Diversification benefits of SA REITs in a mixed asset 

portfolio: one decade and a pandemic later 

 

A Thesis/Dissertation  

presented to 

Faculty of Commerce, Law, and Management  

Wits Business School 

University Of the Witwatersrand 

In partial fulfilment for 

 the requirements for the Degree of 

 Master of Management in Finance and Investments 

By  

Masilo Mphaho 

2023 

Supervisor: Assoc. Prof. Odongo Kodongo 

 



1 
 

PLAGIARISM DECLARATION 
 

1. I know that plagiarism is wrong. Plagiarism is to use another’s work and pretend that it is 

one’s own. 

2. I have used the American Psychological Association (APA) (6th Edition) convention for 

citation and referencing. Each contribution to, and quotation in, this study from the work 

(s) of other people has been attributed and has been cited and referenced. 

3. This study on the Diversification benefits of SA REITs in a mixed asset portfolio: one 

decade and a pandemic later in South Africa is my own work. 

4. I have not allowed and will not allow anyone to copy my work with the intention of passing 

it off as his or her own work. 

5. I acknowledge that copying someone else’s assignment or essay, or part of it, is wrong, 

and declare that this is my own work. 

 

 

Masilo Mphaho 

 

 

 

 

 

 
 

 

 

 

 

 



2 
 

ACKNOWLEDGEMENTS 
 

I would like to take this opportunity to first thank God for giving me the strength to persist with 

this paper. My family and my partner for love, support, and always cheering me on, especially 

when I feel like giving up. A special word of gratitude to my Supervisor, Prof. Kodongo, a 

great leader who is always so patient when providing guidance, and Vukile Property Fund for 

funding my studies. I wish to further extend appreciation to the staff of WBS for all the 

assistance that got me this far. 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

  



3 
 

ABSTRACT 
 

Volatility spillover between financial markets causes inefficiency of diversification. Therefore, 

other investment alternatives are required to build an optimal portfolio, one of them being Real 

Estate Investment Trusts (REITs). The low correlation between REITs and stocks implies an 

advantage of diversification in an investment portfolio containing both assets. An important 

implication of this finding is that if stocks and REITs are incorporated into an investment 

portfolio, the investor will have better diversification benefits. This paper looks at the 

diversification benefits of having REITs in a mixed asset portfolio by conducting an empirical 

study from when the REIT regime came into effect in South Africa 10 years ago, particularly 

focusing on the period between 2013 and 2023. The econometric tools used in this regard 

include cointegration and, time series models (VAR and VECM) for forecasting. The paper 

also considers how the COVID-19 pandemic has affected this relationship by conducting a 

mean-variance spanning test to see if the inclusion of REITs in an existing portfolio dominates 

it. Other measures such as Sharpe ratios and Efficient Frontiers are included for analysing 

portfolio performance. Therefore, providing a mature analysis of REITs continuing from 

current literature and assisting Fund Managers in understanding the impact of including the 

asset class in a portfolio with a long-term investment horizon. This study affirms the low 

correlation between REITs and other stocks and further shows that they are not affected by 

shocks in the bond and stock markets respectively while also having the potential to improve 

the risk-adjusted returns of a Portfolio. Therefore, Fund Managers can consider REITs for their 

portfolio diversification strategies.  

 

 

 

 

 

 



4 
 

 

TABLE OF CONTENTS 
PLAGIARISM DECLARATION .................................................................................................................... 1 

ACKNOWLEDGEMENTS ........................................................................................................................... 2 

ABSTRACT ................................................................................................................................................ 3 

LIST OF TABLES ........................................................................................................................................ 6 

LIST OF FIGURES ...................................................................................................................................... 7 

LIST OF ABBREVIATIONS ......................................................................................................................... 8 

1. CHAPTER ONE - INTRODUCTION ..................................................................................................... 9 

1.1 BACKGROUND ............................................................................................................................. 9 

1.2 CONTEXT OF THE STUDY ........................................................................................................... 11 

1.3 PROBLEM STATEMENT .............................................................................................................. 14 

1.4 RESEARCH OBJECTIVES / RESEARCH QUESTIONS ..................................................................... 15 

1.5 BENEFIT OF THE STUDY (I.E., POLICY, INVESTORS ETC.) ........................................................... 16 

2. CHAPTER TWO - LITERATURE REVIEW .......................................................................................... 17 

2.1 WHAT IS A REAL ESTATE INVESTMENT TRUST? ........................................................................ 17 

2.2 DEBT COVENANTS ..................................................................................................................... 18 

2.2.1 Theoretical Literature ................................................................................................... 18 

2.2.2 Empirical Literature ....................................................................................................... 19 

2.3 DIVERSIFICATION BENEFITS ...................................................................................................... 19 

2.3.1 Theoretical Literature ................................................................................................... 19 

2.3.2 Empirical Literature ....................................................................................................... 20 

2.4 PORTFOLIO OPTIMIZATION ....................................................................................................... 22 

2.4.1 Theoretical Literature ................................................................................................... 22 

2.4.2 Empirical Literature ....................................................................................................... 22 

2.5 DIVIDENDS................................................................................................................................. 23 

2.5.1 Theoretical Literature ................................................................................................... 23 

2.5.2 Empirical Literature ...................................................................................................................... 25 

2.6 COINTEGRATION ....................................................................................................................... 26 

2.6.1 Theoretical Literature .................................................................................................................. 26 

2.6.2 Empirical Literature ...................................................................................................................... 26 

2.7 CONTRIBUTION TO THE LITERATURE ........................................................................................ 27 

3 CHAPTER THREE - METHODOLOGY ............................................................................................... 28 

3.1 INTRODUCTION ......................................................................................................................... 28 

3.2 DESCRIPTIVE STATISTICS ........................................................................................................... 28 



5 
 

3.3 MEAN-VARIANCE SPANNING TEST ........................................................................................... 29 

3.4 COINTEGRATION ....................................................................................................................... 31 

3.4.1 Unit Root Test ............................................................................................................... 31 

3.4.2 Testing for Co-integration ............................................................................................. 31 

3.4.3 Explanation of Results ................................................................................................... 32 

3.4.4 Error correction model (ECM) ....................................................................................... 32 

3.4.5 Forecasting and Analysis ............................................................................................... 32 

3.5 DATA AND DATA SOURCES ....................................................................................................... 32 

3.6 PERIOD OF STUDY ..................................................................................................................... 33 

4 CHAPTER FOUR - DATA ANALYSIS ................................................................................................. 34 

4.1 DATA ......................................................................................................................................... 34 

4.2 DESCRIPTIVE STATISTICS ............................................................................................... 34 

4.3 CORRELATIONS.............................................................................................................. 36 

