor.

Research by Jegadeesh and Titman (1993) show that strategies that go long on past
winners and short past losers realize significant abnormal returns. The abnormal
returns were found to not be due to their systematic risk or other common factors
(such as SMB or HML). The discovered effect on returns is only present short-term

17
and dissipates within two years. Based on these findings, Carhart (1997) developed a
model that added a momentum factor to the Fama & French three-factor model. The
momentum portfolio is constructed by Carhart as the average equal-weighted return
of 30% best performing firms the past 11 months minus the 30% worst-performing
stocks in the same period. The portfolio includes all NYSE, AMEX, and Nasdaq stock,
and are re-formed monthly to get a rolling momentum effect.

The time series regression of the new model has the equation:

𝑅𝑖,𝑡 − 𝑅𝑓,𝑡 = 𝑎𝑖 + 𝛽𝑖 (𝑅𝑀𝑡 − 𝑅𝑓,𝑡 ) + 𝑠𝑖 𝑆𝑀𝐵𝑡 + ℎ𝑖 𝐻𝑀𝐿𝑡 + 𝜌𝑖 𝑀𝑂𝑀 + 𝑒𝑖,𝑡

Where MOM is the momentum factor, and 𝜌𝑖 is the momentum factor exposure. The
momentum factor is also sometimes referred to as UMD for up-minus-down.

3.6 Real Estate Investment Trusts (REITs)

Creating a diversified real estate portfolio is an expensive, labor-intensive, and time-
consuming process. Because of this, REITs were established in the U.S. in the 1960s
to give retail investors access to diversified commercial real estate. It does so by
pooling commercial real estate into financial entities and allowing investors to buy
shares, thus lowering the bar for people and institutions to invest. Essentially, a REIT
is a company that owns and operates rent-generating commercial properties. REITs
can either buy or develop its properties -the important thing is that it operates them.
Thus, real estate developers, construction companies, and other companies developing
properties for sale will by law not qualify as REITs. Most REITs own equity, typically
financed with 40-60% debt (Sando, 2012) but some REITs also invest in real estate
debt and mortgages.

The first REIT index was created in 1977. Since then, REITs have had an average yearly
return of 10.8%. Over the same period, the S&P 500’s yearly return was 11.0% (Sando,
2012). In other words, based on historical returns, listed real estate and listed stocks
seem to offer more or less identical returns in the long run. However, there is a big
difference in the payout structure of these returns: Most REITs generate significantly
higher dividend yields and lower value growth than typical stocks. Out of the ~11%
annual returns generated from REITs and the S&P500, the REIT returns consisted of
76% dividends and 24% value growth, while the S&P 500 returns were 28% dividends

18
and 72% value growth (Sando, 2012). This is no coincidence, as most investors that
buy rent-generating properties are partly doing so because they desire a consistent
stream of cash flows over a long period of time. As REITs function as a way for
stockholders to invest in rent-generating real estate, it should also provide similar cash
flows. This is also embedded in the very definition of a REIT. According to the
Securities and Exchange Commission (SEC), “To qualify as a REIT, a company must
distribute at least 90 percent of its taxable income to shareholders annually in the form
of dividends” (SEC, 2011). In addition to this, REITs are required to invest at least 75%
of its total assets in real estate or cash and derive at least 75% of its gross income from
real estate related business. If these criteria are met, REITs are exempt from paying
corporate tax in the U.S.

3.7 Valuing Real Estate

Real estate assets can in many ways be valued with the same metrics as stocks. For
instance, multiples, yields, and growth rates are all used for valuing real estate.
However, there are also metrics unique to real estate.

Fundamentally, the revenue from all real estate is a combination of value growth and
direct revenue. Combined, these two income streams comprise the total revenue from
a property. The direct revenue comes from rent payments that typically are paid every
month, quarter, or year. The revenue from value growth is the change in the property’s
market value over the holding period. This value growth is a result of the change in the
market rent, change in the yield, or change in the residual value (the value appreciation
that cannot be explained by the two other factors). This includes goodwill and
expectations about future earnings (Syz, 2008).

3.7.1 Damodaran 2012

Some real estate specialists have historically argued that real estate is too different
from stocks, and needs to be valued with unique methods. Aswath Damodaran (2012)
presents the view that the discounted cash flow method commonly used to value
publicly traded stocks also is the best method for real estate assets. Fundamentally, he
says, the value of any cash flow generating asset should be the present value of the
expected cash flows generated by the said asset. I.e. the value of a real estate asset is
determined by the cash flows it generates, the certainty associated with these cash

19
flows, and the expected growth in these cash flows. The higher the level, growth, and
certainty of an asset’s future cash flows, the greater the asset value, other things
remaining equal (Damodaran, 2012).

To use the DCF-method for real estate, a few extra aspects need to be considered both
when quantifying the risks and estimating a discount rate, and when estimating the
right cash flows for the right lifetime.

When discounting cash flows, it is normal to use the weighted average cost of capital
(WACC). This principle also works for real estate. The WACC is given as:

𝐸 𝐷
𝑊𝐴𝐶𝐶 = 𝑟𝐸 ∗ + 𝑟𝐷 ∗ ∗ (1 − 𝑇)
𝑉 𝑉
The cost of debt (𝑟𝐷 ) is the rate at which the investor can borrow money. E/V is the
equity ratio, D/V is the debt ratio and T is the corporate tax rate. These factors work
the same way for real as financial assets, as they merely concern the financing of the
asset. The cost of equity (𝑟𝐸 ) however, leaves more room for discussion. The cost of
equity for an investment is given by the CAPM formula:

𝐸[𝑟𝑖 ] = 𝑟𝑓 + 𝛽𝑖 (𝐸[𝑟𝑀 ] − 𝑟𝑓 )

Hence, it depends on the risk-free rate (𝑟𝑓 ), the asset’s correlation with the market (β),
and the asset’s risk premium (𝑟𝑚 − 𝑟𝑓 ). The cost of equity can also be found by the
arbitrage pricing model. Nevertheless, both CAPM and the arbitrage pricing model
define the risk of any asset as the portion of that asset’s variance that cannot be
mitigated through diversification. For this risk definition to be viable, it is assumed
that the marginal investor in the asset keeps a well-diversified portfolio.

In regards to real estate investments, this view is contested by many. The notion is that
real estate investments tend to be so large that many real estate investors will have
difficulties diversifying sufficiently. If real estate investors do not have diversified
portfolios, then the assumption that only non-diversifiable risk is rewarded does not
stand. In that case, CAPM and the arbitrage pricing model are unsuitable for
estimating the cost of equity.

Damodaran (2012) on the other hand, argues that the marginal real estate investor
should be assumed to hold a well-diversified portfolio. There are primarily three

20
arguments for this. The first is that the investors that are heavily invested in real estate
and not diversified, are doing so by choice because they believe in the sector and/or
because they have specialized knowledge in the field they want to leverage. This is no
different from investors whose portfolios are heavily weighted in a specific industry,
for instance healthcare or tech stocks. The second argument is that the marginal
investor in real estate, just as for stocks, increasingly are becoming large institutional
investors such as mutual funds and life companies, who are making investment
decisions on behalf of aggregated funds. These investors often have hundreds and even
thousands of billions of dollars in assets under management, which is more than
enough to keep a diversified portfolio, and the transaction costs low. The last argument
is that although indeed, real estate assets often are large investments, they are often
split up into smaller pieces. This allows investors to buy shares of property or real
estate portfolios, such as real estate companies and REITs. Such vehicles allow smaller
funds and even retail investors to include real estate assets in diversified portfolios.

Based on this, Damodaran concludes that the CAPM can be used to find the cost of
equity to be used as a discount rate for real estate valuations. Still, the pricing
mechanisms of real estate is different from that of stocks. This has to do with real
estate’s low liquidity, which creates an issue with the credibility of the market beta of
private market real estate assets. This issue will be discussed more thoroughly under
the REITs VS direct real estate section in chapter five.

4 Data
The data chapter represents the data used in the quantitative analysis. It does so by
explaining the indexes and assets that are used for and why it is chosen. It also presents
where the data has been collected, making it possible for other researchers to access
the same data, and to gather updated data of the same indexes.

4.1 Stock Market

This thesis assumes that the market value of publicly traded companies accurately
represents the fundamental value of the companies. For this to be true, all publicly
available information needs to be reflected in the stock prices. Thus, the thesis
assumes that the public exchanges for both stocks, REITs, and bonds are efficient
markets where every relevant change in the companies, the market, and the
macroeconomic environment promptly will be reflected in the market prices.

21
Stocks are traded continuously – Their trading frequency varies over time and between
stocks, but most S&P 500 constituents are traded thousands of times per day. This
means their correct market price is updated just as often. Stocks are also usually traded
on the exchange for extended periods of time, providing plenty of data for comparison
and calculations.

For the quantitative analysis, this historical data is obtained from databases. As the
thesis focuses on the USA, the U.S. stock market is used as a benchmark. Historical
values for the equity market are from the Center for Research in Security Prices, LLC
(CRSP). This data includes the monthly returns of every stock listed on the New York
Stock Exchange (NYSE) and NASDAQ. It also includes all the stocks listed on the
American Stock Exchange (AMEX) for years prior to 2008, when the exchange was
bought and incorporated into the NYSE. These US stock market returns are
downloaded from Kenneth French’s database; “Kenneth R. French – Data Library”
(French, 2020). The returns of the indices are value-weighted based on the value of
the individual stocks on the market and are adjusted for dividends and stock splits.

Portfolios used for Fama-French analysis are also downloaded from the Kenneth
French’s database. These portfolios are constructed as described in the Fama- French
three- and five-factor model sections under theory.

For the COVID-19 related analysis, all stocks on NASDAQ and NYSE are still used as
the benchmark for the stock market. For this part of the quantitative analysis, daily
values and volatility are used instead of monthly. This is to get more data points, as
the analysis is done over a shorter period, from February 1st to April 30th.

4.2 Bond market

Most large-cap investors have a significant part of their funds invested in bonds,
making bond returns important to analyze and compare with real estate. To represent
bond investments, this thesis uses an index to capture the entire U.S. bond market in
an unbiased way. The most representative index for this is the Barclays Capital U.S.
Aggregate Bond Index (AGG). This index tracks the performance of the investment
grade, U.S. dollar-denominated, fixed-rate bond market. The index is the most
commonly used benchmark for analysts and bond traders and is used by more than
90% of U.S bond investors (State Street Global Advisors, 2020). The historical return

22
and yield data for the AGG can be obtained from numerous databases such as Yahoo
Finance (Yahoo Finance, 2020).

The AGG consists of more than 8,000 bonds, treasuries, mortgage-backed securities,
and corporate bonds of different maturities. The weights of the different constituents
as of April 2020 can be found in figure three.

Figure 3: AGG Index constituents

Source: (State Street Global Advisors, 2020)

For the COVID-19 related analysis, the AGG is still used as the benchmark for the stock
market. For this part of the quantitative analysis, daily values and volatility are used
instead of monthly, just as for the stock market data.

In the factor analysis the 10-year constant maturity treasury bond rate is used as a
factor for interest rate exposure. The 10-year treasury rate of return is downloaded
from FRED (Board of Governors of the Federal Reserve System, 2020).

23
4.3 Real Estate Market
The U.S. has many publicly-traded REITs and indexes for different REIT sectors,
making the historical return from these REITS almost as easy to find as for “regular”
stocks. To capture as much of the REIT market as possible, this article uses indexes
comprised of REITs for its analysis. These indexes are delivered from Nareit. The FTSE
Nareit All REITs Index includes all Nareit-membered REITs listed on NYSE, AMEX,
or the NASDAQ National Market List. The index is not free float-adjusted and has no
requirement for minimum size or liquidity. The returns used are dividend-adjusted,
which is extra important as the majority of REIT returns come from dividend
payments.

The index is built up of 20 indexes specialized in 12 unique real estate sectors. In total,
these 20 indexes are comprised of 219 REITs. The All REITs index is the market
capitalization based weighted average of these 219 REITs (Nareit, 2020a). In total,
there are 293 REITs traded on the major stock exchanges in the U.S., meaning the
majority of all publicly-traded REITs in the U.S are part of the All REIT index, making
it representative of the market (REITnotes, 2020), (Nareit, 2020b). The distribution
between sectors in this index, sorted after market capitalization, can be found in figure
four. Most of Nareit’s segment indexes have data from the year 1994. Because of this,
the analysis starts on 01/01/1994.

As for the COVID-19 related analysis, the same index, FTSE Nareit ALL REITs index
is used to represent the impact of the virus on the REITs market. For this part of the
quantitative analysis, daily returns and volatility are used instead of monthly, just as
for the stock and bond data.

24
Figure 4: Nareit ALL REIT Index, segment weights

Source: (Nareit, 2020a)

Over the years, several additions have been added to the index. In 1994, the index
consisted of nine segments: Office REITs, which are owning and operating office
space. Industrial REITs, which are mainly owning terminals for logistics but also some
properties in connection to factories and production of goods. Retail REITs comprise
shopping malls, high street retailers, and some REITs with connection to specific
retailer brands. Residential is the biggest category, and holds REITs specializing in
apartment buildings, single-family homes, and manufactured homes. Diversified
holds REITs who are diversified within themselves. Lodging REITs own hotels and
resorts. Health care REITs hold hospitals, office space in connection to hospitals,
assisted living facilities, medical labs, and research facilities. Self-storage REITs own
storage facilities including vehicle storage, climate-controlled storage, and document
storage for businesses.

In January 2000, mortgage REITs were added to the index. These are split into
commercial and residential mortgages and differ from the other REITs in that they
own collateralized mortgage debt, and not equity in properties. Timber REITs were
added to the index in 2010. This relatively small sector sticks out from the others in
that it doesn’t own buildings, but large pieces of woodlands. Infrastructure was added

25
in 2012 and is now the second biggest sector in the index. The largest segments within
infra-REITs are cell towers, fiber-optic cable networks, and power grids. The latest
extension was in 2015 when data centers and specialty REITs were added to the index.
Data centers have been an important contribution because of their marvelous growth
in pace with the recent technological advancements. Specialty REITs is an interesting
segment in that it is fairly diversified within itself. The segment includes REITs that
specialize in high-security art storage, prisons, farmland, and golf resorts among other
areas. The largest part of the segment however, is casino properties. The list of the ten
largest REITs in the Nareit All REIT Index by market cap and sector can be found in
figure five.

Figure 5: Nareit ALL REIT Index, Biggest constituents (biggest REITs)

Ma r ket ca p
Com pa n y Sect or
(USDb)
In fr a st r u ct u r e
A m er ica n T ower Cor p 1 0 2 ,4
(cell t ow er s)
In fr a st r u ct u r e
Cr own Ca st l e In t er n a t ion a l 6 2 ,3
(cell t ow er s)
In du st r ia l
Pr ol ogis In c. 5 8 ,6
(log ist ics)
Equ in ix In c. Da t a cen t er s 5 0 ,3

Ret a il
Sim on Pr oper t y Gr ou p 4 1 ,1
(sh oppin g
Pu bl ic St or a ge St or a g e 3 9 ,1

Hea lt h ca r e
Wel l t ower In c. (Hospit a ls a n d 3 4 ,4
a ssist ed liv in g )
Residen t ia l
Equ it y Residen t ia l 3 0 ,8
(a pa r t m en t s)
Residen t ia l
A v a l on Ba y Com m u n it ies In c. 3 0 ,2
(a pa r t m en t s)
In fr a st r u ct u r e
SBA Com m u n ica t ion s 2 8 ,2
(cell t ow er s)

Source: (Nareit, 2020a)

26
5 Analysis
In the analysis chapter, research is done and analysis is conducted, built on the
theories presented in the previous chapter. This is done to investigate the thesis
statement and provide perspectives and answers to the sub-questions. The analysis is
divided into two parts: a qualitative analysis and a quantitative analysis.

5.1 Qualitative analysis

The qualitative analysis is the more theoretical of the two. This segment does not
analyze and interpret raw data but is based on the works of other academics. In that
sense, it is mostly meta-analysis. The exception to this is the independent analysis of
the housing data done as part of the COVID-19 related analysis. The analysis explores
the different ways to invest in real estate, the portfolios of some prominent asset
managers, real estate’s typical reactions to macroeconomic factors, and a thorough
analysis of the COVID-19 crisis’ effect on different segments, including a discussion of
potential short and long term effects.

5.1.1 REITs VS direct real estate investments

Considering real estate solely as a financial asset, the main difference between real
estate and stocks is perhaps the price-making mechanisms. The lion’s share of real
estate is not traded on an exchange; it is private properties switching hands. For
commercial real estate, this is done much like how people are selling and buying their
homes, only on a larger scale: Both the seller and the buyer hire brokers and lawyers
and perhaps analysts, who make their value-estimates for the property. For a sought-
after property, there might be a bidding war, where the buyer willing to pay the most
gets to buy the property. If fewer parties are interested, the transaction price will be
set as a compromise between the seller’s and the buyer’s estimated value, essentially
their bid/ask price. When properties are bought like this, on behalf of an investor or
fund, it is called a direct investment. The mechanisms of such a transaction, at its core,
are not that different from when a stock is bought. The main difference is that stock,
in most cases, represents a small share of a company. Thus, listed companies often
have millions of shares on the market, allowing the company to be traded several times
per day, hour, or even second. This way, the actual market price is updated
continuously. This makes historical returns of listed stocks highly accessible for
analysis.