4.4 PORTFOLIO OPTIMIZATION ....................................................................................................... 38 

4.5 MEAN-VARIANCE SPANNING TEST ........................................................................................... 40 

4.5.1 Mean-variance Spanning Tests under normality. ......................................................... 40 

4.5.2 Mean-variance spanning tests under conditional heteroskedasticity. ......................... 41 

4.5.3 Sub-sample analysis ...................................................................................................... 41 

4.6 COINTEGRATION ....................................................................................................................... 42 

4.6.1 Unit Root Test ............................................................................................................... 42 

4.6.2 Lag Length Selection ..................................................................................................... 44 

4.6.3 Testing for Co-integration ............................................................................................. 46 

4.6.4 Vector Autoregression .................................................................................................. 46 

5 CHAPTER FIVE - CONCLUSION ....................................................................................................... 51 

5.1 INTRODUCTION ......................................................................................................................... 51 

5.2 SUMMARY OF FINDINGS ........................................................................................................... 51 

5.3 CONCLUSION ............................................................................................................................. 52 

5.4 IMPLICATIONS ON STAKEHOLDERS ........................................................................................... 52 

5.5 RECOMMENDATIONS FOR FURTHER RESEARCH ...................................................................... 52 

 

 

  



6 
 

LIST OF TABLES 
 

Table 1 Index Correlation ........................................................................................................ 12 
Table 2 REIT Payout Ratios .................................................................................................... 25 
Table 3 Outline of data used. ................................................................................................... 34 
Table 4 Descriptive Statistics for March 2013 to March 2023 ............................................... 35 
Table 5 Descriptive Statistics during national state of disaster between the 27th of March 2020 

to the 8th of April 2022. ........................................................................................................... 35 
Table 6 Correlation Matrix between March 2013 to March 2023 .......................................... 37 
Table 7 Correlation Matrix during national state of disaster between 27th of March 2021 to 

the 8th of April 2022 ................................................................................................................ 37 
Table 8 Correlation matrix for indexes from 2014 to 2023 ..................................................... 38 
Table 9 Correlation matrix for indexes during national state of disaster ............................... 38 
Table 10 Sharpe ratios of the different portfolios during COVID and the full sample period 38 
Table 11 Normality tests for Portfolio 1 and 2 ........................................................................ 41 
Table 12 GMM of Portfolio 1 and 3 during full sample period .............................................. 41 
Table 13 GMM of Portfolio 1 and Portfolio 2 during Covid19 period ................................... 42 
Table 14 All Share Index Unit Root Test ................................................................................. 43 
Table 15 REIT Index Unit Root Test ........................................................................................ 43 
Table 16 Government Bond Index Unit Root Test ................................................................... 44 
Table 17 Lag Length Selection ................................................................................................ 44 
Table 18 Trend and Intercept Test ........................................................................................... 45 
Table 19 Augmented Dickey Fuller test for Error Term.......................................................... 45 
Table 20 Autoregressive Distributed Lag Model ..................................................................... 46 
Table 21 Vector Autoregression Lag Order Selection ............................................................. 47 
Table 22 VAR Residual Serial Correlation LM Tests .............................................................. 47 
Table 23 Roots of Characteristic Polynomial .......................................................................... 47 
Table 24 Vector Autoregression Estimates .............................................................................. 47 
Table 25 VAR Granger Causality ............................................................................................ 48 
Table 26 Forecast error variance decompositions .................................................................. 49 
 

 

 

 

 

 

 

 

 

 

  



7 
 

LIST OF FIGURES 
Figure 1 Cumulative Asset Class Return ................................................................................. 14 
Figure 2 Correlation of REITs with Direct Investments .......................................................... 23 
Figure 3 Portfolio 1 Opportunity Set………………………………………………………………..43       
Figure 4 Portfolio 1 Opportunity Set  during Covid-19 .......................................................... 39 
Figure 5 Portfolio 2 Opportunity Set………………………………………………………………..44                 
Figure 6 Portfolio 2 Opportunity Set during Covid-19 ........................................................... 39 
Figure 7 All Share Index Graph .............................................................................................. 42 
Figure 8 REIT Index Graph ..................................................................................................... 43 
Figure 9 Bond Index Graph ..................................................................................................... 44 
Figure 10 Impulse Response Functions ................................................................................... 50 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 



8 
 

LIST OF ABBREVIATIONS 
 

AIC Akaike Information Criterion 

ALBI All-Bond Index 

ARDL Autoregressive distributed lag 

CAL Capital allocation line 

CAPM Capital asset pricing model 

CPI Consumer Price Index 

DDM Dividend Discount Model 

ECM  Error Correction Model 

EREIT Equity Real Estate Investment Trust 

ETF Exchange Traded Fund 

FCF Free Cash Flows 

FPE Final Prediction Error Criterion 

FVA Fair Value Adjustment 

GDP Gross Domestic Product 

GMMs Generalised Method of Moments 

HQ Hannan-Quinn Criterion 

ICR Interest Coverage Ratio 

JB Jarque–Bera test 

JSE Johannesburg Stock Exchange 

JSEALSH Johannesburg Stock Exchange All Share 

LM Lagrange multiplier 

LTV Loan to Value  

MPT Modern Portfolio Theory 

MREIT Mortgage Real Estate Investment Trust 

NAV Net Asset Value 

NPV Net Present Value 

OLS Ordinary Least Squares 

PLS Property Loan Stocks 

PUT Property Unit Trust 

REIT Real Estate Investment Trusts 

SARB South African Reserve Bank 

SARS South African Revenue Service 

SC Schwarz Criterion 

STATSA Statistics South Africa 

VAR Vector Autoregression 

VECM Vector Error Correction Model 



9 
 

  

1.  CHAPTER ONE - INTRODUCTION  
 

1.1 BACKGROUND 
 

The willingness to participate in the Real Estate market is a result of many different factors, 

from those who want to improve South Africans’ social quality of living to others who 

recognize the potential of this asset class in adding value to their investment portfolio. While 

access to the property market has improved over time, with more previously marginalized 

groups having better access to credit in the form of a mortgage loan/bond since participating in 

the economy, the possibility of entry to larger investment properties is unfortunately limited 

(Massyn, 2015). Only those with an existing financial capital base of reason or access to 

intermediaries who are comfortable lending such large sums of financial capital can participate 

in the commercial property industry. With the average maximum Loan to Value ratio of 80%, 

the monetary value of the equity portion is not readily available to much of the less fortunate 

population (Chatterjee et al., 2022). For this research, we use ‘Real Estate’ and ‘Property’ 

interchangeably. 