27
As for direct real estate, the historical returns are a little more complex to find. As there
is not a small percentage of the property that is bought and sold, but 100% of the
property, the same property will naturally not be bought and sold often, making
properties illiquid assets. Because of this, the price-making mechanisms are somewhat
different. As the historical trading prices for a single property are scarce, property
prices have to rely on the recent pricing of similar buildings. For instance, the price of
a recently sold Downtown Chicago office building will be used as a proxy to value a
similar Downtown Chicago office building. This way of estimating value is called
appraisal. This is not only used for single assets; several real estate indexes are also
based on appraisals. These could be indexes for location and category specified
properties, for instance an index for multi-family homes in Los Angeles or high street
retail properties in Manhattan. Pooled together, these indexes are used by many as a
proxy for the entire real estate market, local as well as aggregated (Damodaran, 2012).
Appraisals work well as a way of estimating the value of a specific property, but there
are two reasons why they tend to give inaccurate estimates for market returns.

The first reason has to do with the very method of appraising market returns. It will
indirectly assume that all properties in the same area and the same category are worth
the same. As parameters such as construction quality, location, floor planning, and age
impact property values significantly, appraisals prove inaccurate. Rebel A. Cole and
Susanne E. Cannon from Chicago University studied the accuracy of real estate
appraisals based on 25 years of appraisals and actual property transactions from
NCREIFs sales data. The data was collected from 1984 to 2010, a period containing
two up and down cycles in the market. They found that appraisals on average are more
than 10% above or below the following sales price. They also found that even in a
portfolio context where positive and negative errors are offsetting each other, the
appraisals were more than five percent off (Cannon & Cole, 2011).

The second reason has to do with how often these appraisals are done. NCREIF’s
National Property Index (NPI) makes its appraisals quarterly (Cannon & Cole, 2011).
This creates a smoothed return as short-term volatility will not be captured. Hence,
appraisal-based indexes such as the NPI and non-public real estate trusts called
Comingled Real Estate Funds (CREFs) tend to underestimate the true volatility of real
estate assets and markets. As a result, if appraisal-based data is used as a proxy to
calculate the return and volatility of the real estate market, analysts might find real

28
estate to seem overly attractive. Because of this, appraisal-based indexes will not be
used in the quantitative analysis of this thesis.

Publicly traded REITs will not have this problem, as they are constantly traded like
any other stock. Thus, REITs are far more usable for comparing with other asset
classes. Still, there is a problem with using REITs as a proxy for the real estate market
as a whole, as properties owned by REITs may not be representative of all real estate.
This is because REITs tend to own certain types of buildings and certain property
classes more frequently, and often hold buildings with higher than average quality and
standard. Additionally, the securitization of real estate and mortgages in REIT
portfolios could cause deviations between REIT returns and real estate returns
(Damodaran, 2012). Looking at historical returns from REITS, CREFs, and the stock
market, it is a clear tendency that assets with appraisal-based valuations such as
CREFs and non-listed real estate, in general, have lower volatility associated with them
than publicly-traded REITs. The data also shows that REITs have a higher correlation
with the stock market than the non-publicly traded assets, indicating that REITs
historically have not been a good proxy for the real estate market as a whole
(Damodaran, 2012).

In separate and more recent research, the European Public Real Estate Association
(ERPA) studied the FTSE EPRA Nareit indexes representing global publicly-traded
REITs, and MSCI indexes of institutional real estate holdings representing the direct
real estate. To adjust for differences in leverage, all returns are unlevered. To
compensate for the smoothed returns created by quarterly appraisals, direct
investment returns are de-smoothed. The analysis then considers returns in the period
from 1998 to 2017. Econometric analysis is conducted on the data to find returns,
volatilities, and correlations. A perhaps surprising find of this analysis was that the un-
leveraging of REITs had an insignificantly small effect on the correlation with direct
real estate (European Public Real Estate Association , 2019). Hence, for simplicity, this
thesis will not unlever any returns in its analysis.

Considering the indexed returns from listed American REITs and American direct real
estate in figure six below, the co-movement is evident. The two asset classes seem to
have been especially dependent on each other until the 2008 financial crisis. This crisis
had a larger effect on the direct market than on REITs. After REITs and direct real
estate turned better in 2009, the two asset classes have moved somewhat equally. Still,

29
the direct real estate has been consistently below REITs with about as much as the
extra loss direct suffered compared to REITs in 2008. The two asset classes show some
fluctuations, but the long term trends are corresponding.

Figure 6: Indexed returns of listed real estate and direct real estate, U.S.

Source: (European Public Real Estate Association , 2019)

This long term similarity between REITs and direct real estate is confirmed by
correlations. As illustrated by the blue line in figure seven below, the correlation
coefficient was a little under 0.5 for investment horizons of one year. The correlation
is consistently increasing over time and is 0.7 for five-year horizons. Doing the same
analysis on REITs compared to the stock market, the opposite trend is observed.
ERPA’s analysis shows that REITs have a high correlation of 0.60 over one year and a
much lower correlation of 0.34 for five-year investments. Historically, REITs have had
a higher correlation with stocks for horizons up to about two years. For horizons over
two years, REITs have been correlating more with direct real estate.

30
Figure 7: Correlations between REITs and direct real estate, and REITs and stocks
over time

Source: (Hoesli & Oikarinen, 2019)

Based on this, ERPA concludes that long term returns from REITs correlate more with
the underlying property market than with the stock market. But in the short term,
REITs correlate more with the stock market and less with the direct real estate market
(Hoesli & Oikarinen, 2019).

CBRE adds to this theory by arguing that especially during large market shifts, REITs
tend not to be priced in line with their underlying asset value. This has been more
evident in the recent bull market after 2008, as a larger share of all stocks is held by
ETFs and index funds. This also applies to REITs, whose majority of shares are held
by ETFs and large mutual funds such as the ones described in the next segment. These
funds tend to invest based on broader market conditions rather than the net asset
value or implied capitalization rate of the REITs’ portfolio of properties. In fact, they
find that only 15% of REIT shares are held by real estate dedicated investors who
analyze the properties at a detailed level (CBRE, 2020).

In short, REITs work well as a way of making real estate investments more diversified.
It also allows investors to get exposure to the real estate market without investing in
the illiquid direct market. In return for being more liquid, REITs are also more volatile,
as they are traded more often, and are part of other types of investors’ portfolios.

31
Because of these findings, this thesis will focus on REITs for the portfolio building and
main analysis, and not real estate as a whole. As REITs are proven often to deviate
from the underlying property values, the discoveries and conclusions made from
analyzing REITs in this thesis will not be used to draw conclusions on behalf of direct
real estate.

5.1.2 Analysis of existing asset managers

Analysis of how typical institutional asset managers allocate money between asset
classes, and how they invest in real estate, can give important insight into what
strategies have worked well in the past. It also serves as a benchmark for this article’s
quantitative analysis, as it provides a solid foundation for comparison.

5.1.2.1 Investment managers

Previous analysis of investment managers has found that typical pension funds
allocate 6% of their portfolio to real estate, while typical hedge funds allocate 3.5%
(Andonov, Eichholtz, & Kok, 2015). This segment will look deeper into specific
investment managers.

The largest portfolio managers in the world are the major American plus some
European investment management companies. The largest of these companies:
BlackRock Inc., Vanguard Group Inc., and State Street Global Advisors all have more
than 2.5 trillion USD in assets under management (P&I, 2020). For many of these
managers, it is difficult to obtain an overview of the portfolio weight of REITs and
other public real estate, as they are sorted under “equities” together with other stocks
in their databases. Looking at the private markets, most of the largest asset managers
own little to no direct real estate. Out of the three largest asset managers, only
BlackRock reported any real estate, a mere 0.50% of its assets under management
(AuM). Most of these asset managers undoubtedly hold some real estate; the exact
numbers are just difficult to access due to their reporting practices.

Because of this issue, this paragraph chooses to focus on the major asset managers
who directly report their investments in listed real estate: BlackRock, JPMorgan Asset
Management, Amundi Asset Management, and Prudential Financial. JPMorgan Asset
Management is the seventh-largest asset manager in the world with 1.7 trillion USD in
AuM. Among the top 10 investors, it is the manager with by far the biggest allocation
to REITs, having 378 billion dollars invested in real estate, corresponding to 22.8% of

32
all its assets. 4% of its portfolio is invested in unlisted real estate and 18.8% is listed
real estate. While JPM invested mainly in listed real estate, U.S Life insurance
company, and the world’s 10th largest asset manager, Prudential, focused more on
unlisted real estate. Prudential was the world’s second-largest owner of unlisted real
estate in 2018, with 129bn USD invested in the sector (Institutional Real Estate, Inc.,
2018). It also had a substantial allocation of 47bn in listed real estate. This amounted
to 8.1% and 3.4% of total AuM respectively. French asset manager Amundi controls
1.6bn USD, out of which 1.8% is reported to be unlisted real estate, and 2.2% is listed
real estate (P&I, 2020). This provides three different real estate strategies and proves
that the major asset managers are not uniform in their investments. Total equity,
bond, and real estate portfolio weights can be found in figure eight.

Figure 8: Selected investment managers' asset allocation

Source: (P&I, 2020), (O'Dea & Lowe, 2018)

Considering pension funds’ and life insurers’ missions, it is no surprise they have
preferred higher allocations of real estate. These companies are obligated to pay out
pensions and/or life-insurance payouts with a specified rate of return, at a predictable
time in the future. The safest way to ensure having the liquidity to meet these expenses
is to generate a long term cash flow on the received premiums matching the rate of
return they have promised their customers. Thus, the fixed cash flows with long
contracts received from direct commercial real estate, as well as bonds, seems to be
perfect for this kind of investor. The long investment horizons of these investors also
make the illiquidity risk of direct real estate, discussed in the previous segment, less of

33
an issue. additionally, they would want to minimize market risk. The combination of
these two preferences could explain why this class of investors tend to prefer direct
real estate above REITs. Direct real estate’s aptness for this purpose is underbuilt by
the fact that there are three insurance companies: Prudential (PGIM), AXA investment
managers, and SwissLife among the biggest ten holders of direct real estate in the
world (Institutional Real Estate, Inc., 2018).

5.1.2.2 Sovereign Wealth Funds

Sovereign wealth funds (SWFs) are some of the largest single funds in the world. By
the end of 2018, the 35 biggest SWFs controlled around seven trillion USD in assets
(Kozlowski, 2020). Large resources are used to find optimal strategies for these funds.
Thus, it is relevant to analyze their asset allocation. In recent years, most SWFs have
gradually changed their portfolios to be more in line with the strategies of the global
asset management industry (Hentov & Petrov, 2020). The specific portfolios are
difficult to find, as many SWFs lack transparency. Nevertheless, the biggest SWF,
Norway’s Government Pension Fund Global, managed by Norges Bank Investment
Management (NBIM) offers full transparency of its portfolio. At the end of 2019,
NBIM’s fund had a total AuM of 1.1 trillion USD. The fund’s real estate portfolio was
worth 47bn USD, about 4.4% of total AuM. Out of the fund’s total real estate
investments, 66% was in unlisted assets, and 34% was in listed assets, mostly REITs.
Hence, real estate is a relatively small part of the fund. NBIM has a mandate to invest
up to 7% of AuM in unlisted real estate, and have no restriction on listed real estate
investments. Of the real estate investments, the U.S. is by far the biggest market,
amounting to ~50% of the fund’s real estate values (NBIM, 2020).

34
Figure 9: NBIM asset allocation

Source: (NBIM, 2020)

For the remaining SWEs, we have to rely on statistics on a more general level by
analyzing how much the funds have allocated to the different markets: Equities, fixed
income, and private markets. For these three categories, Elliot Hentov and Alexander
Petrov in State Street Global Advisors have published statistics covering the largest 35
SWFs (Hentov & Petrov, 2020). As seen in figure 10, The SWFs have a virtually
constant part of their assets invested in listed equities, with the AuM share only
fluctuating between 45% and 46% between 2002 and 2018. This part also includes
listed REITs, as these are not uniquely identifiable in the reporting for most SWFs.
The bigger changes in the portfolios over time have been that considerable funds have
been moved from fixed income (corporate bonds and treasuries) to the private
markets. These private market investments are mainly private equity, direct
infrastructure, and unlisted real estate in the form of direct investments, unlisted
REITs, and CREFs.

35
Figure 10: Average asset allocation for SWFs

Source: (Hentov & Petrov 2020)

When dividing SWFs into small-cap and large-cap (smaller and bigger than 50bn USD
AuM) as seen in figure 11, it is clear that the large funds have caused most of the
transition. In 2002, the large funds held ~55% bonds, and the “small” held ~45%. By
the end of 2018, the large funds had on average more than halved their exposure to
26%. The smaller funds only reduced their bond share with around 10%, down to 37%.
The report forecasts that the development from fixed income into private markets will
stabilize, but slightly increase in the years to come. This is partly due to the belief that
interest rates in the U.S. will stay low. Another reason is the tendency of the bigger
funds acting as market leaders, and the smaller funds tend to follow their strategy.
Thus, State Street predicts that the smaller SWFs will drift towards the allocation of
the bigger, and driving the average bond allocation further down, and the direct real
estate further up (Hentov & Petrov, 2020).

36
Figure 11: Allocation to fixed income over time, by fund size

Source: (Hentov & Petrov 2020)

All the managers considered in this segment should have the same goal: To optimize
the risk-adjusted returns of their investors’ and shareholders’ wealth. With that in
mind, asset managers are found to have surprisingly different investment strategies.
Especially when it comes to real estate, there does not seem to be a defined industry
standard except for the fact that everyone has included some real estate in their
portfolios. Leading experts in the same field such as JPMorgan and BlackRock have
made fundamentally different decisions in regards to real estate allocations, proving
that there is no established optimal strategy. This is a sign that investors should be
open to reconsidering their real estate weighting and consider the strategies found in
this thesis.

5.1.3 Macroeconomic factors

An important aspect when considering the risk of an asset is how the value of the asset
tends to move in relation to changes in the macroeconomic landscape. This segment
will discuss how real estate performance is related to the macroeconomic factors
interest rates, inflation, and recessions.

5.1.3.1 Interest rates

Interest rates are highly important for real estate. As for any other cash flow producing
asset, lower interest rates will result in a higher valuation of the asset. Lower interest
rates give a lower cost of capital, which gives a lower discount rate which gives a higher

37
present value of the free cash flows. Thus, using valuation methods relying on
discounted cash flows (DCF-analyses), this connection is evident.

For real estate there is also another dynamic increasing this connection. This too is
evident through DCF-analysis. Studies show that the majority of a real estate asset’s
enterprise value usually lays in the budget period, whereas it lays in the terminal value
for most other stocks (Damodaran, 2012). This shows that the bigger part of stock
values is the company’s growth expectations, while the bigger part of real estate values
is the actually contracted cash flows, just like for bonds with coupons. For real estate
assets, these cash flows are mostly rent payments from properties. With most property
classes, these rents are close to guaranteed through long term contracts. Thus, the rent
payments for years to come are determined at the time of the transaction. The yield or
capitalization rate of the property is then found as the net rental income divided by the
market price of the property. This is exactly the same method as to how the yield of a
bond is calculated as the coupons divided by the current bond price. Thus, rent
payments from properties strongly resemble coupon payments from bonds. The
inverse relationship between yields and price are apparent in both asset classes. The
yield spread of a bond or real estate asset is the difference between the yield from real
estate and the treasury yield. A higher yield spread makes real estate a more attractive
investment.

Just like other equities, most REITs have some leverage on their investments. This
amplifies the effect of lower interest rates creating higher REIT returns. Mortgage
REITs on the other hand, have opposite exposure. Shifts in interest rates will, after a
short delay, result in shifts in mortgage rates. Thus, falling interest rates will have a
positive effect on equity REITs but a negative effect on mortgage REITs and vice versa
(Ibbotson & Siegel, 1984). The vast majority of REITs mostly hold equity, so REITs
returns, just like other equities, mostly have an inverse relationship with interest rates.
The mortgage REITs somewhat reduces this effect, to some extent making REIT
returns less dependent on interest rates.

5.1.3.2 Inflation
Regardless of the asset class, higher inflation results in lower real returns. This means
a hike in inflations will cause lower bond prices and lower valuation of equities.
Rationally, all cash flow producing assets should be negatively impacted by a hike in
inflation and vice versa. Real estate however, has for a long time been considered an

38
inflation hedge. The notion is that property values and rents typically increase during
times of inflation.

Professor William Wheaton at the Massachusetts Institute of Technology (MIT) has

researched this field, and analyzed different real estate sectors’ effectiveness as
inflation hedges over time. The study analyzes national U.S. data from NCREIF for
each of the property types: Office, apartments, retail, and industrial. From this data, a
“property value” series and a “net income” series have been calculated quarterly from
1978 through 2016. The income and value of the four segments have then been
compared to the U.S. cumulative CPI index. Elasticities between the CPI index and
value and income respectively, have been estimated with an error correction model,
and the results are clear: Some subsectors of commercial real estate have consistently
worked as an inflation hedge over the last 40 years. All elasticities were positive,
meaning that for all of the four sectors the study analyzed, a positive change in CPI
resulted in a positive change in both rental income and value. Focusing on rental
income, retail proved to be the best hedge, with income rising a little more than one
percent when CPI rise one percent. Income from industrial properties and apartments
have provided a partial hedge for inflation. Office income, which was the worst of the
four, provided practically no hedge. Even so, the segment provided a slightly positive
response 0f 0.18% rise in income from a 1% increase in CPI. Keeping in mind that the
general stock and bond market has a negative elasticity, this is still a decent result
(Wheaton, 2017).

Figure 12: Real Estate CPI elasticities

Pr oper t y t y pe In com e V a l u e
Ret a il 1 ,0 2 1 ,0 7
In du st r ia l 0 ,7 0 ,9 1
A pa r t m en t 0 ,5 6 0 ,9 8
Office 0 ,1 8 0 ,7 4

Source: (Wheaton, 2017)

The research shows that the elasticity for property values has a stronger connection to
inflation than the elasticity for rents. Considering the historical development in
property values, these have worked better than income from a CPI-hedge perspective.
both retail and apartments are complete inflation hedges. Industrial properties were
only marginally behind and have virtually worked as a complete hedge. Even office
39
property, whose income did not provide much of a hedge, had a solid elasticity of 0.74,
meaning office property too, has been a decent inflation hedge in the past.

The reason for the difference in income and value responses, according to Wheaton, is
the secular trend downwards in cap rates (yields) for commercial real estate over the
last 30 years. The decline in cap rates has almost matched the decline in government
interest rates, both in real and nominal terms. In this decline, property values have
kept up with inflation even though a significant part of the income has not (Wheaton,
2017).

There is of course no guarantee this dynamic will stay as strong in the future. Retail is
extra uncertain, as the sector is experiencing some issues with e-commerce stealing
market shares. For most commercial real estate however, rents are adjusted annually
for changes in the consumer price index (CPI-adjusted), and residential property rents
are usually adjusted annually to the market price, including change in CPI. In total,
the notion that real estate works as an inflation hedge is confirmed with historic data,
and it seems plausible this will continue in the future, at least to a certain degree.

5.1.3.3 Recession
In the U.S., a recession is defined as “a significant decline in economic activity spread
across the economy, lasting more than a few months, normally visible in real GDP, real
income, employment, industrial production, and wholesale-retail sales “. It follows
from the definition that most industries will suffer from this, and real estate has
historically not been an exception. It has however tended to affect different subsectors
with different severity.

Lodging and retail tend to move in line with, or even more than the rest of the
economy. As people are more concerned about their financial situation and job
security, they will spend less on consumer goods and travel. Especially the luxury
goods and high-end hotels can be one of the earlier to suffer in a recession. Office,
industry, healthcare, and other segments that have professional renters tend to be
somewhat more robust. For a business to move location is an expensive and time-
consuming process and something that is avoided at most costs. Still, pressure on the
market can result in lower rents which leads to lower earnings from the segment. If
the situation gets severe enough, companies will default or be forced to move, at which
point real estate owners will be in bigger trouble.

40
Residential real estate tends to be even more robust. High unemployment will put a
strain on personal economies. Still, welfare systems tend to keep most people from
having to move out of their homes. In the long term, rents will go down and so will
property values, but a collapse in the residential markets will take a long-lasting crisis,
or uncontrolled use of leverage like seen in the 2007 crisis. The market is now more
regulated and controlled, making a large-scale bust in the residential real estate
market less likely.

In total, fluctuations in real estate value seem to be less directly linked to

macroeconomic shifts than stocks. All real estate is in some way dependent on leases
and customers, and even though real estate is a core asset and the assets themselves
will not change by a recession, the change in companies’ and people’s economy and
liquidity tend to slow down the demand for properties, eventually driving property
values and rents levels down. This analysis will continue in the next chapter, where the
COVID-19 crisis provides a specific example of a recession.

5.1.4 The COVID-19 crisis

After 2008, the stock market and real estate market both experienced a more or less
uninterrupted bull market of 11 years. During the creation of this thesis, the bull
market came to a sudden end, when the U.S. stock market shifted downwards in the
last week of February 2020. The COVID-19 crisis has now critically impacted our
society and the international economy. In absolute terms, the USA has been hit the
hardest by the virus as of mid-April 2020, with the most confirmed cases and the most
deaths from the virus (Johns Hopkins University & Medicine, 2020). This has led the
U.S. to the brink of a recession. In late March 2020, Congress passed the CARES Act,
providing stimulus packages to support directly and indirectly all sectors of the
economy. The CARES Act is unprecedented in size, injecting 2.1 trillion dollars into
the economy. Still, big consequences for the economy and financial markets seem
unavoidable. The ultimate severity of the impact is still unknown, with different
experts providing varying outlooks. Most experts seem to agree that the impact will
differ greatly between industries, and also amid the real estate sector. This chapter’s
objective is to provide clarity amongst the many deviating views and the general
chaotic situation by presenting an overview of and discussing different scenarios.

41
As of May 2020, it is evident that some parts of the real estate sector are highly affected
by the crisis. The geographic location of assets is of course a highly significant factor,
as some cities are far more impacted by the crisis than others. This will not be analyzed
further, as the more or less impacted cities can change quickly, and a geographical
analysis has less relevance for the research question. It is more relevant to analyze the
different real estate segments’ reactions to the crisis. As seen in figure 13, property
segments with greater human density have been hit the hardest: regional malls,
lodging, healthcare facilities (except hospitals), and student housing have seen a
drastic decrease in operating income (Gujarl, Palter, Sanghvi, & Vickery, 2020). The
following paragraphs will take a deeper look at some core real estate segments that
have experienced more or less of an impact.

Figure 13: Change in U.S. unlevered and levered value from Feb 21st to Apr 12th

Source: (Lachance & Tibone, 2020)

42
5.1.4.1 Lodging
One of the first sectors to suffer from COVID-19 was the hotel industry. On March 14th,
the U.S. closed its borders towards the outside world. This set an immediate stop for
international tourism and business travels alike. Already at this point, the hotels were
suffering greatly. Then, on March 19, the state department issued a “do not travel”
advisory, robbing hotels of their domestic customers too. These new rules in
combination with the general fear of going out among people are all factors that have
made many hotels close their doors temporarily and even go bankrupt. As of April 3 rd
2020, Green Street Advisors estimates the unlevered enterprise property value of
American lodging assets has already declined by 37%. This makes lodging the
undisputed loser among the real estate sectors in the COVID-19 crisis (V. Gujarl et al.,
2020).

Marriott Hotels, the largest hotel operator in the world in terms of hotel rooms and
total revenue, reported numerous shutdowns. Approximately 25% of Marriott’s 7,300
hotels worldwide are temporarily closed. In the U.S., 870 Marriott hotels have
temporarily shut its doors, amounting to 16% of all its U.S. locations (Oliver, 2020).
16% might not seem that extreme, but the hotels that are still open are facing
drastically fewer guests, lower daily rates, and lower revenues.

Hospitality analyst company STR analyzes the U.S. hotel market on a continuous
basis. Its report of week 15 (April 6-12th) shows that compared to the same week in
2019, average occupancy was down 69.8% to an unprecedented 21%. This means that
approximately 79% of all hotel rooms in America were empty. In STR’s data going back
to 1987, the previous worst occupancy level was 54.6% during the financial crisis in
2009 (STR & Tourism Economics, 2020). Naturally, this low occupancy also resulted
in lower room prices, with average daily rates (ADR) down 45.6% year over year,
averaging a nightly rate of 71 dollars. These two factors combined result in a terrible
outcome for the most common hotel valuation metric: revenue per available room
(RevPAR). RevPAR was, on average, down 84% compared to week 15, 2019. For week
18 (the week ending May 2nd), occupancy, prices, and RevPar were up around five
percent compared to week 15 (STR, 2020). This indicates that these staggeringly low
hotel earnings have become the new normal during the crisis, and will probably stay
about as bad for as long as there are travel restrictions (STR, 2020).

43
The big hotel brands such as Marriott and Hilton mostly do not own their hotels; they
operate them and rent the properties from an investor, typically a REIT. As opposed
to most residential and office properties, where rent is paid as a fixed amount, hotels
normally pay a variable rent. In short, if a hotel does well, they will pay the REIT more
rent, and if it does poorly, it will pay less rent. This makes hotel owners more directly
connected with the tenants’ struggles than the owners of many other real estate
sections, further worsening the situation for hotel owners.

5.1.4.2 Retail
Retail has been a struggling segment long before the coronavirus hit the market.
Department stores have closed down locations and independent stores have gone
bankrupt. This already stressed segment is now being severely tested, as government-
imposed social distancing is forcing stores to close down. Real estate owners are
experiencing permanent loss of cash flows from retail tenants and an elevated chance
of bankruptcies, despite government support.

From the crisis began in late February until April 15th, the share prices of publicly
traded retail REITs have declined by 45% in the U.S. and 43% in Europe. The effect is
so bad that retail owners most likely are seeing a worse situation now than during the
2008 crisis (Lachance & Tibone, 2020).

5.1.4.3 Office
In the Nordic region, more office deals were executed in Q1 2020 than in Q1 2019.
From the statistics, it looks like the market is as liquid as before. An issue is that the
deals that were signed and executed in March 2020 counts towards this statistic. Large
office deals usually take months to negotiate, so the deals that were publicly
announced at the beginning of the COVID-19 crisis were essentially agreed upon
before the crisis hit Europe and the U.S. (Bjølgerud, 2020). This example is based on
Nordic data because a general U.S. statistic was not found regarding office
transactions. The notion of a delayed effect on transactions however is also true in the
U.S., as the transactions here are even more complex.

Like most other stocks, office REITs have taken a beating. Considering the three
largest office REITs in the Nareit index: Boston properties Inc. and Kilroy Realty
corporation are down about 40% since the end of February, as of May 11th. The second-
largest REIT, Alexandria Real Estate Equities, was also down almost 30 percent, but

44
the stock turned march 23rd and is now almost back at the same level as at the end of
February (Yahoo Finance, 2020d). One explanation for this could be that the office
REITs fell because the entire market was plunging, not because the crisis was bad for
office properties. In that case, the office REITs fell more than the asset values of the
buildings in the REITs. As discussed in the REITs VS direct real estate investments
paragraph, REITs have a higher correlation with the general stock market than the real
property value for short investment horizons, especially during crises. Thus, in a crisis,
REITs are typically not priced in line with their underlying real estate value (CBRE,
2020). This indicates that the large loss in office REITs could be an overreaction, that
the office prices will stabilize, and that property prices will not go as low as the REIT
prices indicate. Nevertheless, it is too early in the crisis to make conclusions, and office
prices, both direct and REITs could still go even lower.

5.1.4.4 Residential
As discussed under the recession paragraph, residential real estate typically sees a
delayed effect from economic crises. In this crisis, people are advised to stay inside
their homes unless absolutely necessary. This is one of the reasons people are not going
house-hunting. They simply are following government advice and also do not want to
contaminate themselves with the virus. In addition to this, with businesses struggling
and more people than ever before in American history are getting laid off their jobs,
these are uncertain economic times for most Americans. One would believe that people
who fear for their financial future inherently do not want to do big investments like
buying a home or sign a new lease on an apartment. Because of this, the development
in the number of transactions in the housing market is analyzed.

The complete output of sales data for the U.S. and its two largest cities is found in
figure 14 below. The numbers are adjusted for variations in seasons, as transactions
and prices usually are higher in the spring, and lower in the fall and winter (Zillow,
2020). The adjustment is minimal but still makes it easier to separate between normal
seasonal differences and market trends.

45
Figure 14: Home sales per month

Source: (Zillow, 2020)

February

As figure 14 shows, sales rose from January to February across the U.S, including New
York, the city hardest hit by the coronavirus. This indicates no immediate effect on the
housing market. In Los Angeles, the U.S. second largest metropolitan area, and with
the second-most cases of the virus, the number of transactions fell with 3% from
January to February, indicating a slight effect.

March

In March, the situation gets more interesting, as the U.S total sales numbers were
down 6.8% since February. Considering how stable this metric has been historically,

46
this is a significant drop. Still, month over month sales has dropped more on several
occasions over the last decade, last time in 2017, so the reaction is not considered
overly extreme (Zillow, 2020).

Surprisingly, New York’s sales are still up from February to March, despite the
pandemic worsening rapidly in the area. The same is the case in Los Angeles, where
sales numbers were up from February to March, but are still lower in March than in
January.

April

As of May 10th, most cities, Los Angeles included, have not updated their statistics for
April. Because of this, the total U.S. metric is also unavailable for April. New York has
published April numbers, showing the number of transactions fell with 15% from
March to April. This is the City’s biggest month over month drop in sales in Zillow’s
10-year database history (Zillow, 2020). New York’s drop confirms that areas directly
affected by the virus will suffer a serious cool-down of the housing market. In the lack
of national data, the next cities on the list are examined. Among the 100 biggest U.S.
cities, 20 of them have updated April sales numbers. All of these 20 cities had
significantly fewer transactions in April than March. The city whose sales were down
the least was Denver, Colorado, with a -6% change in sales. The city where sales were
down the most was Springfield, Massachusetts, with a -19% change in sales. The
average of the 20 cities was a decrease of 11,5% from March to April. The average drop
in sales from January to April was 5.9%. This shows that even urban areas with the
virus more contained see transactions drop significantly. Even though this is only a
sample of the U.S. housing market, the numbers all point in the same direction, and is
a clear indication that the whole country’s housing market is affected.

Zillow also provides data on the transaction prices of all the properties in the “home
sales per month” statistic. The development in monthly median prices, seen in figure
15, shows that the coronavirus has not made a statistically visible impact on home
prices as of the end of April. Numbers from both U.S. total, New York, and greater Los
Angeles show that the median sale prices in March were higher than in February and
January. New York prices were only marginally higher in February than January, but
this cannot be regarded as an anomaly in a market where growth has been minimal
over the last three years.

47
Figure 15: Median transaction values, all homes

Source: (Zillow, 2020)

Just as for the sales count, April numbers for U.S. total and the majority of U.S. cities
were not available by mid-May. The development of sales-prices in April is therefore
analyzed through the 20 out of the top 100 U.S. urban areas that have released their
data.

The transaction prices tell a different story than the sales numbers. In April, New York
alone had more cases of COVID-19 than any other country in the world (Johns
Hopkins University & Medicine, 2020). Still, median sales prices were up 3.5% from
March to April. Among the twenty cities investigated, five saw a decrease in median

48
sales price from March to April. None of these five cities were among the biggest 50.
Only one city, Des Moines, Iowa, had lower median sales prices in April than in
January. In total over the 20 cities, median sales prices were 1.8% higher in April than
in March, and 3.2% higher in April than in January (Zillow, 2020). As for the sales
count analysis, this is based on a sample of urban areas, but the data is also here
pointing in the same direction: As of April 2020, the crisis cannot be said to have made
a significant impact on the U.S. housing prices.

The key takeaway from the housing market is that the impact was not immediate, but
that it has slowly affected the residential market. Transaction volumes are significantly
down in April, and will likely continue down in May. Sales are down regardless of
geographic location, with effects seen in areas with and without a large presence of the
virus.

Even though fewer properties are changing hands, the transaction values are not
affected. There can be several reasons for this, but one likely factor is that the
transactions we see in March and even April could have been more or less agreed upon
before, or at the beginning of the crisis. To buy and sell a house is as earlier mentioned
a large and time-consuming process, which will cause some delay between a shock
hitting the market, and that shock affecting the prices. In Norway, a country where the
virus spread earlier than the U.S., housing prices were affected negatively already in
March, falling 1.4% (DNB Markets, 2020). Despite most experts predicting Norway’s
prices to fall even more in April, the seasonally adjusted prices were down only 0.2%
over the month (Eiendom Norge, 2020). COVID-19 has impacted the U.S. significantly
worse than Norway, so if Norway can be used as an indication for the U.S., the
American housing prices should also go down a few percent, and this will likely happen
soon.

5.1.4.5 Forecasts for the rest of the crisis

How the crisis will continue to affect the market is difficult to forecast. The COVID-19
crisis is unprecedented in our modern economy, and it is unknown for how long it will
last, and how bad it will get. The forecasts of the general economy should, however, be
a useful indicator.

49
The economic downturn experienced in the last two weeks of March were enough to
offset the growth of Q1 2020, bringing the U.S. GDP growth of Q1 to approximately
0%. Considering the expectations of the most prominent American experts and
investment banks, the crisis will continue in Q2, and result in severe losses for the U.S.
economy. As seen in figure 16, the GDP expectations were highly divergent, the most
optimistic being -9%, and the most pessimistic -50% annualized compared to Q1
(Charles Schwab, 2020). This large spread itself is highly unusual, with experts
normally having more or less the same forecast, usually only a few percent apart. The
current large spread illustrates the uncertainty of the situation. All these institutions
have experts making forecasts, and when they come up with such different answers, it
proves that nobody knows what will happen, even just two months into the future. It
seems the only thing experts agree on, is that the U.S. real GDP will continue to drop
over 2Q. The average expected drop is almost 30% in one quarter. This will by far be
the largest quarterly GDP drop in U.S. history, as the largest quarterly GDP drop so far
was 10% in 1958. Back then, some 2 million people lost their jobs. So far, the corona
crisis has resulted in over 16 million new jobless claims as of April 1oth 2020.
JPMorgan forecasts unemployment to hit 20% in 2Q (Klebnikov, 2020). Regardless of
which analyst is right, it seems inevitable that a downturn of this magnitude will shake
more or less the whole economy.