 

Such interest in the asset class, combined with a lack of affordability, has seen many property 

holders utilize the opportunity to go public and provide this offering to a wider group through 

the financial markets and other platforms such as crowdfunding (Montgomery,2018). The 

expected growth in emerging markets has opened an opportunity for domestic and, more 

especially, foreign investors. Furthermore, the potential diversification benefit is a widespread 

notion that has led many researchers to investigate if this is truly an add-on or if it could just 

be an assumption as to why many institutional investors intend to have the vehicle within their 

mixed asset portfolio to reap the possible diversification benefits attributed to property 

investment (Feng, 2009). 

 

To understand Real Estate performance, you can distinguish between the asset classes split into 

four major sub-classes: office, industrial, retail, and residential, or consider the sector 

holistically. The short-term movements of these sub-classes are driven by their respective 

drivers, which include inter alia, population, employment, and real income per capita, some of 

which overlap between the sub-classes. 

 



10 
 

South Africa’s introduction of the Real Estate Investment Structure (REIT) in 2013, according 

to Ntuli et al. (2017), had the intention of encouraging more local and international investment. 

With the conversion from property companies to REITs, many entities started showing 

qualities of listed equity and property-backed assets. With the former introducing more market 

volatility, this raises the question of whether REIT performance should continue to be classified 

in the same category as other property companies and how large a gap could be realized. This 

would affect the way investors value these assets, whether putting more weight on the 

underlying asset through property appraisal prices or capital market information. The first 

method which is more linked to the economic theory of hedonic prices (Lisi, 2022) introduces 

what is known as the property appraisal evaluation risk, whilst the latter introduces the risk of 

investor sentiment that is priced in asset prices, which could lead to mispricing.  

The background of REITs in South Africa, as explained by Klerk (2019), originates from back 

in 2006 when the local listed property sector was facing specific tax challenges and the various 

sector representatives began engaging with the South African National Treasury to promote 

the concept of a REIT, which was a fast-growing trend over the rest of the world. For South 

Africa to keep up, they had to implement a similar vehicle tailored to the local investment 

culture (Ntuli, 2020). Treasury formally published the REIT tax legislation for South Africa on 

the 25th of October 2012 in the 2012 Taxation Laws Amendment Bill. This took the form of a 

new section (section 25BB) in the Income Tax Act, which Klerk (2019) refers to bringing South 

Africa’s publicly traded property sector in line with international standards, with the JSE 

publishing new listing requirements on 28th March 2013 facilitating the SA REIT. 

 

The following was set as the guide on what structure these entities should retain: minimum 

share capital of R300 million, total loan-to-value ratio of 60 percent with the source of income 

being 75 percent from the rental and direct/indirect income from property owned, and a 

minimum distribution of 75 percent of distributable income to Shareholders (De Klerk, 2013). 

 

The various attributes of a REIT that aim to, in summary, simplify and maximize real estate 

investment and returns might have resultant effects on the performance of these instruments 

with different economic cycles. Newell, et al. (2013), Newell et al. (2015); Newell et al. (2016) 

being one of the few that distinguish the asset class, find that, in established REIT markets, the 

risk-adjusted performance of REITs tends to be greater than that of stocks and listed property 

and that in these markets, the risk-adjusted performance of REITs tends to be greater than that 

of stocks and listed property.  



11 
 

 

Since property companies expose themselves to more criticism when converting to REITS, 

there ought to be a great enough opportunity cost for them to pursue that conversion. According 

to Wang (2022), the decision may depend on each country’s regulatory restrictiveness, market 

sentiment, the actions of peers, and company-specific factors. This includes taxation and how 

listed property in each country is affected by taxation. Gyourko et al. (1999) investigated this 

and they found that tax savings played a significant role in this regard. Furthermore, REIT 

regimes are different in each country; where some countries are lenient on capital gains tax 

some REIT regimes require taxation of unrealized capital gains in the property portfolio, and 

this could demotivate property companies showing interest in acquiring REIT status. 

 

That said, there is a distinction to be acknowledged between this and other asset classes from 

a performance and portfolio formation point of view. For diversification benefits, this is 

primarily because REITs provide a ‘real asset’ hedge to systematic risk. Unlike most shares on 

the stock market, REITs, like commodity and other infrastructure companies, are assumed a 

balance sheet, which is reinforced by their ability to hold decent value during periods of 

economic slowdown however this notion, according to Chyi (2012), is inaccurate. Chang 

(2017) suggests that this asset class can contain a hedge against inflation, which allows 

investors to protect their portfolios from the harsh impacts of inflation on earnings and capital 

growth. 

 

Another factor to consider is the accretive nature of dividends provided by REITS. Since REITs 

are required to distribute 75% of their distributable income, most of these funds compete to 

achieve a lot more than this requirement such that investors (usually income-focused) weigh 

their dividend-earning capabilities of the pool of entities. Most REITs on the Johannesburg 

Stock Exchange (JSE) make it a point to issue more than 90% of their earnings to investors 

with further consistent growth guidance in dividends per share. The race to pay dividends, a 

peculiar quality across asset classes, allows REITs to stand out.  

 

1.2 CONTEXT OF THE STUDY 
 

Recently, there have been peculiar movements across the financial markets globally and 

domestically. According to a report by Anchor Stockbrokers, a local brokerage firm, on indexes 

cumulative asset class returns to May 2023, the property sector has somewhat gone against the 

recovery trend apparent in common equity and bonds, showing a slowdown for most of the 



12 
 

period. This comes as landlords are faced with rising interest rates and increased operating 

costs due to load shedding and municipal expenses (Kelly, 2018). Coupled with business plight 

as consumers are struggling to keep up with the higher costs of living, this results in a ‘double 

whammy’ on landlords who are often approached with rent freezes and, in some cases, negative 

rental reversions. This is particularly worse for portfolios containing rental office properties 

that currently face all-time low occupancy rates, resulting in less income and lower valuations 

from fair value adjustments. 

As a result, the relationship between the performance of REITs has shown a negative 

correlation, as shown in Table 1 below, supporting what has been described above. This is 

based on a pre-study of this research using data from Yahoo Finance based on weekly return 

data, collected over ten years.  

 

Table 1 Index Correlation of REITS vs All Share and Government Bonds 

                              vs   

J805TR INDEX JALSH INDEX -0,312 

J805TR INDEX IGOV TR INDEX -0,370 

 

Where the J805TR index is the JSE REIT index, the JSEALSH index is the JSE all share index, 

and IGOV TR INDEX is government bonds, the REIT index has exhibited low negative 

correlation when paired with the All share and the Government bond indexes, respectively. 