50
Figure 16: Institutions' GDP forecasts for 2Q, made at the beginning of Q2

2Q real GDP forecasts (Q/Q)

Firm 2Q 2020 GDP estim ates
Bloomberg Economics -9,0 %
TSLombard -17,7 %
Wells Fargo -22,3 %
Cornerstone Macro -25,0 %
JP Morgan -25,0 %
UBS -25,0 %
IHS Economics -26,5 %
Bank of America Merrill Lynch -30,0 %
Oxford Economics -32,0 %
Deutsche Bank -33,0 %
Strategas -33,3 %
Credit Suisse -33,5 %
Goldman Sachs -34,0 %
Barclays -35,0 %
Morgan Stanley -37,9 %
Capital Economics -40,0 %
Evercore ISI -50,0 %
Average -29,95 %

Source: (Charles Schwab, 2020)

Focusing on the specific business sectors, it is interesting to see what the businesses
are expecting themselves. Figure 17 shows an overview of companies’ earnings
guidance in response to COVID-19, given fairly early in the crisis, 03/06/2020, sorted
by sector. All companies in the survey are MSCI world index constituents. The figure
represents the numbers in absolute companies, not the percentage of the sector. Thus,
the negative responses need to be compared to the positive responses to get a good
foundation for comparison.

It is important to emphasize that earnings guidances are reported by the companies

themselves, and thus may contain a positive bias. This could be an explanation as to
why the total (average) expectation from the businesses themselves are far less
negative than the GDP forecasts made by less partial analysts. Still, it is evident that
some industries stick out as more optimistic, some more pessimistic, and some
disagreeing within the sector. Health care and financials are the two most consistently
optimistic industries. The most pessimistic is, perhaps not surprisingly, consumer
discretionary (Charles Schwab, 2020).

51
Figure 17: Overview of companies' earnings guidances as of March 6th

Source: (Charles Schwab, 2020)

52
Real estate companies are, together with industrials, IT, and energy, among the less
unison industries. I.e. the industries where some companies are optimistic and other
pessimistic. Among the real estate companies, eight reported negative expectations
and 10 reported positive expectations. One reason for this could be the high diversity
within the real estate segment. Real estate companies focused on lodging, retail
(especially discretionary consumption goods), and industrials, are likely to share the
views of the companies in this sector. Real estate companies focusing on segments
such as health care, utilities, and office space for financials would likely be more
positive. Regardless of that, considering the real estate segment as a whole like this
statistic, real estate is one of the few segments that expect COVID-19 to have a neutral
or even positive impact on the earnings. It has to be emphasized that without a global
pandemic, close to all of the real estate companies would probably issue positive
earnings guidance, as that is “business as usual”. Thus, these numbers should not be
interpreted as COVID-19 being good for the real estate sector. What it shows however,
is that the real estate sector, compared to most other sectors, is not expecting the crisis
to affect them as severely.

A consequence of the COVID-19 crisis expected to affect all real estate sectors is the
lack of liquidity that is caused by the scarcity of capital. Bård Bjølgerud, CEO of Pangea
Property Partners, a Nordic commercial real estate brokerage and analysis company,
describes the dynamic like this: With many businesses struggling, the banks focus on
the clients they already have, and the assets these clients already own. They will give
out extensions on loans, accept later payments, and provide other amenities to keep
customers’ operations afloat and avoid bankruptcies. But, the banks are hesitant to
finance investors wanting to expand their portfolios and do investments. As most
investors are dependent on some debt financing to afford, or to turn a positive NPV on
their real estate investments, this stops many transactions from happening. Lower
liquidity in the market results in higher liquidity risk, which in turn lowers the
property values. This is a surprising effect, as one would believe that the lower interest
rates would make debt financing more accessible and have a positive effect on real
estate transactions. The problem however, is not the interest rate. Because of the
higher perceived risk in the market, banks are requiring higher margins on loans. The
effect is that despite lower borrowing rates, the cost of debt could be unchanged or in

53
some cases even higher now than before the crisis, and the commercial real estate
transaction market is expected to more or less halt completely in Q2 2020. It will take
some time to start his market again, so if society does not reopen during Q2 or Q3,
2020 will be considered a lost year when it comes to commercial property transactions
(Bjølgerud, 2020).

5.1.4.6 Long term consequences of COVID-19

After the pandemic is over, one has to expect the society and economy to stabilize.
However, things will likely not go exactly back to “normal”. Experts at CBRE and
McKinsey among others, predict that some corona-caused changes in consumer habits
and the conduction of business will likely stay, and that the virus will accelerate long-
term trends. This paragraph presents some perspectives on potential long term
impacts from the COVID-19 crisis.

Residential

The market for residential real estate could see some long-term effects of the
pandemic. Trends in where Americans choose to live is a complicated field on its own,
and no urban area is the same. But perhaps the most defining long term trend has been
that America saw a huge boom in people migrating to the suburbs after WW2. After
the 2008 financial crisis, the suburbs have seen a decline in popularity, as millennials
have preferred living in urban areas. In the last few years however, there has been a
slight shift, as more millennials are moving to the suburbs. The reason is, most likely,
that they have reached a new stage in their lives: they are starting families and getting
children, which is more convenient and affordable when not living in the downtown of
a big city (Adamczyk, 2019).

There are several reasons the COVID-19 virus could contribute to single-family homes
in the suburbs increasing more in popularity. The most direct reason is that more
people are now are experiencing how unpleasant it is to live through quarantine in a
small studio- or one-bedroom apartment. With virtually everything being closed, none
of the arguments for living in the big city applies anymore. The dependence on public
transport and elevators, and the general crowded nature of city centers, makes it not
just less comfortable, but also riskier in terms of contamination. In addition to this,
there are a lot fewer supermarkets and grocery stores per person, which made the
stores run out of food and toilet paper a lot earlier in the cities than in the suburbs,

54
when people started to stock up. In the suburbs, single-family homes are more
affordable, and if you have to stay inside for weeks or months, a house with a backyard
and more rooms to recreate would naturally be preferable. This could not only cause
more people to move from the cities to the suburbs, but it could also make the people
wanting to move from the suburbs to the city to change their minds. Many “Empty-
nesters” i.e. couples whose children have moved out, and are now living in a house
much bigger than what they need, have historically sold their big suburban houses and
moved to smaller three or four-room apartments in more urban areas. One could
imagine many of these families are now reconsidering (Rivera, 2020).

The indirect effect that could accelerate the same trend, is that the crisis is causing
people to lose their bonuses, parts of their savings (through bonds and stocks), or even
lose their jobs. As living costs in the cities are higher than in most suburbs, the virus
could force people out, as they will get more space for the same price, or the same space
for less, if they move further out from the city center. Many suburbs also have good
public schools, whereas expensive private schools can be the only option for inner-city
families with ambitions for their children. In Boston, this effect is already happening,
as the suburb houses for sale are now on the market for an average of five days instead
of the normal average of 42 days (Rivera, 2020). it seems a likely scenario that suburbs
and smaller metropolitan areas will get more popular in the years to come.

Office

Within office real estate, there has over the last decades been a trend towards
densification and open landscape based office layouts. Public health officials could
decide to increasingly amend building codes to minimize the risk of pandemics in the
future. This could affect the standards for square meters per person and the maximum
amount of people in an enclosed space (V. Gujarl et al., 2020). If measures like this are
taken into action, the trend of increasingly open office plans could slow down, or even
reverse back into the old model with more separate offices, and more space between
the employees.

Already before the crisis, many large companies had begun to cut down on the number
of desks, and are in fact having fewer desks than employees because they know a few
percent of the workforce will be sick, have time off, be on business travel, etc. at any

55
given time. James Gorman, the CEO of the international investment bank Morgan
Stanley reflected on this in an interview with Bloomberg Television. The bank has
moved approximately 90% of its more than 80,000 employees to work from home
during the crisis. Gorman says this transition has been “surprisingly smooth”. He said
that after life returns to normal, he of course would want most staff to return to the
office, but they have proved that the company can operate efficiently with much less
real estate. “Can I see a future where part of every week, certainly part of every month,
a lot of our employees will work from home? Absolutely” he said (Gorman, 2020). If
many companies start to think like this, it would over time result in lower demand for
corporate office buildings.

These two effects could to some degree weigh each other out, in that companies will
want fewer employees in offices at a time, but these fewer employees might require
more space per person. Regardless, the most plausible total effect on the office market
seems to be more uncertainty and a negative net effect.

Health Care

The U.S. has a relatively healthy demographic pyramid. However, it does have a
slightly aging population, with 12.25% of the U.S. population being between 50 and 70
years old (PopulationPyramid.net, 2020). This demographic, the baby boomers, will
over the next years and decades grow into the prime age for independent and assisted
living facilities. The large outbreaks of COVID-19 in such facilities have created strong
fears as their occupants are among the people taking the most damage from the virus.
It is reasonable to think the fear of the assisted living facilities turning into hubs for
pandemics, combined with rules against taking visitors from friends and families into,
will make these facilities less popular in the years to come. Some elderly people of
course may not have any other choice, but for relatively healthy people who are not
dependent on assistance, they may want t0 to stay in their homes longer because of
this fear. This fear could again potentially result in the assisted living facilities
changing their services, offering m0re physical space per person, and stricter
operational requirements (V. Gujarl et al., 2020).

Lodging

It is possible that the fear of pandemics changes how people do their leisure travels. In
the short term, it is plausible that people will avoid traveling overseas, which could

56
give more domestic visitors to hotels, but at the same time fewer international guests.
After a few years however, it is unlikely that a significant amount of people will change
their vacation habits because of the fear of a new pandemic.

Business travel could however change more permanently. The dynamic described in
the office paragraph, of the surprisingly few negative effects of working from home,
could likely affect business travel more than office properties. Sending employees on
planes to other cities and have them staying in hotels just to attend a few meetings can
be a highly inefficient use of the workforce and company funds. On many occasions,
digital meetings solutions are sufficient or even preferable. As companies’ workforces
are now forced to work from home, more companies are implementing and getting
used to having virtual meetings and video conferences. Thus, the corona crisis could
help pave the way for virtual meetings replacing a bigger share of business travel. This
could lead to great losses for hotels, especially in cities with big business centers, but
few tourist attractions, such as Houston, Texas, and Charlotte, North Carolina.

Retail

One of the areas where experts seem to agree the most about the future is retail. For
many years now, traditional retailers have been struggling. It is nothing new that e-
commerce is taking over market shares from brick and mortar retailers. Still, when
people are not able or do not want to leave their homes, it will likely force the many
customers still preferring to shop in stores to do it online. Depending on how long this
pandemic lasts, it could also force people who have hesitated to try online shopping,
the late majority, and laggards, to open up accounts and give internet shopping a try.
Many small local retailers are going bankrupt, and big department stores and malls
like Macy’s and Nordstrom which were already struggling before the crisis are closing
down stores and limiting their physical presence. As physical stores are being pushed
over the edge by the crisis, the local shopping opportunities for many Americans will
be more limited, again contributing to e-commerce’s competitive advantage. Through
these dynamics, COVID-19 is likely to accelerate the shift towards e-commerce (V.
Gujarl et al., 2020).

Industrial

With a change in the retail industry follows a shift in logistics. More goods being sent
directly to the end-user instead of retailers make the logistic more complicated, as the

57
transportation chains are being disrupted. In short, consumers demanding more
deliveries at higher speeds make logistics more important. Already in 2016, a study
conducted by Cushman & Wakefield found that industrial vacancies were at a 15-year
low. The reason for this, they concluded, was the need for industrial properties for
large logistic halls, processing centers, and last-mile distribution centers (Selko, 2016).
Industrial real estate, including logistics, has already been among the best performing
real estate sectors over the last 20 years, and with this trend, it is expected to continue.

Other big potential industry shifts caused by the crisis are on-shoring and near-
shoring. Businesses see that it is risky to have production, storage, and other parts of
the value chain abroad, or generally far away from the end-user, when transport is
stopped and trade is halted. This could drive more American industries to move larger
parts of their value chains back to the U.S., and perhaps closer to the urban areas. A
hike in domestic production would not only be good for the industrial property market;
it would also increase logistics and warehouse demand (CBRE, 2020).

5.2 Quantitative analysis

The quantitative analysis consists of independent analysis based on raw data from
different indexes and assets. This sector is divided into two subsectors: the All REIT
Index, analyzing the REIT segment as a whole, and the sector analysis, going a little
deeper and analyzing the REIT segments separately. The historical returns and risks
of the different asset classes and sub-sectors are presented and evaluated, and this
historical data is then used to create optimal mixed-asset portfolios.

5.2.1 All REIT Index

The FTSE Nareit All REITs Index represents the majority of the U.S REIT market and
serves well as a benchmark for REIT investments. Focusing on REITs as a whole, this
segment will analyze the returns and volatility of the U.S REIT market, and compare
REITs to other financial assets. MPT theory will be used to construct mixed-asset
portfolios, and find optimal weights of REITs for a mean-variance investor.

5.2.1.1 Return
By looking at the indexes for REIT, Stock and Bond return in figure 18, it is
immediately evident that there is a correlation between REITs and the stock market.
Since January 1994, both the REITs and the stock market have grown considerably.

58
The stock market started stronger in the late 90s, mainly due to the IT-boom. This
bubble busted in 2000, and the three asset classes were back at the same level in late
2001. From 2002, the REITs increased significantly more than stocks and bonds as
financing and mortgages were cheap and the real estate market was booming until
2007. The 2007-2008 global financial crisis hit the real estate market extra hard,
bringing the REITs back down to the same level as the stock market and bond market.
Since 2009 there has been a long bull market for both stocks and REITs. At the same
time, bond yields have gradually decreased, resulting in lower returns from bonds.
This has resulted in stocks and REIT investments growing a lot more than bond
investments. By the end of 2017, the REIT and stock market returns were exactly the
same, but the REITs experienced slightly higher growth in 2018 and 2019.

Figure 18: All REIT Index vs U.S. stock market and bond market. Dec 1993- Jan 2020

All REIT Index vs US stock market and Bond market

1 400
1 200
1 000
800
600
400
200
0
93 95 97 99 01 03 05 07 09 11 13 15 17 19
All REITs Stock market Barclays aggregated bond index

Source: (Nareit, 2020c), (French, 2020), (Yahoo Finance, 2020)

In total, the average yearly return of the ALL REIT index was 12.04% between
01/01/1994 and 01/01/2020. In the same period, the average yearly return of the stock
market was 10.99%, and the bond market 5.18%. The worst five-year period for REITs
was 1994-1999, with average annual returns of 7.27%, followed by 2005-2012 when
annual returns were 7.51% on average. For stocks, the worst five-year periods were
2000-2005 with only 0.68%, and 2005-2010 with a 4.80% average annual return. For

59
the bond market, the two worst periods were notably 2016-2020 with 3.12% and 2011-
2015 3.84% average annual return. Diminishing bond yields over the last ten years
have made bonds less attractive, as illustrated by the steadily sinking development of
the ten-year U.S treasury yield from 1993 to 2020 in figure 19.

As for the best performing periods, this was 2000-2005 and 2011-2015 for REITs, with
21.20% and 12.74% average annual return respectively. For the stock market it was
1994-1999 and 2016-2020 with average annual returns of 23.71% and 13.89% return.
Bonds did best from 2000-2005 and 1994-1999 with 6.45% and 6.05% return. The
complete overview of returns over the five-year periods can be found in appendix 7.1.

Figure 19: 10y U.S. treasury yield

10y U.S. treasury yield
9%
8%
7%
6%

5%
4%
3%
2%
1%
0%
93 95 97 99 01 03 05 07 09 11 13 15 17 19

Source: (Yahoo Finance, 2020c)

5.2.1.2 Volatility
Volatility measures the movement in value of an asset in both directions, and thus it is
a representation of the dispersion of an investment. Volatility is the core of the most
common risk measurements for financial assets, making volatility as significant to
analyze as returns.

Historical volatility, also known as statistical volatility, is calculated from the

fluctuations of historical asset prices. The formula for sample variance is used, as the
monthly adjusted prices used in the data set do not contain all the data points in the

60
population. The sample variance uses “n-1” instead of “n” in the denominator, creating
an unbiased estimate with a slightly higher variance compensating for the uncertainty
from not having all the data points. The sample variance is given by the formula:

𝑛
∑𝑡=1(𝑥𝑖 − 𝑥̅ )2
𝜎2 =
𝑛−1

Where:

𝜎 2 = sample variance
𝑥𝑖 = 𝑉𝑎𝑙𝑢𝑒 𝑜𝑓 𝑖 𝑡ℎ data point
𝑥̅ = 𝑆𝑎𝑚𝑝𝑙𝑒 𝑚𝑒𝑎𝑛
𝑛 = 𝑠𝑎𝑚𝑝𝑙𝑒 𝑠𝑖𝑧𝑒

Standard deviation (𝜎 ) is calculated as the square root of the variance, and used
because it is a better parameter for comparing volatility between assets.

A one year moving annualized standard deviation is calculated for every month after
December 1994. This is calculated as the standard deviation over the last 12 months
for every month since January 1994. This calculation is made for the REIT index, the
stock market index, and the bond index. The output can be found in figure 20. The
output shows, perhaps surprisingly, that REITs were the most volatile asset class. This
has a significant connection with the fact that REITs were even more volatile in the
2007-2008 global financial crisis than stocks and especially bonds.

61
Figure 20: One-year moving standard deviations, annualized
One year moving standard deviation, annualized
0,7

0,6

0,5

0,4

0,3

0,2

0,1

0
94 99 04 09 14 19

All REIT Return Stock market Bond index

Source: (Nareit, 2020c), (French, 2020), (Yahoo Finance, 2020)

It is even more evident when observing the standard deviation over five-year periods
in figure 21. The stock market was more volatile than REITs in the periods from 1994
until 2005, but REITs were a lot more volatile during and shortly after the crisis. The
real estate market was perhaps the segment affected most strongly by the crisis.
Especially mortgage REITs, as an extreme amount of both residential and commercial
mortgages defaulted. In the most recent time-period, 2016-2020, the REITs and
stocks were back at the same volatility level. Still, in total, REITs had an annualized
monthly standard deviation of 0.18 since 1994, slightly above the stock market with
0.15, and far above the bonds index with a standard deviation of only 0.04.

62
 Standard
Figure 21: Annualized deviation
standard over
deviations overfive-year periods,
five-year periods
annualized
0,35

0,3

0,25

0,2

0,15

0,1

0,05

0
93-99 00-05 05-10 11-15 16-20 Total

All REIT Stock market Bond index

Source: (Nareit, 2020c), (French, 2020), (Yahoo Finance, 2020)

This volatility analysis indicates that over time there is not much difference in the
standard deviation of REITs and stocks. REITs did have slightly higher volatility, but
this does not seem to be significant, as the main reason is a specific crisis that
happened to affect the real estate market more.

5.2.1.3 Correlation
When creating portfolios, the volatility of the individual assets in the portfolio is of
limited importance. What really matters is the total volatility of the portfolio. The
entire benefit of diversification is based on the notion that different investments follow
different paths and are not correlating 100%. Thus, assets with a low correlation
between them are good for creating portfolios with low volatility.

The correlation between two assets, x and y, is calculated as with the formula:
Σ(𝑥𝑖 − 𝑥̅ )(𝑦𝑖 − 𝑦̅)
𝜌𝑥,𝑦 =
√Σ(𝑥𝑖 − 𝑥̅ )2 (𝑦𝑖 − 𝑦̅)2
Where:

63
 𝜌𝑥,𝑦 = 𝑐𝑜𝑟𝑟𝑒𝑙𝑎𝑡𝑖𝑜𝑛 𝑏𝑒𝑡𝑤𝑒𝑒𝑛 𝑎𝑠𝑠𝑒𝑡 𝑥 𝑎𝑛𝑑 𝑎𝑠𝑠𝑒𝑡 𝑦
𝑥𝑖 𝑎𝑛𝑑 𝑦𝑖 = 𝑠𝑎𝑚𝑝𝑙𝑒 𝑒𝑥𝑒𝑠𝑠 𝑟𝑒𝑡𝑢𝑟𝑛𝑠 𝑜𝑓 𝑎𝑠𝑠𝑒𝑡 𝑥 𝑎𝑛𝑑 𝑦
𝑥̅ 𝑎𝑛𝑑 𝑦̅ = 𝑚𝑒𝑎𝑛 𝑒𝑥𝑒𝑠𝑠 𝑟𝑒𝑡𝑢𝑟𝑛 𝑜𝑓 𝑎𝑠𝑠𝑒𝑡 𝑥 𝑎𝑛𝑑 𝑦
𝐸𝑥𝑒𝑠𝑠 𝑟𝑒𝑡𝑢𝑟𝑛 = 𝑎𝑠𝑠𝑒𝑡 𝑟𝑒𝑡𝑢𝑟𝑛 − 𝑟𝑒𝑡𝑢𝑟𝑛 𝑜𝑓 𝑡ℎ𝑒 𝑟𝑖𝑠𝑘 𝑓𝑟𝑒𝑒 𝑟𝑎𝑡𝑒

The correlations between the three asset classes: REITs, stocks, and bonds, have not
been constant over time. Considering the first period (1994-1999), it is notable that
the difference in correlation between the asset classes was smaller. In this time, all the
correlations were between 0.18 and 0.45, a spread of only 0.27. After 2006, the spread
has been drastically higher, going from 0.72 in 2006-2010 to 0.77 in 2011-2020. One
reason for this is that interest rates were higher in the 1990s, resulting in bond returns
being more similar to that of REITs and stocks. Comparing this period to the others in
figure 22, it is clear that the reason was the higher correlation between stocks and
bonds in the first period and not lower volatility between REITs and stocks in the later
periods.

Figure 22: Correlations between segments over time

Source: (Nareit, 2020c), (French, 2020), (Yahoo Finance, 2020)

Considering what the different periods have in common, it is evident that REITs and
the stock market have the highest correlation among the asset classes in all the periods.
This is not surprising, considering that REITs and stocks are both equities, and that
the REITs are publicly traded and thus a part of the stock market. Since 1999, stocks
and bonds had the lowest volatility, and this volatility is often even negative.

64
Most of the periods, the REITs & Bonds correlations have been somewhere between
the REITs & stocks and the Stocks & bonds correlations. This shows that REITs’
correlation with bonds is higher than stocks’ correlation with bonds. In the interest
rate section under Macroeconomic factors in chapter 5.1.3, it was found that real
estate as an asset class has significant similarities with bonds. The correlations support
this, as it shows that REITs historically have co-moved more with bonds than what
stocks have. Thus, the argument that real estate can act as a hybrid between stocks and
bonds is supported by historical evidence.

5.2.1.4 Beta
An asset’s co-variance with the return of the market portfolio is denoted as beta
(Petersen, Plenborg, & Kinserdal, 2017). Hence, beta is an asset’s sensitivity to
volatility in the market portfolio. Beta represents the systematic risk, and is, according
to CAPM, the only risk an investor will be compensated for taking. Aswath Damodaran
states that the market portfolio should represent the marginal investor’s diversified
portfolio (Damodaran, 1999). As most investors are heaviest weighted in stocks, the
return of the stock market as a whole is used as a proxy for the market portfolio return.
Thus, the group of stock indexes used to represent the stock market in the analysis also
serves as the market portfolio.

The estimation of beta based on historical returns is found using a simple linear
regression with the asset returns as the Y input and the market portfolio return as the
X input. The slope of this regression represents beta. Beta can also be found using the
formula given under CAPM in the theory section. Historical betas are found for REITs
and bonds for horizons of 26 years (all the data), 10, five, and two years.

The return interval used for the calculations has a significant effect on the beta
estimate. This is because assets do not trade on a truly continuous basis, creating a
non-trading bias. This results in assets with high liquidity getting a higher beta
estimate, while lower liquidity leads to a lower beta, everything else being equal
(Damodaran, 1999). The solution for the non-trading problem is to use a monthly time
interval of returns, so the difference in liquidity on the assets won’t affect the beta
estimates.

As most large investors will have a large exposure to the market, low-beta, or negative
beta investments will be preferable from a diversification stance. This is simply

65
because an investment with high beta will amplify market movements, while a low beta
investment will reduce the exposure to market volatility.

Figure 23: REIT, bonds and stock betas

Bet a for differ en t in v est m en t h or izon s

A sset 26 y ea r s 20 y ea r s 10 y ea r s 5 y ea r s 2 y ea r s
REIT s 0 ,6 8 0 ,7 8 0 ,6 9 0 ,5 0 0 ,4 8
Bon ds 0 ,0 0 -0 ,0 2 -0 ,0 6 -0 ,0 6 -0 ,0 4
St ocks 1 ,0 0 1 ,0 0 1 ,0 0 1 ,0 0 1 ,0 0
Source: (Nareit, 2020c), (French, 2020), (Yahoo Finance, 2020)

As seen in figure 23, the choice of horizon affects the beta estimates. The goal is to
estimate the most representative beta for the future. Thus, the choice of period and
horizon is a trade-off. To precisely capture the systematic risk, the beta-estimate
requires a substantial amount of observations. At the same time, the market dynamics
as well as the characteristics of both REITs and other companies are constantly
changing. Furthermore, index constituents in both the REITs, bond, and stock indices
are changing over time. The heavyweight companies of the indices in 1994 are to a
large extent not the same companies that create the bulk of the indices today. Hence,
the more recent data points are more representative of the future beta than the earlier
data points (Damodaran, 1999). Because of this, the 26-year-old data points are not
considered the most relevant. A regression should contain a minimum of 60 data
points to provide a significant result and capture the true systematic risk (Koller,
Goedhart, & Wessels, 2010). This criterion makes two years too short of a reference
period.

The NAREIT All REITs index has added several subsectors of REITs over the years.
The latest extension was in 2015 when data centers and specialty REITs were added to
the index. There have been a few new single REITs added since then, but the index has
been virtually the same since 2015. Hence, the index of the last five years is found to
be the most representative of today’s index.

The five-year time-frame includes only the bare minimum of data points according to
Koller et al., and the 20-year time-frame includes data that may be too old to accurately
represent today's markets. Hence, both perspectives have their pros and cons. In this
trade-off they are at each end of the scale, and they both provide a good picture of how

66
REITs have been connected to the market in the past, which could both be indicative
of the future market beta. Because of this, the 20-year and five-year historical betas
are given some extra focus.

Figure 24: Regression plots against the market portfolio, data from the last five
years

Source: (Nareit, 2020c), (French, 2020), (Yahoo Finance, 2020)

Figure 25: Regression plots against the market portfolio, data from the last 20 years

Source: (Nareit, 2020c), (French, 2020), (Yahoo Finance, 2020)

67
Comparing the 5-year and 20-year regression plots in figures 24 and 25, the difference
in the number of data points is evident. This shows how the 20-year beta has a more
solid basis on data, which is also evident in the higher 𝑅 2 . The bond’s returns are a lot
more unison, indicating a clear pattern. Still, bonds betas for both five and 20 years
have insignificant R-squares of 0.05 for the five-year and 0.009 for the 20-year beta.

The regressions show that REITs have a low to medium beta in the range 0.5-0.78,
while bonds have a slightly negative beta of -0.06 to -0.02. Considering the 5-year
beta, this means that a 1% rise in market returns results in a 0.5% rise in REITs and a
0.06% drop in bonds. Conversely, a 1% fall in the market gives a 0.5% fall in REITs
and a 0.06% rise in bonds. Based on this, REITs seem to work decently for
diversification. Bonds however, seem to be a great way to diversify a stock portfolio.

Even though the beta estimates show that bonds are better for differentiation than
REITs, the significance of the beta estimates is very different. The REIT estimation has
a 𝑅 2 of 0.24. This is, according to Cohen J. (1992) a moderately high degree of
explanation. Considering 20 years of data points, the degree of explanation was
somewhat higher, with an 𝑅 2 of 0.36. The REIT beta also had a highly significant P-
value, indicating that the true degree of explanation cannot, in fact, be zero. The bonds
beta, however, had insignificant R-squares and a P-value of 0,08. This P-value
indicates that the true degree of explanation could, in fact, be zero. The estimated bond
beta is already very close to zero, so it is not surprising if the true degree of explanation
is not statistically significant. Thus, the bond beta is found to be very close to 0, and
the REIT beta in the range of 0.5 to 0.8.

5.2.1.5 Portfolio building

Following the modern portfolio theory (MPT), portfolios are constructed based on the
historical return, volatility, and correlations of the three asset classes. For the chosen
historical periods, tangency portfolios are found, giving the portfolio with the highest
risk-adjusted return for each period.

These portfolios are constructed in four steps: The portfolio return is found as the sum
of the product of the weights of the three asset classes and the historical returns from
each asset class. The returns used are the average annualized excess returns over the
relevant years for each asset class.

Portfolio standard deviation is then calculated with the following formula:

68
𝜎𝑝 = √𝑤12 𝜎12 + 𝑤22 𝜎22 + 𝑤32 𝜎32 + (2𝑤1 𝑤2 𝐶𝑜𝑣1,2 ) + (2𝑤1 𝑤3 𝐶𝑜𝑣1,3 ) + (2𝑤2 𝑤3 𝐶𝑜𝑣2,3 )

Where:

- 𝑤1 , 𝑤2 and 𝑤3 are the portfolio weights of REITs, stocks, and bonds

- 𝜎1 , 𝜎2 and 𝜎3 are the historical variance of REITs, stocks, and bonds
- 𝐶𝑜𝑣1,2 , 𝐶𝑜𝑣1,3 and 𝐶𝑜𝑣2,3 are the covariance between the asset classes,
calculated as:

𝐶𝑜𝑣1,2 = 𝜌1,2 ∗ 𝜎1 ∗ 𝜎2

The portfolios’ Sharpe ratios are then calculated as the portfolios' excess return divided
by the portfolio standard deviation as shown in the Sharpe ratio paragraph under
theory. Once these formulas are set up, the tangency portfolios are found through
iteration, as the weights resulting in the highest portfolio Sharpe ratios. This is done
with the Solver tool in Microsoft Excel.

The tangency portfolios for the different periods can be found in figure 26, and it is
evident the weights are varying over time. This is not surprising, as both returns,
volatility, and correlations have been fluctuating significantly over the 26 years
observed. What is recurring, is that bonds are getting a bigger allocation than the other
asset classes, with a portfolio weight of between 40% and 95%.

The ultimate situation for portfolio building is a negative correlation, as the volatility
of the different investments will then offset each other. Thus, looking at correlations,
bonds and stocks seem like the perfect combination for portfolio building. Historically,
this combination has worked very well. Still, despite the negative correlation, bonds
have become less attractive in many funds, as seen in the analysis of existing asset
managers segment. This is once again because of the low returns from bonds in recent
years. Bonds’ negative correlation with stocks does not make a large impact on a
portfolio’s volatility if the bonds are barely fluctuating in value compared to the
portfolio’s other assets. For such bonds to make an impact big on a portfolio, the bond
weight would have to be very high, and the investor would maybe even have to use
leverage on the bond investments. This can be problematic because of two things; first,
the high weight in bonds compared to stocks would be against the policies of many

69
asset managers. Secondly, financial gearing on the bond investments would be
impossible for many asset managers, especially pension funds, who have limited
mandate for using leverage (Loader, 2016). Because of this, the tangency portfolios
illustrated in figure 26 might not be optimal or realistic for all professional investors.
However, in a world with no constraints, the best asset allocation for risk-adjusted
returns based on the entire 26-year period would be a high weight of about 70% bonds,
18% Stocks, and 12% REITs.

Circling back to the portfolios of the large institutional asset managers, Prudential was
the only asset manager who held as much bonds as the tangency portfolios. Except for
Prudential, the average investment manager and SWF bond holdings were far lower,
at between 25% and 45% in 2018. This is an indication that the high bond weights of
~70% are not realistic for real-world investors, which is further discussed in the sector
analysis.

The development of the tangency portfolios’ Sharpe ratios, illustrated with the purple
line in figure 26, has been fluctuating as well. The highest Sharpe ratio was obtained
in 2011-2015 when it was 1.64. The lowest was, not surprisingly, in 2006-2010, with a
ratio of 0.88. Most years, it is the asset class with the highest Sharpe ratio that has the
highest weight in the tangency portfolio. Conversely, in 2000-2005 and 2016-2020,
REITs and stocks respectively had the highest Sharpe ratios among the asset classes,
but the tangency portfolios still weighted bonds the heaviest in both those instances.
This shows how important correlation is for portfolio volatility, and thus portfolio
weights.

As illustrated with the blue line set against the purple line, the tangency portfolios
always have a higher risk-adjusted return than portfolios consisting of equal parts
REITs, stocks, and bonds. The tangency portfolios also have higher Sharpe ratios than
the single asset class with the highest Sharpe ratio for all of the periods. This can be
observed in the complete table showing weights and Sharpe ratios for all three asset
classes in appendix one.

70
Figure 26: Tangency portfolios weights and Sharpe ratios over time

The Sharpe ratios are measured on the right-side Y-axis. Source: (Nareit, 2020c), (French, 2020), (Yahoo
Finance, 2020)

To more directly showcase the effect REITs have on portfolio allocation, there has been
constructed tangency portfolios for the same periods, but without REITs. Thus, REITs’
effect on risk-adjusted returns is isolated and quantified. The tangency portfolios
without REITs were constructed the same way as the other tangency portfolios, but
the REIT weight is set to zero, and the optimal weights of stocks and bonds are then
found through iteration to give the highest possible Sharpe-ratios without REITs. As
illustrated with the orange versus the blue line in figure 26, the tangency portfolio with
REITs had higher Sharpe-ratios than the ones without REITs. This was through the
entire reference period, although the difference was marginal in 1994-1999 and 2006-
2010, two periods where REITs underperformed compared to other stocks. The
biggest gap was in 2000-2005, where the portfolio with REITs provided 61% higher
Sharpe-ratio than the portfolio without REITs. For the last ten years, the difference
was ~12%, and for the whole period, the average difference was 12.4%. This is a strong
indication that for a mean-variance investor, a tangency portfolio with an element of
REITs, historically, performs better than a tangency portfolio without REITs.

71
The most representative tangency portfolio

In the beta paragraph, the “last five years” and “last 20 years” time perspectives were
both found to be useful indications for the future beta. Following this logic, the
tangency portfolios based on the last five and twenty years of data should also be the
most representative for future returns and volatilities of the REIT index, stocks, and
bonds. The tangency portfolios based on historical returns and volatilities from the
last five years and the last 20 years are calculated in the same manner as the other time
perspectives.