 

The following are factors affecting the performance of the Real Estate sector recently: 

Electricity load shedding worsened in 2019, resulting in disrupted trade for tenants in the retail 

and industrial subclasses, followed by corporate companies adopting the remote working 

practice for their employees practiced by many corporates in 2020, which led to many landlords 

experiencing significant numbers of vacant offices and subsequent loss of income. Coupled 

with municipal expenses containing double CPI increases, according to Kelly, (2018), the 2021 

social unrest in the KZN region resulted in many commercial property owners either losing 

their properties or having to at least increase their guarding expenses with no foreseeable return 

but as a defensive measure (Elumalai,2022). Furthermore, the Russia/Ukraine war affecting 

global food security and prices resulted in reduced consumer discretionary spend, followed by 

the 2022 severe floods and landslides in KZN, Eastern Cape, and recently (2023 Q3) Western 

Cape, which caused damage to properties and government infrastructure like roads and sewer 



13 
 

networks at the same time when most REITs face interest rate increases affecting cost of 

finance. 

 

All the events mentioned above have an impact on most REIT portfolios of different nature 

and size. The interest rate spikes, municipal expenses, and electricity load shedding (spending 

on backup solutions and indirectly through tenant trade downtime) directly reflect on the 

financials. In their most recent results, the two largest retailers stress how generator running 

costs (diesel and maintenance) have impacted their financials, with one reporting the first-ever 

loss since listing (Child,2023). The resultant effect is that tenants are becoming stricter on 

rentals to reduce the total cost of occupation. The remainder of the issues also eventually affect 

performance and maybe more severely, like the disruption caused by RIOTs placing R7 billion 

worth of property at risk in eThekwini and floods removing R25 billion from the economy 

(South African Institute of Valuers,2022). 

  

Other financial risks of a systematic nature, like interest rate hikes, which increased from 3.75% 

in 2021 to 8.25% in 2023 (from levels of 6.25% before the Covid-19 pandemic), appear to 

show a greater impact on Real Estate, which accord with Anderson (2022) on the effect of 

interest rates on REITs valuation and future return. This is mainly due to the industry levels of 

leverage (loan-to-value ratios) and the magnitude (size of capital) thereof. The sector relies 

greatly on external capital, which mostly comes from credit providers such as banks (Feng, 

2007), and where such facilities have not been hedged using swap agreements, the interest rate 

risks and movements that take place during monetary policy tightening reduce the amount of 

cash available to satisfy lender repayments. This and other issues like financial covenants affect 

confidence in the sector, which increases the cost of capital in this alternative market and 

impedes growth even further by so doing (Borgonovo, 2013). The graph that follows shows 

the slow recovery compared with other asset classes on a cumulative return basis: 

 

What is apparent is that in May 2023, listed property lagged far from a cumulative asset class 

return perspective compared to common stock and bonds. The three financial instruments have 

had periods where they move collectively; however, from November 2022, listed properties 

(which mostly consist of REITs) have started bucking that trend. To better understand this, 

sub-samples need to be extracted for these cycles and matched with the respective business 

cycle or major economic activities.  



14 
 

 

Figure 1 Cumulative Asset Class Return of Property, Equity and Bonds over the period of August 2021 
to May 2023 

 

Source: Anchor Stockbrokers 

In measuring the relevance of REITs in current mixed-asset portfolios, this paper investigates 

their performance using different asset types as a reference and further taking into consideration 

other factors which are prudent when analyzing the performance of the REITs by assessing the 

impact of the various REIT rules imposed on these companies. To this effect, we also consider 

how the distribution requirements affect the performance in tight economic conditions and how 

investors perceive this as a value add by looking into the movements in the stock price during 

times when investors failed to deliver on this requirement, i.e., during Covid-19 when cash 

flow was tight and the preceding period interest rate hikes.  

 

1.3 PROBLEM STATEMENT 
 

The COVID-19 pandemic, coupled with the technical recession experienced by many countries 

globally before the closure of economies, has exacerbated challenges for companies across the 

board. For REITs, who have certain obligations to meet to retain their listing status and reap 

the benefits of having this status, this has been further challenging as their clients (mostly 

retailers) could not trade unless they provided essential services (Akinsomi, 2020).  

The matter becomes if REITs still hold a diversification benefit for investors and how they 

bode in a defensive portfolio, especially through major economic headwinds. What risk does 



15 
 

the latter bear on shareholders who had considered REITs as part of their investment strategy? 

Should these parties revisit the guidelines, and strategies given that it has been 10 ten years 

since they were implemented, and now there is a long enough period to test the relevance of 

the initial REIT structure and or/ business case and observe how it competes with pre-existing 

and newer financial instruments.  

Key literature like Carstens (2019), shows the adverse effects of this legislation on a REITs 

internal and external capital availability due to the lessened retention rate and increased 

distribution rate.  (Akinsomi, 2020)‘s early investigation into the impact of COVID-19 shows 

the importance of understanding REIT performance on an asset class level, as distinct events 

have different results on each.  

Share structures have also become a focal point with Fortress (REIT), now formally known as 

Fortress ‘Real Estate Investments’ Investments’, having lost its REIT status due to its inability 

to issue dividend payouts to its shareholders emerging out of the pandemic in 2022 when 

interest rates were starting to normalize. The company has since described this as an 

unfortunate event. However, they claim that they are now able to reduce their LTV and redirect 

the capital to income-enhancing developments and capital expenditure. Others like Hyprop,  

Attacq, and Texton also failed to issue dividends in the 2020 to 2022 period with the latter’s 

cash generation in 2020 being below 75% of its distributable income which is the required 

distribution to retain REIT status. Could the REIT diversification benefit have been a fallacy 

and is the risk of other listed companies with REIT status is a broader problem that this paper 

seeks to solve adding COVID-19 as a test. Although other Authors have completed earlier 

studies on South African REITs during COVID-19 this paper also offers a 10-year examination 

which is currently found in a global context therefore intending to assist local investors. 

 

1.4 RESEARCH OBJECTIVES / RESEARCH QUESTIONS 
 

The overall objective of this study is to examine REIT investment performance when included 

in a mixed portfolio of fixed-income instruments (bonds) and common stock, continuing from 

previous studies that consider the benefits of having a REIT in a mixed asset portfolio for 

diversification purposes. 

Specifically, the study seeks to:  

1) Study REIT diversification benefits for the ten years since introduction and, 



16 
 

2) Further to the above, due to the sensitivity of financial markets to economic cycles, this 

paper will perform tests on how the REIT performance was affected by the COVID-19 

pandemic as GDP growth was under strain. This study examines the safe haven 

characteristics of this asset class as their ability to hedge against prevalent economic 

downturns, placing additional focus on the COVID-19 pandemic. 