The results are found in Figures 27 and 28 and it is evident that all the excess returns,
volatilities, and Sharpe ratios are significantly different in the two perspectives. All
over, both absolute and risk-adjusted returns were higher over the last five than the
last 20 years. This is not surprising, as the 20 includes a global financial crisis while
five is mostly bull market. REITs however, stick out with having higher annual returns
over the last 20 than the last five years. Despite these large differences in risks and
returns, the weights of the tangency portfolios are highly similar. The largest difference
is the weight in stocks, which is almost double in the five compared to the 20 years’
time perspective. REITs and bonds however, were weighed about the same for the two
perspectives.

Figure 27: Tangency portfolio based on the last five years of data

Last 5 years

Sharpe
Asset Weight Excess return Standard deviation
ratio
REITs 14 % 8,11 % 0,1257 0,65
Stocks 20 % 9,95 % 0,1232 0,81
Bonds 66 % 2,34 % 0,0328 0,71
Portfolio 100 % 4,67 % 0,0377 1,24

Source: (Nareit, 2020c), (French, 2020), (Yahoo Finance, 2020)

72
Figure 28: Tangency portfolio based on the last 20 years of data

Last 20 years

Sharpe
Asset Weight Excess return Standard deviation
ratio
REITs 15 % 11,13 % 0,1935 0,58
Stocks 12 % 5,50 % 0,1501 0,37
Bonds 73 % 1,66 % 0,0347 0,48
Portfolio 100 % 3,56 % 0,0436 0,82

Source: (Nareit, 2020c), (French, 2020), (Yahoo Finance, 2020)

The optimal portfolio consists of about 15% REITs, 12-20% stocks, and 70% bonds.
The return and Sharpe ratio of this portfolio will depend on the overall market but
based on historical movements it produces an annual excess return of around 4% and
a Sharpe ratio of approximately 1.0. An important takeaway is that even though the
returns and risks of the market were completely different in the two considered time
perspectives, the optimal tangency portfolio included almost exactly the same weight
of REITs. This indicates that with a long time horizon, a 15% allocation to REITs can
be good for the portfolio regardless of the economic climate. This optimal REIT weight
of 15% is significantly larger than what the majority of large-cap investors are holding,
demonstrating that most large-cap investors would benefit from investing more in
REITs.

5.2.1.6 Portfolio reaction to the COVID-19 crisis

As of May 1st, 2020, stocks, REITs, and bonds all seem to have passed the worst part
of the COVID-19 crisis, at least for now. Regardless of the final result not being known,
the part of the crisis that has already passed offers interesting information about the
asset classes’ reaction to the crisis.

Figure 29 is indexed so that the asset classes and the tangency portfolio all start at the
same level (100) on February 1st 2020. The daily returns indicate that the asset classes
co-move similarly during the crisis as when the market was “normal”. I.e. Stocks and
REITs are still more volatile than bonds, and they also have a higher correlation with
each other. This is where the similarities with the “normal” market conditions stop.

73
Not surprisingly, the three months from February through April have been extremely
volatile. As seen in figure 31, compared with the standard deviations in figure 27, the
annualized standard deviation of the stocks and bond markets were more than five
times as high as the last five-year average. REIT volatilities rose even more, with an
annualized standard deviation of 0.74, which is six times as high as its five-year
average.

Since the market started turning downward on February 19th until the lowest point so
far in the crisis: March 23rd (stocks), March 24th (REITs), and March 18th (bonds), the
three asset classes had lost 34%, 42%, and 7% respectively. Since then, all three asset
classes have improved significantly through April. The stock market did a spectacular
rebound despite the high number of layoffs and general market uncertainty, resulting
in the S&P-500 seeing its highest monthly return since 1987, and the best April return
since 1938 (DeCambre, 2020). Because of this, as of May 1st, the losses were reduced,
and REITs were down 22%, stocks were down 14%, bonds were up 1%, and the five-
year tangency portfolio down 3% since the crisis begun to impact the market (Nareit,
c, 2020) (Yahoo Finance, 2020) (French, 2020). As the two tangency portfolios have
such similar weights, their returns have been about the same over the crisis. Because
of this, only the tangency portfolio based on five years of data is included in the
comparison.

74
Figure 29: 5y Tangency portfolio VS All REIT index, stock market, and bond market

Source: (Nareit, c, 2020) (Yahoo Finance, 2020) (French, 2020)

In the “Beta” segment, it was found that the REIT index had a market beta of 0.5-0.8.
This should indicate that when the market falls 14%, REITs should fall around 7-10%.
Conversely, REITs fell 22%, proving the historical beta estimate drastically wrong.
When an asset or index moves the same direction as the market, but to a higher degree
than the market, it has a positive beta. Thus, from February to May seen as a whole,
REITs have had a significantly positive market beta, which of course is the opposite of
what investors want when they are seeking portfolio diversification. Calculations from
the daily returns between February 1st and May 1st confirms this, as seen in the
regression in figure 30.

REITs’ market beta during the COVID-19 crisis has been 1.03, around twice as high as
before the crisis. It is also worth noticing that the beta’s coefficient of determination
(R^2) now is 0.73, which is significantly higher than before COVID-19 (Nareit, c,
2020) (French, 2020). Both these parameters indicate that the reason REITs are
falling is the general movements of the market and not changes in the underlying real
estate assets.

75
Figure 30: REITs market beta and R2 from Feb to May

Circling back to the COVID-19 segment of the qualitative analysis, it was found that
several real estate segments were doing relatively fine. For instance, house prices have
not fallen, and data-centers and cell-towers are doing well. Intuitively, the same
conclusion could be drawn from this as from the beta-analysis: It can seem like an
overreaction that REITs prices are falling so much more than other stocks.

Figure 31: Annualized returns and volatilities in the COVID-19 crisis

Annualized from daily data, February to May 2020

Exess return Standard deviation Sharpe ratio
REITs -43,05 % 0,74 -0,58
Stocks -20,29 % 0,62 -0,33
Bonds 13,36 % 0,16 0,82
5y Tangency portfolio -4,92 % 0,20 -0,24

Source: (Nareit, c, 2020) (Yahoo Finance, 2020) (French, 2020)

The pandemic is far from over, and there is a real possibility we have still only seen the
beginning of the crisis’ impact on the financial markets. Hence, this should not be
interpreted as a final answer for which asset classes made it best and worst from the
crisis. Still, as of May 1st, 2020, bonds were the undisputed winner between the assets
analyzed. REITs have so far been doing the worst during the crisis, 22% down from
the top in February. Overreaction or not, this indicates that in pandemic-induced
financial crises where people have to quarantine, the markets punish real estate more
than the average stock on the Nasdaq and NYSE.

76
5.2.2 Sector Analysis
The REIT sector has now been explored as a single entity, but there are differences and
nuances within REITs across different sectors. In this paragraph, different REIT
sectors will be analyzed and compared on their past returns, volatility, and the
correlation between them and other assets. The sub-sector REIT indexes provide a
more focused and detailed picture of the real estate market.

The sector analysis will have a time perspective of 20 years, from January 2000 to
January 2020. The time interval is selected because some sectors do not have available
return data before 2000.

5.2.2.1 Returns
Figure 32 shows the annualized average returns of each sector for the time interval.
The period was very good for REITs, with every sector beating the performance of the
stock index.

Figure 32: Annualized Average Returns, 2000 - 2020

Source: (Nareit, 2020c) (French, 2020)

To visualize the relative performance of each sector the returns are indexed in figure
33, for the years 2000-2020. The figure shows how superior the performance of Self-
Storage and Healthcare have been in the period, which distorts and hides the
performance of the other sectors. A logarithmic scale version of the graph can be seen
in appendix two and shows that the various sectors experience much of the same
variations in returns and mostly follow the same trend. The sectors differ in their
sensitivity to those trends which results in large differences in total return over a 20-
year period between some sectors.

77
Figure 33: Indexed Returns since 2000 by REIT sector

Source: (Nareit, 2020c) (French, 2020)

In the period 2000-2020 the best performing sectors where Self-Storage and
Healthcare with annualized average returns of 19.0% and 17.8% over the period
compared to the All REIT-index’s return of 13.5%. On the other end of the scale are
Commercial Mortgage and Lodging, only producing annualized returns of 3.2% and
6.5% in the same period. Stocks performed poorly compared to REITs during these 20
years with only a 6.2% annual return, only outperforming Commercial Mortgage
REITs. Stocks’ poor performance is due to two financial crises. The Dot-com in 2001-
03 and the global financial crisis in 2007-08 which both saw the stock market fall over
40% from all-time high peaks.

In the years prior to the global financial crisis, all sectors performed very well, and
where largely unaffected by the Dot-com bubble. In this period, Commercial Mortgage
REITs outperformed stocks and all other REIT sectors, with average annual returns of
26.2% compared to an unweighted average of 16.7% for the other REITs between
2000-2007. As mentioned before, REITs were affected significantly by the global
financial crisis of 07-08, and Commercial mortgage REITs were hit especially hard
losing 90% of their value from the start of 2007 to the end of 2009. While REITs
overall recovered quickly and had regained much of their losses by late 2010, mortgage
REITs still have not recovered from their losses in the 07-08 crash.

78
In the ten years from 2000 to 2010, Healthcare was the best performing sector,
returning average annual returns of 23.9%. Average annual returns are the average of
all monthly returns in a given time period which are then annualized. In figures 34 and
35, the annualized average returns for each sector can be seen for two 10-year periods,
confirming the image given by the indexed returns in figure 33. In the first decade,
REITs where an overall very good asset, all outperforming stocks besides commercial
mortgage.

Figure 34: Average Yearly Returns, 2000-2010

Source: (Nareit, 2020c) (French, 2020)

Returns for 2010 to 2020 are graphed under in figure 35 and shows that Self-Storage
continued to deliver high returns while Healthcare performed significantly worse than
the previous decade. Industrial, Commercial Mortgage, and Residential are the best-
performing sectors following Self-storage, with annual returns off around 18%.
Commercial Mortgage performed much better as it was recovering from the crisis,
returning 18,7% annually. As mentioned, 2010 to 2020 was a much worse period for
Healthcare than the previous, with returns 12 percentage points lower. Retail is
another sector that saw substantially lower returns, falling nearly 6 percentage points.

79
Figure 35: Average Yearly Returns, 2010 – 2020

Source: (Nareit, 2020c) (French, 2020)

5.2.2.2 Volatility
To measure volatility in returns the annualized standard deviations are calculated for
each sector. In modern portfolio theory, volatility is seen as a measure of how risky an
asset is. Higher volatility in past returns is believed to indicate a higher risk for a
portfolio. In figure 36, the annual standard deviation for each sector is shown for four
time periods of five years. The volatility over time for each sector has followed the same
trends as the All REIT index shown in section 5.2.1. The overall volatility increased
drastically during the financial crisis of 07-08 before it leveled out at approximately
the same level as before the crisis.

The sectors which historically have had the most volatile returns are Lodging,
Industrial, and Commercial Mortgage. The three sectors all had returns with annual
standard deviations of around 0.29 for the past 20 years. While the rest of the REIT
sectors show standard deviations around 0.20 over the same period. Industrial and
Commercial Mortgage returns were especially volatile during the financial crisis, with
standard deviations of around 0.5. Both have since seen less volatility and are not
among the sectors with the highest standard deviations the following decade. Lodging
returns has been the most volatile of all sectors for all periods except between ‘05 and
’10. Stock returns have been less volatile than all REIT sectors for the full period and
also for most of the sub-periods of 5 years.

80
The least volatile sectors are the group of six to the right in the chart in figure 36, with
similar standard deviations in returns. The two least volatile of them are Residential
and Self-Storage at 0.186 and 0.193.

Figure 36: Standard Deviation per Sector, 5-year Periods Annualized

Source: (Nareit, 2020c) (French, 2020)

The Sharpe ratio of each sector for the last 20 years can be seen in the table below. The
most volatile sectors also have the lowest Sharpe ratios, meaning that the investor has
not been compensated for higher risk in these assets. Self-Storage and Healthcare have
the highest risk-adjusted returns of the REITs and have much higher Sharpe ratios
than stocks. Bonds does, however, have the highest Sharpe ratio of all the assets.

Figure 37: Sharpe Ratio 2000 - 2020

Source: (Nareit, 2020c) (French, 2020)

81
5.2.2.3 Correlation
When looking at specific sectors to include in a portfolio the correlation with the rest
of the portfolio is important to consider. In the graph under, correlation in returns
between different REIT sectors and the stocks and bonds can be seen. Home Mortgage,
Self-Storage, and Healthcare have the lowest correlation with stocks, while Lodging
and Office have the highest correlation coefficient with stocks. Individual REIT
sectors' correlation with stocks over time have not differed from the trend of the ALL
REIT Index. Correlation for all sectors increased substantially for the years 2005-2010
and remained at a high level the following five-year period, before decreasing again
between 2015-2020. For the last five-year period the correlation coefficient of Self-
Storage and Healthcare decreased substantially more than the other sectors, to levels
close to zero at 0.03 and 0.14.

Figure 38: Correlation of REIT sectors with stocks and bonds. 2000 - 2020

Source: (Nareit, 2020c) (French, 2020)

Bonds are another asset class often included in a portfolio and it is therefore also
relevant to look at the correlation of REIT sectors correlation to bonds. Previously in
the paper it was found that REITs’ returns overall are more correlated with stocks than

82
with bonds, and it is, therefore, no surprise that all REIT sectors are less correlated
with bonds than stocks. There is also a clear trend that sectors that are less correlated
with stocks are more correlated with bonds.

Correlations between the sectors of REITs are as expected high, especially among the
equity REITs. The Mortgage Home sector is considerably less correlated with the other
sectors, with a coefficient of around 0.3 for most sectors. Among the equity REITs,
Lodging and Self-Storage are the least correlated with the rest of the sectors, with
coefficients of around 0.7 which is still high.

5.2.2.4 Tangency Portfolio

In this section, individual REIT sectors are added to portfolios to explore the
differentiation benefit of them. The portfolios are constructed the same way as
described in the previous chapter, but using data from individual sectors instead of the
All REIT Index.

The portfolio that maximizes the Sharp ratio, also when individual REIT sectors are
available, consist primarily of bonds. Under the assumption of no short sale and with
no other constraints, the tangency portfolio for the time interval, 2000 – 2020,
allocates 82.9% to bonds, 5.5% to stocks, and 11.6% REITs. Of the REITs, 8.9% are
allocated to Self-Storage, 1.7% to Home Mortgage, and 1.1% to Residential. Tangency
portfolios for different time intervals can be seen in figure 39 below.

83
Figure 39: Tangency Portfolio of bonds, stocks, REIT sectors.

Source: (Nareit, 2020c) (French, 2020)

Bonds have the highest Sharp ratio and the lowest correlation to any asset or sector
and are therefore naturally the primary component in most of the portfolios. Stocks
have as mentioned a lower Sharpe ratio than most of the REIT sectors, but stocks also
have a small negative correlation with bonds which is preferable in a portfolio.
Therefore, stocks still have a presence in optimal portfolios ahead of sectors that offer
higher Sharpe ratios. Under different constraints to the minimum and maximum
allocation to different assets (see appendix 7.3), the REIT allocation is made up of
mostly of Self-storage, Home Mortgage, and Residential. Self-Storage is the favored
REIT sector in most portfolios, as it has the highest Sharpe ratio. Even though it is also
one of the highest correlated sectors to bonds which constitutes a major part of most
portfolios.

The healthcare sector has the second-highest Sharpe ratio of the REITs but has the
highest correlation with bonds of the sectors and is therefore not optimal in a portfolio
with a large weight in bonds. If the allocation to bonds is restricted to a low level,
Healthcare becomes a more attractive sector and is included.

84
The tangency portfolio for the period from 2010 to 2020 still consists of a large
allocation to bonds, making up 79% of the portfolio. Stocks performed much better
this decade and make up 16.5% of the portfolio. REITs made up just 5% of the
portfolio, with Self-storage, still the largest component, while Healthcare is not
included.

Looking at just the last five years, from 2015-2020, the tangency portfolio is very
different. Industrial REITs have performed very well and have a much higher Sharpe
ratio than any other assets, at 1.77 against the second highest at 0.93. The large SR of
Industrial causes the tangency portfolio to allocate 77% to it, and 23% to bonds. If the
Industrial REIT sector is excluded, the tangency portfolio weights are almost identical
to that from the last 20 years, with 11.6% allocated to REITs, 68.3% to bonds, and
20.1% to stocks.

Overall, the inclusion of different REIT sectors to the portfolio building does not affect
the overall allocation to REITs in any substantial way, compared to the All-REIT
Index. The allocation to REIT sectors was a bit lower than to the All-REIT Index, but
still at a similar level.

5.2.3 Multi-Factor Regression

The expected return of an asset reflects the compensation for the time value of money
and risk. The most basic model for the relationship between risk and return is the
CAPM model, in which exposure to market fluctuations is the only risk factor. There
are, however, multiple models that explore different components that compensate for
various sources of risk. The Fama French 3 factor model is perhaps the most well-
known of these models. In this section, REIT’s exposures to different systematic risks
are explored to find out what kind of portfolios REITs fit best with.

In this section, a total of 8 different factors will be used to explain the excess return of
REITs. The factor models used are the Fama and French five-factor model and
Carhart’s momentum factor. Additionally, two bond factors are explored as an attempt
to capture REITs' sensitivity to interest rate changes.