 

1.5 BENEFIT OF THE STUDY (I.E., POLICY, INVESTORS ETC.) 
 

This paper will firstly assist investors in understanding more of the impact/strength of South 

African REITs and if they would be a value add to their portfolios, as well as policymakers in 

ensuring that the structure remains relevant and beneficial to these investors in the fast-

changing environment. As the worst of the pandemic now seems to be over, this paper aims to 

continue where previous researchers left off and draw attention to how REITs behave during 

times of uncertainty and extremely weak economic headwinds. 

 

 

 

 

 

 

 

 

 

 

 

 

 

 



17 
 

2. CHAPTER TWO - LITERATURE REVIEW 
 

2.1 WHAT IS A REAL ESTATE INVESTMENT TRUST? 
 

The property sector is well known for its long-term growth and reliable returns reinforced by 

the underlying brick-and-mortar and land value, which tends to appreciate over time 

(Omokhomion, 2018). Introducing this asset class to the financial markets has sparked much 

interest from institutional investors, inter alia: life insurance companies, pension funds, and 

mutual, hedge, and superannuation funds (Andonov et al., 2015).  

An investment vehicle that later became known as a Real Estate Investment Trust (REIT) was 

established to expand the offering and make the properties more attractive to investors. This 

asset trades as an alternative investment on the stock exchange, like ETFs and other non-

conventional asset classes, which the market perceives.  (Robinson, 2018) describes a REIT as 

an entity that owns income-generating assets that offer investors an opportunity to gain access 

to the capital-intensive world of property.  

The introduction of this asset to the capital markets has resulted in supernatural growth of the 

sector and, consequently, its contribution to economic growth; as statistics have it, the sector 

has grown to make up a substantial portion of global gross domestic product with assets of 

about $2 trillion (Brown, 2019). This begs the question of how much we understand of this 

asset class - from the underlying asset to other corporate finance matters like capital funding 

strategies and, on the other end, the potential risk it poses to the world economy as seen in the 

housing market-induced global economic crisis of 07/08 (Bhuyan, 2015). 

This evolution of REITs has undergone various changes, progressively arriving at the most 

efficient structure. Some studies, such as Nareit (2010), deal with when the first REIT Act was 

passed by Mr. Eisenhower, former president of the United States, which was contained in the 

Cigar Excise Tax Extension of 1960. There are also different types of REITs (Equity REITs, 

Mortgage REITs, and Hybrid REITs), and each can be further broken down into subcategories 

(S&P Dow Jones Indices, 2023).  

A REIT, however, is simply a property asset structure that allows the Founders/Managers 

growth by allowing them to pass on the tax obligation to the investor provided that certain 

requirements set by the REIT legislature are met. These rules include a minimum distribution 

of, i.e., at least 70% and 75% in Nigeria and SA, respectively, while in the UK, at least 90% 



18 
 

(as region-specific requirements) of its distributable earnings to shareholders. REITs are 

pushed to disperse profits from their commercial real estate portfolios to take advantage of tax 

breaks for shareholders, which earn relatively small net income margins (Omokhomion, 2018). 

According to Friday (1999), REITs are closed-end investment companies providing investors 

a passive channel to invest in income-producing real estate. A REIT's investing goal is to 

produce dividend income for investors, typically through rental revenue and gearing from 

capital gains on real estate assets. 

Haslama (2015) says there is a significant amount of value at stake if predictions about future 

values turn out to be overly optimistic because Fair Value Adjustments (FVA) modify asset 

values in the present based on future financial expectations about earnings that have not yet 

been realized. If asset values are reduced, these changes must be covered by shareholder funds, 

which might or might not be enough to reduce financial instability and avert insolvency. 

Haslama (2015) further, finds that because of the favorable development potential and financial 

returns that FVA and legal changes combined to offer to real estate investors in 2007, 

represented a fresh chance for the REIT business model. However, in 2008, during the financial 

crisis, when the value of commercial property fell by nearly 50%, the climate in which this 

business model operated was badly damaged. The stock market value of this group of 

companies fell in line with the decrease in real estate values in 2007-2008. 

 

2.2 DEBT COVENANTS 
 

2.2.1 Theoretical Literature 

Earlier research has also recommended debt covenants to curb Agency Problems (Jensen et al., 

1976) Smith et al., 1979). One of the common attributes of REITs is the covenants placed by 

debt financiers when providing a facility to a REIT for a property. These create an 

accountability tool for managers when undertaking projects and protect debtholders from 

value-reducing projects. These financial covenants normally include limits on leverage, 

prescribed fixed charges, and minimum interest coverage ratios, also known as ICRs, and are 

mostly more prevalent in investment-grade REITs (Tsang, 2016)  

It should be noted that the distribution requirements, which limit ploughing back, also prohibit 

management’s expropriation and require the agent to make an effective investment decision 

that provides long-term benefits to shareholders. Omokhomion1, (2018), Berle et al. (1932), 



19 
 

and Hartzell et al. (2003) indicate that having a large family/founder concentration limits this 

Agency problem with inside ownership positively correlated with the REITs TobinQ. 

2.2.2 Empirical Literature 

Tsang (2016) demonstrates that covenants only reduce the cost of debt for investment-grade 

REITs that have a higher potential for agency conflicts and liquidity risk and even though the 

covenant requirements are probably non-binding, investment-grade REITs can gain from using 

covenant clauses as a sign of the quality of their debt when they offer loans. Their results 

demonstrate that investment-grade REITs may still have agency conflicts between shareholders 

and debt holders. However, they may be lessened by covenant provisions, which would lower 

the cost of debt. This is also supported by Riddiough (2020), who found agency conflicts by 

documenting how they help manage the conflict between managers and shareholders over spare 

debt capacity. 

 

2.3 DIVERSIFICATION BENEFITS  
 

2.3.1 Theoretical Literature 

Diversification is an important concept in portfolio formation as it involves spreading 

investments across different asset classes, sectors, geographies, and securities to reduce risk 

and potentially enhance returns (Biswas, 2015). The idea behind diversification is that different 

investments behave differently under different market conditions. By diversifying a portfolio, 

investors can reduce the impact of negative events affecting any one particular investment, 

asset class, or sector on their overall portfolio. This means that if one investment performs 

poorly, it may be offset by the positive performance of another investment in the portfolio. 

Diversification also helps manage risk by ensuring that the portfolio is not overly exposed to 

any security or asset class (Koumou, 2020). This can help investors avoid the risks associated 

with concentrated positions, such as any risk of a company-specific event that negatively 

impacts the stock price. 

Real estate is an investment that shows a specific risk/return characteristic and low co-

movements with investments in the domestic bond and stock markets (Maurer et al., 2002). 

This implies that including real estate in a portfolio of stocks and bonds presumably enhances 

the risk/return profile from the perspective of a domestic investor (Maurer et al., 2002). On the 

other hand, according to Chandrashekaran (1999), REITs do appear to offer substantial 

diversification benefits, at least during certain periods, and therefore, dynamic asset allocation 



20 
 

strategies that invest in REITs are likely to achieve superior risk and return profiles. Since the 

overall goal of asset allocation is to realize the best possible level of return at the lowest level 

of risk, real estate asset allocation, thus, aims to achieve diversification among different classes 

of property assets combined in an investment portfolio (Darst et al., 2008) and  (Bhuyan, 2015).   

2.3.2 Empirical Literature  

Glascock (2018) looks at the difference between REITs and Listed Property companies by 

using diversification properties (or benefits) of holding both in one portfolio, which is more 

aligned with what this paper intends to explore. This allows mixed Fund Managers to weigh 

their opportunities of investing more in real estate and spreading their risk even further, as 

property investment itself is considered to provide diversification. Using various portfolio 

constructs (six in particular), the authors were able to identify the results of including or 

removing these assets from a portfolio.  

Starting by comparing the portfolio performance, which initially shows that REITs and public 

property companies provide diversification benefits if both play a part in the portfolio 

formation; however, a mixed asset portfolio that already contains REITs will not achieve an 

additional benefit by including listed property companies in their portfolio. To reduce the error 

that the market conditions could induce, Glascock (2018) continues to conduct the test 

throughout different economic cycles. With different property asset classes expected to 

perform differently depending on the cyclical environment, this also calls for a proper 

robustness test. The results were almost close to the initial test conducted. He finds that REITs 

and listed property companies contain diversification benefits from a mixed asset portfolio 

perspective but do not yield noteworthy gains when holding both stocks in one portfolio. Whilst 

listed property companies do provide more opportunities; the poor substitutability qualities of 

these assets should be considered by investors who wish to invest in real estate.  

 This is similar to Seiler’s journal investigation of the possibility of using ‘Equity REIT 

(EREITs) shares to rebalance/ diversify private real estate portfolios’ (Seiler, 2001).  Seiler 

used different real estate asset classes (retail, etc.) in all private and public real estate through 

re-allocating each asset class into a portfolio repeatedly, i.e., comparing a private real estate-

only portfolio comprising of apartments, industrial, office, and retail versus one without the 

private industrial and replacing it with public industrial (EREITs). The percentage allocations 

in each property type and the efficient frontiers were compared before and after. The more 

dissimilar the pre- and post-allocations and efficient sets are, the lower the degree of 

substitutability. This means that the two assets provide less diversification benefits. Seiler 



21 
 

found that in the US, the substitution of public real estate for private real estate in a private real 

estate-only portfolio varied to a considerable extent and would not result in in-efficient 

portfolio formations. However, efficient frontiers were almost the same. They proved that there 

can be a swap of a public real estate asset for a private one but with a similar efficiency level. 

The paper also goes on to evaluate the ability of short selling in a public real estate portfolio, 

but this method was proven to be fruitless in a sense. 

When looking at REITs' absorption to various ‘conventional and unconventional monetary 

policy shocks,’ Feng et al. (2022) note a level of heterogeneity between the various real estate 

asset classes (e.g., office, hospitality, etc.). However, the sector tends to respond similarly to 

these shifts. The paper uses the Inoue (2021) method of identifying monetary policy shocks 

according to shifts in the entire term structure of government bond yields, looking at 

unconventional periods such as the COVID-19 pandemic experienced in 2020-2021 and the 

07/08 Global Financial Crisis. 

Bossman (2022) deals with the impact of the COVID-19 pandemic by examining the physical 

aspects of lockdown restrictions on performance and the strain induced by the ‘heightened 

financial uncertainty’ due to the unstable market sentiment. They found that the pandemic has 

resulted in a heightened bearish market and suggested a more robust policy oversight to control 

market sentiment and protect REIT performance. The return from investment however not only 

controlled by market sentiment but also the response of an investor to the market according to  

Vallespin et al. (2021) who studied REITs performance pre and during the COVID-19 

pandemic. 

Hanif (2013) studies the diversification potential of REITs from 2016 to 2022 against the 

economic downturn of the bearish crude oil market and COVID-19 and tests for haven 

properties using the regression analysis of REIT returns vs Oil returns. This study found that 

during the pandemic, the Japanese REIT index amongst 13 other indexes (including the Global 

REITs index) proved to be the only one with hedging qualities with a beta equal to 0.060 at a 

level one significance. Fuss (2011), after finding a correlation between REITs and other asset 

classes, the Johansen co-integration test to assess the potential for diversification by including 

international REITs in the strategic asset allocation of a mixed-asset portfolio and finds the 

asset class to offer both active investors and those with long-term investment horizons ample 

diversification potential through the achievement of continuous cash flows. 



22 
 

BOVA11, which represents the Brazilian variable income market, and the SPDR S&P500, 

which represents real estate investment, were used by Penabad (2017) to measure and analyze 

the diversification by investing in REITs and equities for an investor from Brazil. Their 

findings prove that the introduction of REITs is optimal in a mixed asset portfolio by individual 

and institutional investors searching for assets with attractive potential for profitability and risk 

diversification (Penabad, 2017). 

2.4 PORTFOLIO OPTIMIZATION  
 

2.4.1 Theoretical Literature 

Mean-variance optimization, another name for portfolio optimization, is a quantitative 

technique used in finance to build investment portfolios to minimize risk for a given level of 

projected returns or maximize returns for a given level of risk  (Markowitz, 1952). Depending 

on the investor's risk tolerance and investment goals, the main principle underlying portfolio 

optimization is to arrange assets so that the final portfolio provides the optimum trade-off 

between risk and return. The Modern Portfolio Theory was created by the famous Harry 

Markowitz in the 1950s and aims to optimize the trade-off between risk and return in the 

construction of an investment portfolio.  (Markowitz, 1952) It is crucial to remember that 

portfolio optimization is predicated on assumptions and previous data, and future success is 

not guaranteed. Moreover, it assumes that market conditions remain constant, which is 

frequently not the case (Petter et al., 2014). 

2.4.2 Empirical Literature 

Cheng (2017) extends the classic Morden Portfolio Theory to analyze mixed-asset portfolios 

and explain the actual portfolio allocation to real estate without resorting to special data 

manipulations such as smoothing techniques and retaining basic MPT principles of greater 

returns at the least levels of risk. They found that the holding period T is irrelevant to optimal 

portfolio choice. Other papers also find the respective significance of French, Singapore and 

UK REITs in mixed-asset portfolios in the post-global financial crisis (GFC) environment 

(Newell et al., 2013) (Newell et al., 2014) (Newell et al., 2016). 