The methodology for capturing interest rate exposure is based on the methodology of
Van Nieuwerburgh in the article “Why are REITS Currently So Expensive?” (Van
Nieuwerburgh, 2019). The article uses the 10-year constant maturity treasury bond
rate as interest exposure as Van Nieuwerburgh claims it is the most salient for real

85
estate investors. This factor is denoted as GS10 in the models and referred to as
treasury factor. The other bond factor is the Barclays Capital U.S. Aggregate Bond
Index (AGG) used in other sections of this paper and gives exposure to investment-
grade fixed-rate bonds in the US. This factor is denoted as a bond index in the models
and is not used together with the treasury factor, as treasury bonds are a significant
part of the bond index.

The complete risk factor model, with every factor included for expected excess return,
is given by the following formula:

𝑅𝑖,𝑡 −𝑅𝑓,𝑡 = 𝑎𝑖 +𝛽𝑖 (𝑅𝑀𝑡 − 𝑅𝑓,𝑡 )+ 𝑔𝑖 (𝑅𝐺𝑆10 − 𝑅𝑓,𝑡 ) + 𝑠𝑖 𝑆𝑀𝐵𝑡 +ℎ𝑖 𝐻𝑀𝐿𝑡
+ 𝑟𝑖 𝑅𝑀𝑊 + 𝑐𝑖 𝐶𝑀𝐴 + 𝜌𝑖 𝑀𝑂𝑀

Figure 40 shows the results of the regressions on the excess return of the All REIT
index as the dependent y-variable and the various factors as independent x-variables.
The table shows the estimated beta coefficients for the various factors, as well as the
p-value for the estimation. The table is divided into different combinations of factor
models and factors in a multiple linear regression. The standard one-factor CAPM
model in the first column shows a beta to the stock market of 0.78 which is significantly
different from one. The two-factor model with the excess return on stock and the 10-
year treasury, show a negative treasury beta, meaning that when the yield on interest
increase, the REITs value decrease. The estimate is however not statistically
significantly different from zero at significance level set at 0.05. The p-value of the
treasury beta is 0.6353, which means that there is a 63.53% probability of obtaining a
result at least as extreme given the null hypothesis, that beta is zero, is true.

86
Figure 40: Regression output on All REIT Index as dependent variable

Source: (Nareit, 2020c) (French, 2020) (FRED, 2020)

The last row in the table gives the adjusted R-square of the model. R-square gives the
proportion of the variance in the dependent variable that is explained by the
independent variables. The adjusted R-square is adjusted for the automatic increase
adding more variables to the model leads to.

The table shows that the CAPM explains 35.9% of the variance in REITs, this is in line
with the average explanatory power of CAPM on stocks, which is approximately 30-
40% (Benninga, 2014, s. 92).

The treasury factor is consistently statistically insignificant and lowers the adjusted R-
square coefficient of the model with various factors. The treasury factor explains no
variation in the returns of the All-REIT index for all combinations of factor models and
is therefore excluded from other multifactor models.

In the belief that some REITs have exposure to fixed income-oriented risks given the
underlying businesses, another factor is added considering the treasury factor's

87
inability to explain variation. Regression against the bond factor and the market factor
reveals that REITs have a high factor exposure against fixed income-oriented risk. The
exposure is estimated to 1.47 and is highly significant. The R-squared increases around
6 pct. points from the CAPM-model, and 9 pct. points from the five-factor model.

Column 4 shows the result of the regression against the Fama French 3 factor model,
where size and value factors are added. REITs have an SMB (small-minus-big)
exposure of 0.28 and HML (high-minus-low) exposure of 0.56, which both are
statistically significant, meaning that exposures to these risks explain parts of the
return in REITs. The positive factor exposures to SMB and HML indicates that REITs
behave like stocks with relatively low market capitalization and stocks with high book-
to-market ratio. Intuitively it is expected that REITs have a high book-to-market ratio,
as REITs are asset-heavy and required to have a high payout ratio.

The three-factor model increases the R-square of the model substantially, from 36%
to 47%, compared to the CAPM. The alpha estimate for the CAPM was only just
statistically insignificant with a P-value of 0.0532. The three-factor model gives a
lower alpha estimate with a higher P-value, meaning it is less plausible that the
estimate is different for zero.

Column 6 shows Carhart’s four-factor model, in which a momentum factor is added to

the Fama French 3 factor model. REITs have a small negative exposure to the
momentum factor, but the estimate is insignificantly different from zero.

Fama French’s updated five-factor model can be seen to not explain substantially more
than the three-factor model, as R-square only marginally increases. REITs have a
positive exposure against RMW, but the estimate is insignificant for all model
compositions. The investment factor, CMA, does also give insignificant estimates, so
neither RMW nor CMA have any explanatory power for the variation in REITs.

The alpha estimate becomes lower for each factor that is added to the regression,
indicating that the factors explain more of the variation. The estimates are all
statistically insignificant at confidence level 95%, indicating that REITs do not produce
an additional return that cannot be explained by market exposure or other common
equity risk exposure such as value and size. This may indicate that REITs do not have
characteristics that set them meaningfully apart from other stocks, as their return can
be explained by the same risk factors as stocks. This weakens the argument for that

88
listed real estate should be treated as an individual asset class the same way bonds and
private real estate is.

The five-factor model does not add additional explanatory power over the three-factor
model, and evidently REITs do not have exposure to the investment nor the
profitability factor. The best model for estimating the expected return of REITs are
therefore the three-factor model added the bond index factor. The full model with all
the factors is, however, still used further down in the model, as it is deemed relevant
to explore the factors that are on individual sectors as well.

5.2.3.1 Industry comparison

Figure 41: Regression output on various stock industries, 2000 – 2020.

Source: (Nareit, 2020c) (French, 2020) (FRED, 2020)

To add perspective on the factor model regressions, the same has been applied to 10
different stock industries. The industry returns are value-weighted indexed returns
from 10 different industries obtained from Kenneth French’s data library (French,
2020). The description of the industries can be found in appendix 7.3.

REITs have a similar exposure to the market as Non-Durable consumer goods, Health
care, and Utilities. These are all industries that intuitively are less exposed to market

89
downturns, as the products can be described as essential, and are among the last
services that are cut in economic hardship.

REITs have among the highest exposure to size and value factors of the industries. The
comparison underlines that REITs behave like small firms with considerable assets on
their books. REITs are not exposed to the profitability, investment, or momentum
factors. Six of the industries show significant exposures to profitability, in which all
have positive exposure except HiTec. Industries with lower market beta seem to have
a higher exposure to the investment factor, as the industries with the lowest beta to
market have the highest significant exposure to CMA.

REITs have, as mentioned, very high exposure to the bond index, an exposure it is
unique in having. Only Utilities have high positive exposure to the bond factor, which
is not surprising as utilities, like REITs, are well known for high dividends. The
implication is that REIT behaves more like investment grade fixed income products
than other stock categories.

The R-squares of the model on the various industries show that the factors have
relatively low explanatory power on REITs. While the R-squares on REITs are low,
industries such as Energy and Utilities have lower. With the bond index factor added,
the R-square of Healthcare is also lower. But, considering only the equity factors of the
established models, the R-square of REITs is consistently lower.

By comparing REITs and various stock industries we can see that REITs’ exposure to
common factor models is not unique and further weakens the argument for REITs to
be considered a unique asset class.

5.2.3.2 Sector exposures

REIT sectors are also explored using the same 7 factors as above, to explore inter-
sector differences. The stock beta varies substantially across sectors, with Lodging and
Industrial at the high end at 1.26 and 1.11, and Healthcare and Self-Storage on the low
end at 0.51 and 0.49. The two mortgage REITs have very different exposures to the
stock market, as home has the lowest beta at 0.41 and commercial are among the
highest at 0.97. The exposure to the stock market beta can be interpreted as exposure
to the economic cycle and would be consistent with how the sectors intuitively are
perceived to be affected by economic conditions. Healthcare and Self-storage offer

90
services that are necessities and not something where the need and consumption
increase or decrease with consumers' economic situation. Lodging is, however,
especially exposed since a large portion of hotel stays are for vacations.

Figure 42: Regression output of various REIT sectors

Source: (Nareit, 2020c) (French, 2020) (FRED, 2020)

Van Nieuwerburgh (2019) found patterns in differences among sectors' exposure to

interest rate factors that can be explained under a lease duration perspective. Sectors
with typically long lease agreements, like Industrials and Healthcare, have higher
exposure to interest rate changes. Sectors like Lodging and Residential which have
shorter leases are less exposed to interest rate risk. In our data a similar pattern can
be seen in the differences between sectors' exposure to the bond index AGG. Lodging
has, at 0.84, far lower exposure to the bond index than any other sector, and
Residential has the second-lowest exposure. Industrial and Healthcare have exposure
coefficients at 2.80 and 2.40, which is far above the other sectors.

Exposures across sectors are the most consistent for the value and size factors. All
sectors have positive significant exposure to size and value, except Healthcare and
Commercial Mortgage for which the size factor is insignificant. Exposure to the

91
profitability and investment factors are for the majority insignificant, with only one
sector having significant exposure to each factor. Relative to the difference in exposure
seen across stock industries in the section above, the variation among REIT sectors are
minor. Overall, all sectors trade like small value stocks that have much in common
with bonds.

6 Conclusions
6.1 Alternative ways to invest in real estate
Real estate includes many asset classes, both the obvious like residential and offices,
to the less obvious such as cell towers and timberland. Numerous specialized financial
products will give investors exposure to the real estate market. This thesis has focused
on the most common, which is direct real estate, and REITs.

Both direct real estate and REITs were found to have their pros and cons. Perhaps
most notably, direct real estate tends to be highly illiquid, as it takes weeks or months
of preparation, analysis, due diligence, and legal work to sell or buy a property.
Publicly traded REITs allow investors to buy a share of a property or portfolio of
properties, just like any other publicly traded stock. Thus, REITs allow investors to
make more liquid placements in the real estate market. The throwback of course, is
that REITs then become more volatile, and have a higher correlation with the stock
market as a result of it being listed. The choice does, in a way, provide a trade-off
between liquidity risk and market risk.

6.2 Optimal allocation for REITs in mixed-asset portfolios

The quantitative analysis is based on historical returns, volatilities, and correlations of
hundreds of REITs, and thousands of stocks and bonds from the public American
markets between 1994 and 2020. Investments have been ranked after their risk-
adjusted returns, where the investments with the highest Sharpe ratios are regarded
as optimal.

In the reference period, REITs had the highest returns, closely followed by stocks.
Bonds were far behind in terms of absolute returns but for a lot of the time it had the
highest Sharpe ratio due to its low volatility. For most investment horizons, REITs and
stocks were interchanging having the highest Sharpe ratio. Still, REITs’ relatively low

92
beta of around 0.5 towards the stock market and bonds give a diversification effect
significantly improving the Sharpe ratios of the constructed tangency portfolios. For
most of the different investment horizons and periods over the last 25 years, the
highest risk-adjusted returns were achieved by allocating between 10% and 25% of the
total portfolio to REITs. Judged by the two periods declared the most representative
for the future, mean-variance investors should invest 15% of their total portfolio in
REITs.

6.3 How do the sub-sectors of REITs contribute to portfolios?

The analysis of sub-sectors of REITs is executed similarly to the analysis of the All-
REITs Index. The analysis shows that almost all REIT sectors have had higher average
returns over the past 20 years, with especially Self-Storage and Healthcare performing
very good returning over 15% annually. While REITs sectors have delivered good
returns, the volatility in the returns has also been high. Sub-sectors of REITs have all
been more volatile than the stock market, for all five-year periods.

Using the sub-sectors of REITs in the portfolio building did not change the weight
allocated to REITs substantially. Bonds remained the biggest component of the
optimal portfolio. The Sharpe ratio of the optimal portfolio did, however, get a much
higher Sharpe ratio, than when using the All REIT index. This is mainly because this
allowed for a higher weight of Self-Storage REITs, which had a much higher Sharpe
ratio than the All-REIT index.

6.4 What investor types benefit more or less from real estate?
The analysis finds little consensus among large-cap investors’ portfolios when it comes
to real estate investments. Some managers held only 0.5% real estate, others held close
to 25%. Traditionally, real estate investments have been popular among insurance
companies and pension funds, as real estate’s predictable and long term cash flows
match these investors' predictable and long term expenses. Research of asset
managers’ portfolios found this to still be the case, especially when it comes to direct
real estate. Based on this, it is found that investors who are more than averagely risk-
averse, have a long investment horizon and an endurance for illiquidity could benefit
from a high exposure to direct real estate. This group of investors includes family

93
offices, life insurance companies, pension funds, and mutual funds with a low-risk
profile.

Conversely, investors with a shorter investment horizon, that need the ability to
liquidate on short notice, are seeking higher returns, and has a higher appetite for risk
should avoid direct real estate.

When investing in REITs, the illiquidity risk is lower, but the higher correlation with
the stock market results in higher volatility and market risk. The fact that REITs are
more similar to stocks also makes them more suiting for the common investor. The
diversification effect historically made all mean-variance investors benefit from
REITs. Thus, REITs are good assets for all investors seeking to maximize their risk-
adjusted return. The exemption is investors holding large amounts of unlisted real
estate, as they could perhaps get a larger exposure towards real estate than they are
seeking. Unless the fund in question is one specializing in a specific asset, such as pure
real estate funds, all large-cap investors should make sure to diversify their
investments.

6.5 What are the effects of COVID-19?

At this point in time, nobody can yet know for certain what effects the COVID-19 crisis
will have on the real estate market or REITs. So far, the crisis has affected some sectors
strongly while others have been less affected. Hotels, shopping malls, and casinos are
facing unparalleled losses, while a few REITs, mostly in the data-center and cell tower
industry, have seen their market values rise during the crisis. Most REITs however
have lost about as much or a little more than the over-all stock market, only with a
slight delay.

The unlisted property market has not been the main focus of the thesis. It is also more
difficult to accurately monitor this market. Yet, data from the private housing market
was made available and has been analyzed. This revealed a far less dramatic change in
the housing market than in REITs, as of May 2020. The housing market, just as REITs,
is not a sufficient proxy for the whole unlisted real estate market. Still, the vast
difference in COVID-19 reactions for the two markets does contribute to the notion
that REITs may have dropped so much so far in the crisis because of their high short
term correlation with the stock market, and not because of a collapse in underlying

94
property values. This shows that REITs did not offer any downside protection, at least
not in the short term, during the COVID-19 crisis.

Bonds were down a lot less than the two asset classes, perhaps not surprising
considering the low historical volatility. Because of the mixed tangency portfolio’s high
weight of bonds, the mixed-asset portfolio returned 3% from February 1st to May 1st.
Considering the stock market was down 14% over the same period, this is a relatively
solid return.

Bond investments have been increasingly unpopular after the 2008 financial crisis, as
stocks have performed astonishingly well, and bonds have not. Even in this period,
investment-grade corporate bonds have returned strong Sharpe ratios. As such, bonds
have been an underrated asset class but was the savior of the tangency portfolio during
the corona crisis. As more investors and analysis realize this, a long term effect of the
COVID-19 crisis could be an upswing in bond investments.

There are plenty of different possible scenarios for the long term effects: Changes in
where and how we want to live, work, travel, and shop could all be affected by the
crisis. Some changes will probably be good for the real estate sector; other changes will
be bad. The only thing that seems for certain, is that behavioral changes will cause
alterations in the demand for real estate in a post-corona world and that real estate
investors and operators will be wise to keep an eye on the future and take action to
improve their portfolios in reaction to the changes.

6.6 Final Conclusion

To sum up, this thesis has found several answers to the main research question Why
should large-cap investors include REITs in their mixed-asset portfolios?

REITs were found to be the simplest way to include diversified real estate investments
into mixed-asset portfolios. Although REITs do not perform exactly like their
underlying properties, they succeed in giving the investors some of the same exposure
to the real estate market as direct real estate investments. These exposures include
characteristics such as secure long term dividend cash flows, Low market beta, and a
natural complete or at least partial hedge against inflation.

95
Focusing on the numbers, the quantitative analysis found that REITs had not been as
stable and “boring” as most people, the authors included, would expect. In fact, since
1994, REITs had a slightly higher return and volatility than the total stock market. An
investor owning only REITs over the last 20 years would have achieved a higher risk-
adjusted return than an investor owning the market portfolio.

Still, to achieve the highest possible risk-adjusted return, investors will want to
diversify, and mean-variance investors were found to be best off when holding about
15% of their assets in REITs. This can be done very easily, even without analyzing or
doing any research as to what REITs to buy. Merely by investing in an index consisting
of all publicly traded American REITs in combination with stock and bond indexes,
investors will improve their risk-adjusted return compared to not including any
REITs. This has been proved by analyzing historical returns and was the case for all
different time perspectives after 1994.

Focusing on the big picture, real estate is a core asset class, and a big asset class: For
most people, real estate is the asset representing the largest allocation of their total
wealth. It is an asset that everyone will be dependent on for the foreseeable future; not
only in the form of their home, workplace, schools, hospitals, and shops. But also
because the production of goods is dependent on it and the postal and transportation
services are dependent on it. Our phones wouldn’t work without cell towers, and
without data centers the internet would collapse. Considering this, it is no surprise
that such a fundamental part of our lives should also be part of most large-cap
investors’ portfolios.