Another study by Hoesli (2019) finds that due to the significant transaction costs usually 

associated with the asset class, the allocation to direct real estate in a portfolio that seeks to 

maximize the Sharpe ratio and contains equities and bonds is zero for investment periods 

shorter than 2.5 years indicating less value add. However, even over the long term, REITs are 

proven to have the lowest correlation with direct investments, as shown in the figure below. 



23 
 

 Figure 2 Correlation of REITs with Stocks, Bonds, Direct Real Estate, Core Funds, Value Added Funds 
and Opportunistic Funds 

 

Source: (Hoesli, 2019) 

2.5 DIVIDENDS 
 

2.5.1 Theoretical Literature 

Dividend policy has been a topic of research, with many trying to establish the relevance and 

effect of its existence, such as the well-sought Corporate Finance paper of Miller et al. (1961). 

This indicates that, in efficient markets with fixed investment policies, all feasible capital 

structures and dividend policies are optimal because they produce the same shareholder wealth. 

So, the choice among them is irrelevant. A later paper by DeAngelo et al. (2005) titled, ‘The 

irrelevance of the MM dividend irrelevance theorem’ strongly differs from the findings of 

Miller and Modigliani based on the assumptions taken. They argue that this Theory ignores the 

assorted options that companies have in how much they decide to distribute (MM states that 

100% distribution is implemented in one way or the other) and retention or a pay-out of less 

than the full NPV of Free Cash Flows (FCF) affects a firm’s value. 

According to DeAngelo et al. (2005), the MM theory, including others that support the 

phenomenon that the dividend policy of an entity does not affect shareholder value, is quite 

flawed, and it would be/has been misleading to investors and researchers thereof to rely on this 

theory. In their words, “Failure to recognize that MM’s dividend irrelevance theorem does not 

apply to pay-out/retention decisions can cause serious mischief.” They further criticize the 

Black (1976) dividend puzzle, which emphasizes the tax implications of issuing dividends by 

reducing the real income earned by the shareholder, and firms should retain most of their 



24 
 

distributable earnings. This theorem will result in firms showing minimal to no value when 

using the Dividend Discount Model (DDM), which relies on the present value of all expected 

future dividends of an entity, consequently affecting a firm that seeks to raise capital. The 

counter to the Dividend puzzle is that in the normal cause of dividends being taxed, investors 

can select assets that are favorable to them.  

In conclusion, DeAngelo et al. (2005), whilst acknowledging the foundation set by MM in 

Morden corporate finance theory, proves that a pay-out ratio is most certainly relevant in a 

world where previous limitations are removed from retention. 

One of the characteristics of REITs mentioned above is the pay-out required ratio. Since REITs 

are required to pay out most of their distributable earnings to shareholders using dividends, the 

question becomes how elastic investors are to the distribution ratio above this threshold 

compared to other property companies that do not have a dividend requirement (Roberts et al., 

2023). 

To understand the impact of the pay-out decisions on shareholder sentiment, we can attempt to 

study what is known as the ex-dividend share price of the entity. Experts have attributed this 

to an exact offset for the dividend amount as a qualification for the drop in share price. 

Considering the tax effects of paying out dividends versus plowing that income into profitable 

projects and allowing the investors to delay the tax - Capital gains tax arises when you dispose 

of an asset for more than what you paid for it - then why do companies then prefer to pay-out 

as much as they can? This is the opening question by Wang (1993) in their paper titled: 

Dividend policies and dividend announcement effects for real estate investment trusts. Instead 

of retaining some of the earnings at no cost and channeling those funds into their investments, 

companies normally issue a maximum pay-out and do a capital raise (which entails a floatation 

cost that could have been averted). 

A general justification for high dividend pay-out ratios is the market signaling potential 

embedded in dividends and how this helps reduce the chances of agency costs Jensen (1986). 

This is exacerbated by the advisors used by entities who might have conflicting values. REITs 

also issue dividends at levels over 100% of their distributable earnings, according to Wang, 

(1993); this is made possible by accrued revenue and expenses (depending on the accounting 

methods used), depreciation method, and refinancing of their properties. 

Allowing these companies to go beyond the regulatory requirements and concluding that these 

factors allow firms to set a tone in the market on their performance. This was later confirmed 



25 
 

by Mooradian (2001), who found that listed property companies use more debt and pay less 

dividends than REIT entities. His study, which was based on hotel REITS vs non-REIT hotels, 

indicated that the former usually has free cash flow retention levels. 

2.5.2 Empirical Literature 

REITs in South Africa have become popular for their high dividend practices, well above the 

75% requirement leaning much closer to the one ratio to their distributable income before the 

pandemic. Carstens (2019) finds an increase in sustainable growth and profitability since the 

introduction of the SA REIT structure in 2013, which has enabled REITs to offer shareholders 

long-term stable and even increasing returns. Large players in the market have maintained a 

payout ratio of 95% even through tough economic conditions and elevated Operating Costs. 

According to Hill (2008), these excess dividends inform the capital markets that Management 

is aware that they will need more funding and that they are prepared to enhance the value of 

their shares to gain access to them in the future. 

Below is a sample of 4 large SA REITs and their payout ratios: 

Table 2 REIT Full Year Distributable Income Payout Ratios during the pandemic period 

Entity Payout Ratio 

Fairvest Ltd. 100% 

Investec Property Fund Ltd. 95% 

SA Corporate Real Estate Ltd.  90% 

Growth Point Properties Ltd. 82.5% 

Redefine Properties Ltd. 80% 

Source: 2022 Company Results 

As a result, REITs appear to have become more attractive to shareholders, making them more 

competitive and stimulating future shareholder investment, which in turn may increase future 

profitability and growth (Roberts et al., 2023). This study also confirms that leverage and profit 

retention diminished after the introduction of legislation, suggesting that the new legislation 

may limit capital availability, which is why most REITs regularly go through a period of asset 

disposals during their tenure. On the one hand, this has allowed this market to get a more 

reliable understanding of their NAV over and above the desktop valuations conducted and 

created space for new entrants. 



26 
 

In 2020, at the height of the COVID-19 pandemic, the local body of REITs in South Africa, 

formally known as SAREIT, requested the South African National Treasury to forego the 75% 

dividend payment obligation for two years ending 2022 (Akinsomi, 2020). This was because 

REITs were grappling with low turnovers, resulting in minimal or non-positive distributable 

earnings. 

2.6 COINTEGRATION 
 

2.6.1 Theoretical Literature 

Engle (1987) introduced the cointegration method, which employs error correction models 

instead of correlations to determine the long-horizon equilibrium between a group of non-

stationary variables. Cointegration approaches are used in well-known real estate research like 

Myer (1997) and Hansz (2017) to identify long-term relationships between real estate and 

alternative asset classes. It is important to study cointegration because it provides an 

understanding of long-term equilibrium correlations between the price series of two data sets, 

especially because long-term relationships are particularly important for long-term investors. 