96
7 Appendix
7.1 Appendix 1, Tangency portfolios, returns, volatilities and Sharpe ratios of
ALL REIT index, All stocks index, and bonds index
Total

Asset Weight Excess return Standard deviation Sharpe

REITs 10 % 9,47 % 0,1794 0,52804
Stocks 13 % 8,37 % 0,1483 0,56462
Bonds 77 % 2,75 % 0,0359 0,7679
Portfolio 1 4,16 % 0,0388 1,07224

1994-1999

Asset Weight Excess return Standard deviation Sharpe

REITs 7% 2,23 % 0,1197 0,18673
Stocks 53 % 16,66 % 0,1403 1,18683
Bonds 40 % 1,08 % 0,0395 0,27347
Portfolio 1 9,42 % 0,0772 1,22064

2000-2005

Asset Weight Excess return Standard deviation Sharpe

REITs 25 % 16,67 % 0,1402 1,18868
Stocks 0% -2,03 % 0,1598 -0,12725
Bonds 75 % 3,55 % 0,0377 0,94298
Portfolio 1 6,83 % 0,0451 1,51392
2006-2010
Asset Weight Excess return Standard deviation Sharpe
REITs 2% 5,05 % 0,3075 0,16422
Stocks 3% 2,49 % 0,1816 0,13688
Bonds 95 % 3,29 % 0,0383 0,85752
Portfolio 1 3,30 % 0,0375 0,87899
2011-2015
Asset Weight Excess return Standard deviation Sharpe
REITs 10 % 12,03 % 0,1457 0,82555
Stocks 16 % 12,12 % 0,1213 0,99895
Bonds 73 % 3,05 % 0,0289 1,05671
Portfolio 1 5,46 % 0,0333 1,6401
2016-2020
Asset Weight Excess return Standard deviation Sharpe
REITs 14 % 9,17 % 0,1261 0,72706
Stocks 22 % 12,69 % 0,1233 1,02945
Bonds 64 % 2,42 % 0,0320 0,75605
Portfolio 1 5,65 % 0,0391 1,44606

Last 5 years

Sharpe
Asset Weight Excess return Standard deviation
ratio
REITs 14 % 8,11 % 0,1257 0,65
Stocks 20 % 9,95 % 0,1232 0,81
Bonds 66 % 2,34 % 0,0328 0,71
Portfolio 100 % 4,67 % 0,0377 1,24

Last 20 years

Sharpe
Asset Weight Excess return Standard deviation
ratio
REITs 15 % 11,13 % 0,1935 0,58
Stocks 12 % 5,50 % 0,1501 0,37
Bonds 73 % 1,66 % 0,0347 0,48
Portfolio 100 % 3,56 % 0,0436 0,82

97
7.2 Portfolios of REIT sectors with various constraints, 2000-2020

7.3 Industry Classification

98
8 Bibliography
Ackoff, R. L. (1962). Scientific Method: Optimizing Applied Research Decisions. New
York: Wiley.
Adamczyk, A. (2019, September 29). Millennials are fleeing big cities for the
suburbs. Retrieved from CNBC.com:
https://www.cnbc.com/2019/09/29/millennials-are-fleeing-big-cities-for-
the-suburbs.html
Andonov, A., Eichholtz, P., & Kok, N. (2015). Intermediated investment management
in private markets: Evidence from pension fund investments in real estate.
Journal of Financial Markets, 73-103.
Banz, R. (1981). The relationship between return and market value of common
stocks. Journal of Financial Economics 9, pp. 3-18.
Benninga, S. (2014). Financial Modeling. The MIT Press.
Bitsch Olsen, P. (2003). Problemorienteret Projektarbeide, 3rd edition.
Frederiksberg: Roskilde Universitetsforlag.
Bjølgerud, B. (2020, April 13). CEO Pangea Property Partners. (M. Lorentzen,
Interviewer)
Blitz, D. (2018, September 15). Fama-French 5-factor model: why more is not
always better. Retrieved from Robeco:
https://www.robeco.com/en/insights/2015/10/fama-french-5-factor-model-
why-more-is-not-always-better.html
Board of Governors of the Federal Reserve System. (2020, 04 20). 10-Year Treasury
Constant Maturity Rate [GS10]. Retrieved from FRED, Federal Reserve Bank
of St. Louis: https://fred.stlouisfed.org/series/GS10
Cannon, S. E., & Cole, R. A. (2011). How Accurate Are Commercial Real Estate
Appraisals? Evidence from 25 Years of NCREIF Sales Data. Chicago.
Carhart, M. M. (1997). On Persistence in Mutual Fund Performance. The Journal of
Finance Vol. 52, pp. 57-82.
CBRE. (2020, April 1). CBRE Flash Call, COVID-19 impact on industrial and logistic
real estate. Retrieved from CBRE.com: https://www.cbre.no/-
/media/cbre/countryunitedstates/us-research/covid-19-flash-call/covid-19-
impact-on-industrial/04-02-2020-covid-19-industrial-flash-call-summary-
v3.pdf?la=en
Charles Schwab. (2020, March 13). Schwab market perspective: Coronavirus hits
market hard. Retrieved from Shwab.com:
https://www.schwab.com/resource-center/insights/content/market-
perspective
Cohen, J. (1992). A power primer. Psychological Bulletin, 112 (1), 155-159.

99
Damodaran, A. (1999). Estimating Risk Parameters. New York: NYU Stern School of
Business.
Damodaran, A. (2012). Investment Valuation: Tools and Techniques for
Determining the Value of Any Asset (3rd ed.). Wiley.
DeCambre, M. (2020, May 1). After the best April in 82 years, is "sell in May" in the
coronavirus era a smart strategy? Retrieved from MarketWatch.com:
https://www.marketwatch.com/story/after-a-blockbuster-april-for-the-dow-
and-sp-500-is-sell-in-may-in-the-coronavirus-era-a-smart-strategy-2020-04-
30
DNB Markets. (2020). Norway Update: Fast reopening and public spending could
save the Norwegian economy from long slump. Oslo: DNB.
Eiendom Norge. (2020, May 6). Forskjeller i boligprisutviklingen i april. Retrieved
from eiendomnorge.no: https://eiendomnorge.no/boligprisstatistikk/
European Public Real Estate Association . (2019). Does Listed Real Estate Behave
Like Direct Real Estate? Brussels: EPRA.
Fama, E. F., & French, K. R. (1992). Common risk factors in the returns on stocks
and bonds. Journal of Financial Economics, 3-56.
Fama, E. F., & Kenneth, F. (1993). Common risk factors in the returns on stocks and
bonds. Journal of Financial Economics 33, pp. 3-56.
Fama, E., & French, K. (1992, June). The Cross-Section of Expected Stock Returns.
The Journal of Finance Vol. 47. No.2, pp. 427-465.
Fama, E., & French, K. (2015). A Five-Factor Asset Pricing Model. Journal of
Financial Economics 116, pp. 1-22. Retrieved from
https://ssrn.com/abstract=2287202
Fama, F. E., & French, K. (1996, March). Multifactor Explanations of Asset Pricing
Anomalies. The Journal of Finance, pp. 55-84.
French, K. R. (2020). Kenneth R. French - Data Library. Retrieved from
MBA.Tuck.Darthmouth.edu:
http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html
Friedman, H. C. (1971). Real Estate Investment and Portfolio Theory. The Journal of
Financial and Quantitative Analysis, Vol. 6, No. 2, 861-874.
Gorman, J. (2020, April 16). CEO of Morgan Stanley Investment bank. (B.
Television, Interviewer)
Gujarl, V., Palter, R., Sanghvi, A., & Vickery, B. (2020, April 15). Commercial real
estate must do more than merely adapt to coronavirus. Retrieved from
McKinsey.com: https://www.mckinsey.com/industries/private-equity-and-
principal-investors/our-insights/commercial-real-estate-must-do-more-than-
merely-adapt-to-coronavirus#

100
Heale, R., & Twycross, A. (2015). Validity and reliability in quantitative studies.
Ontario: CrossMark.
Hentov, E., & Petrov, A. (2020). How Do Sovereign Wealth Funds Invest? Less and
Less Contrarian. State Street Global Advisors.
Hoesli, M., & Oikarinen, E. (2019). Does Listed Real Estate Behave Like Direct Real
Estate? Brussels: European Public Real Estate Association (ERPA).
Ibbotson, R. G., & Siegel, L. B. (1984). Real Estate Returns: A Comparison with Other
Investments. AREUEA Journal, Vol. 12, No. 3, 219-242.
Institutional Real Estate, Inc. (2018). Global Investment Managers 2018. San
Ramon, CA: IREI.
Jegadeesh, N., & Titman, S. (1993, March). Returns to buying winners and selling
losers: Implications for stock market efficiency. The Journal of Finance 48,
NO. 1, pp. 65-91.
Johns Hopkins University & Medicine. (2020, April 16). coronavirus.jhu.edu.
Retrieved from Johns Hopkins University:
https://coronavirus.jhu.edu/map.html
Klebnikov, S. (2020, April 10). JPMorgan Forecasts 20% unemployment and 40%
hit to second-quarter GDP. Retrieved from Forbes.com:
https://www.forbes.com/sites/sergeiklebnikov/2020/04/10/jpmorgan-
forecasts-20-unemployment-and-40-hit-to-second-quarter-
gdp/#1e1e8bb64489
Koller, T., Goedhart, M., & Wessels, D. (2010). Valuation: Measuring and Managing
the Value of Companies. McKinsey & Company.
Kozlowski, R. (2020, January 30). Sovereign wealth funds stay the course on asset
allocation. Retrieved from Pensions&Investments.com:
https://www.pionline.com/sovereign-wealth-funds/sovereign-wealth-funds-
stay-course-asset-allocation
Lachance, C., & Tibone, V. (2020, April 15). Is Coronavirus the Retail Armageddon?
Retrieved from Green Street Advisors:
https://www.greenstreetadvisors.com/insights/blog/is-coronavirus-the-
retail-armageddon
Lintner, J. (1965, Feb). The Valuation of Risk Assets and the Selection of Risky
Investments in Stock Portfolios and Capital Budgets. The Review of
Economics and Statistics, Vol. 47, No. 1, pp. 13-37.
Loader, D. (2016). Fund Custody and Administration. Academic Press.
Markowitz, H. (1952, Mar). Portfolio Selection. The Journal of Finance, Vol. 7, No. 1.,
pp. 77-91.
Mossin, J. (1966, Oct). Equilibrium in a Capital Asset Market. Econometrica Vol.34
No.4, pp. 768-783.

101
Munk, C. (2018). Financial Markets and Investments. Copenhagen Business School.
Nareit. (2020a). Constituent companies of the FTSE Nareit All REITs Index. Nareit.
Nareit. (2020b, March 05). reit.com. Retrieved from REITs by the numbers:
https://www.reit.com/data-research/data/reits-numbers
Nareit. (2020c, 03 05). Monthly index values and returns. Retrieved from reit.com:
https://www.reit.com/data-research/reit-indexes/monthly-index-values-
returns
Nareit, c. (2020, May 1). FTSE NAREIT ALL REITs . Retrieved from investing.com:
https://www.investing.com/indices/ftse-nareit-all-equity-reits-historical-
data?fbclid=IwAR0UHHxD_k48RXwEq2l70Y-
nEhgLw8aznuUH2I7xk08rchq4QDeap7iogmA
Nasdaq. (2019, September 17). NYSE may be bigger, but Nasdaq is growing faster.
Retrieved from nasdaq.com: https://www.nasdaq.com/articles/nyse-may-be-
bigger-but-nasdaq-is-growing-faster-2019-09-17
NBIM. (2020). Government Pension Fund Global -Annual report 2019. Oslo:
Norges Bank Investment Management.
Nieuwerburgh, S. V. (2019). Why are REITs Currently So Expensive? Real Estate
Economics V47, 18-65.
Novy-Marx. (2013). The other side of value: The gross profitability premium.
Journal of Financial Economics 108, pp. 1-28.
O'Dea, C., & Lowe, R. (2018). Top 100 Real Estate Investment Managers 2018. IPE
Real Assets.
Oliver, D. (2020, April 14). About 25% of Marriott hotels shuttered worldwide due
to coronavirus. Retrieved from USAtoday.com:
https://eu.usatoday.com/story/travel/hotels/2020/04/14/coronavirus-
marriott-hotels-closed-worldwide/2989610001/
P&I. (2020, March 19). P&I Research Center. Retrieved from P&I:
https://researchcenter.pionline.com/v3/rankings/money-
manager/profiles/34224/overview
Petersen, C., Plenborg, T., & Kinserdal, F. (2017). Financial Statement Analysis -
Valuation - Credit Analysis - Performance evaluation. Bergen:
Fagbokforlaget.
Phillips, D. T., Ravindran, A., & Solberg, J. (1976). Operations Research, Principles
and Practice. New York: Wiley.
PopulationPyramid.net. (2020, April 20). United States of America. Retrieved from
PopulationPyramid.net: https://www.populationpyramid.net/united-states-
of-america/2019/

102
REITnotes. (2020, 03 05). list of publicly-traded REITs in the United States.
Retrieved from reitnotes.com: https://www.reitnotes.com/list-of-public-
reits/country/United%20States
Rivera, S. (2020, April 3). Will coronavirus cause an urban exodus? Retrieved from
BostonMagazine:
https://www.bostonmagazine.com/property/2020/04/03/coronavirus-
boston-suburban-exodus/
ROBECO. (2018, 03 27). Fama-French 5-factor model: five major concerns.
Retrieved from Robeco:
https://www.robeco.com/en/insights/2018/03/concerns-regarding-the-new-
fama-french-5-factor-model.html
Roll, R. (1977, March). A critique of the asset pricing theory's tests Part I: On past
and potential testability of the theory. Journal of Financial Economics, pp.
129-176.
Rosenberg, B., Reid, K., & Lanstein, R. (1985). Persuasive evidence of market
inefficiency. journal of Portfolio Management 11, pp. 9-17.
Sando, C. E. (2012). ODIN om eiendom. Oslo: Odin Forvaltning AS.
SEC. (2011, 12 9). Real Estate Investment Trusts (REITs), Investor.gov. Retrieved
from Investor.gov: https://www.investor.gov/additional-resources/news-
alerts/alerts-bulletins/real-estate-investment-trusts-reits
Selko, A. (2016, March 11). E-commerce drives the demand for logistics real estate.
Retrieved from Material Handling and Logistics:
https://www.mhlnews.com/global-supply-
chain/article/22051522/ecommerce-drives-the-demand-for-logistics-real-
estate
Sharpe, W. (1964, Sep). Capital Asset Prices: A Theory of Market Equilibrium under
Conditions of Risk. The Journal of Finance Vol.19 No.3, pp. 425-442.
Sharpe, W. F. (1966, Jan). Mutual Fund Performance. The Journal of Business, Vol.
39, No. 1, Part 2: Supplement on Security Prices, pp. 119-138.
State Street Global Advisors. (2020). SPDR Bloomberg Barclays U.S. Aggregate
Bond UCITS ETF (Dist). State Street.
Statista.com. (2019, Oct 29). Market cap of the REITs in the U.S 1975-2018.
Retrieved from statista.com:
https://www.statista.com/statistics/916665/market-cap-reits-usa/
Stattman, D. (1980). Book values and stock returns. The Chicago MBA: A Journal of
Selected Papers 4, pp. 25-45.
STR & Tourism Economics. (2020). U.S. hotel RevPAR forecasted to drop 50.6% for
2020. Hendersonville: STR.
STR. (2020). U.S. hotel results for the week ending 11 April. Hendersonville: STR.

103
STR. (2020, May 6). U.S. hotel results for the week ending 2 May. Retrieved from
str.com: https://str.com/press-release/str-us-hotel-results-week-ending-2-
may
Syz, J. (2008). Property Derivatives - Pricing, Hedging and Applications.
Chichester: John Wiley & Sons Ltd.
Titman, Wei, & Xie. (2004). Capital investments and stock returns. Journal of
Financial and Quantitative Analysis 39, pp. 677-700.
V. Gujarl et al. (2020).
Van Nieuwerburgh, S. (2019). Why are REITs Currently So Expensive? Real Estate
Economics V47, 18-65.
Wheaton, W. (2017). economics.mit.edu. Retrieved from mit.edu:
https://economics.mit.edu/files/14673
Yahoo Finance. (2020, 03 03). Yahoo finance, AGG. Retrieved from Yahoo finance:
https://finance.yahoo.com/quote/AGG/?guccounter=1&guce_referrer=aHR0
cHM6Ly93d3cuZ29vZ2xlLmNvbS8&guce_referrer_sig=AQAAAJLnzllxIHrID
ZYyNBZxCaNIHikKH87h7c6YHfMYk3E3x85unw7x-
jjvMs4ibyeX2p1cC6qtRy9g4VVTYgtBe66-
CJ7LjrhYLiNO5hdF6QOX84vkvkdqynONsVfVrajvCzFsnktLN
Yahoo Finance. (2020b, February 21). Apple Inc. (AAPL). Retrieved from yahoo
finance: https://finance.yahoo.com/quote/AAPL?p=AAPL&.tsrc=fin-srch
Yahoo Finance. (2020c, February 10). Treasury Yield 10 Years (^TNX). Retrieved
from finance.yahoo.com:
https://finance.yahoo.com/quote/%5ETNX?p=%5ETNX
Yahoo Finance. (2020d, May 11). Boston Properties stock information. Retrieved
from finance.yahoo.com: https://finance.yahoo.com/quote/BXP/
Zillow. (2020, April 20). Zillow research data. Retrieved from Zillow.com:
https://www.zillow.com/research/data/

104