Additionally, cointegration deals with the significant drawback of conventional current 

portfolio-based approaches to a certain extent (Glascock, 2018). This is addressed by the 

robustness of cointegration methods against correlation instability (Glascock, 2018). Crucially, 

cointegrated assets have similar risk profiles and move in unison, offering little to no benefits 

for diversification.  

2.6.2 Empirical Literature 

To build a connection between risk diversification and cointegration theory, Glascock (2018) 

expands theoretical frameworks by examining the implications of portfolios empirically and 

assessing the long- and short-horizon inter-relationships between REITs and other listed 

properties using Granger causality approaches and Johansen cointegrations. In his paper Hansz 

(2017) finds that in a cointegrative system, EREITs are regarded as the leading asset, MREITs 

as the subordinate asset, and vice versa, implying that although long-term cointegrated prices 

of MREITs and REITs are stable in the cointegrated vectors, MREITs are a subordinate asset 

in the system and EREITs are the leading asset due to the unilateral Granger-causality from 

EREIT returns making the latter superior in ranking for asset allocation (Hansz, 2017). 

Glascock (2018); however, it was established that cointegrated real estate assets would be 

superfluous as diversifiers and can thus be omitted from portfolios. 

  



27 
 

2.7 CONTRIBUTION TO THE LITERATURE 
 

This research paper considers the effects of recent business cycles on the performance of REITs 

and how they measure when combined with stocks and fixed-income securities. Employing 

methods such as co-integration as Hansz (2017) and Glascock (2018) but for a broader 

understanding of diversification benefits in a mixed asset portfolio instead of a real estate 

concentrated portfolio. 

Since there has been minimal study on REIT performance and diversification potential post-

pandemic, this study will add substance to studies like Akinsomi (2020) in understanding the 

effects of the tough economic conditions and, especially considering that this asset class was 

particularly affected by delayed dividend distributions as observed in the pandemic because 

this vehicle is popular for dividend distributions.  

In closing, Chapter 2 conceptualizes REIT diversification benefits in a mixed asset portfolio 

by first discussing the physical nature of the underlying asset which tends to retain value over 

time compared to other asset classes. When considering different papers produced like Feng et 

al., (2022) economic shock resistance can be considered together with the low correlation 

between REITs and others covered by Hoesli (2019) this increases the potential of 

diversification. In addition to this, the elevated payout nature of REITs studied by DeAngelo 

et al. (2005) makes them more attractive to investors and has an impact of portfolio growth 

(Roberts et al., 2023). As alluded to in the Research Problem, this paper will add to existing 

literature and further close the gap between global studies on REITs in mixed asset portfolios 

as covered by papers such as Newell, et al. (2016), Darst et al. (2008) and Bhuyan (2015) by 

considering REITs in South Africa.   

 

 

 

 

 

 

 

 

 



28 
 

3 CHAPTER THREE - METHODOLOGY  
 

3.1 INTRODUCTION 
 

This methodology allows for a structured approach to deal with the research problem. An 

overview of the data sources, the types of statistical methods, their relevance, period, and 

interpretation of results is detailed here. The chapter begins with a description of the research 

approach and strategy before detailing the research design, data collection, and instruments. 

Thereafter, the data and methodological limitations and assumptions are identified and 

discussed. 

3.2 DESCRIPTIVE STATISTICS 
 

Descriptive statistics are first conducted to assist in summarizing the key characteristics of 

individual assets within a portfolio, such as their returns and risks (Kulkarni, 2019). This 

summary allows investors to understand the properties of each asset before considering their 

inclusion in a diversified portfolio. Co-movements between REITs, common stock, and bonds 

are also considered to understand if any volatility spillovers exist between the asset classes. 

To compare the diversification benefits provided by REITs during the 10-year period and the 

pandemic respectively, four multi-class portfolios are created as follows:  

1. Portfolio (I), consisting of bonds and common stock stocks,  

2. Portfolio (II) consisting of bonds, common stock, and REITs, 

3. Portfolio (III), consisting of bonds and common stock stocks during the pandemic and 

4. Portfolio (II) consisting of bonds, common stock, and REITs during the pandemic. 

Excel Solver is then used to derive an optimal portfolio using the combination of portfolios 

above, where the overall return is maximized at the lowest level of risk. Results from a pre-

study exercise show that a portfolio with REITs produces a lower Sharpe ratio than a portfolio 

without REITs. 

Below is the Sharpe ratio formula which adjusts expected returns by risk: 

SR'i = (
𝐸𝑅𝑖 

ϭ 
)            (1) 

where, 

SRi = Sharpe Ratio 



29 
 

ERi = excess return of portfolio, (ERi = ui – rf)  

ϭi= standard deviation of the excess portfolio return 

3.3 MEAN-VARIANCE SPANNING TEST 
 

To introduce a test for REIT diversification benefits, especially for 2020 to 2021 during the 

Covid-19 pandemic, we deploy various mean-variance spanning tests by (Huberman (1987), 

which are normally used to assess the diversification benefits of financial assets. Bélanger 

(2020) lays out the different types of mean-variance  (M-V) spanning tests that will be 

considered for this paper.  

As a financial and statistical test, the Mean-Variance Spanning Test is used to evaluate the 

effectiveness and diversification of an asset portfolio. It is sometimes referred to as the 

spanning test or the mean-variance efficiency test. The modern portfolio theory  (MPT), which 

was created by Harry Markowitz, serves as the foundation for the test. The main concept behind 

MPT is to build a portfolio that either minimizes risk for a given level of expected returns or 

maximizes expected returns for a given level of risk  (volatility). By comparing returns and 

risk, the Mean-Variance Spanning Test helps establish whether one portfolio may be regarded 

as efficient in comparison to another. 

Mean-variance spanning tests under normality. 

If the minimum-variance frontier of the K assets is the same as the minimum-variance frontier 

of the K assets plus additional N assets, then K risky assets span a wider set of N + K hazardous 

assets. The benchmark assets are the first set, while the test assets are the second set. Investors 

who care about the mean and variance of their portfolios will only be interested in the tangency 

portfolio of the risky assets  (i.e., the one that maximizes the Sharpe ratio) if there is a risk-free 

asset and unlimited lending and borrowing is permitted at the risk-free rate. (Bélanger, 2020) 

Investors in that situation are solely interested in determining whether the tangency portfolio 

utilizing K benchmark risky assets is identical to the one formed using all N + K risky assets. 

This can be summarized as follows: 

R2t = 𝛼 +𝛽R1t + 𝜖t           (2) 

R1 is a T x K matrix of the reference assets’ returns, R2, a T x N matrix of the test 

assets’ returns, 𝛽, a K x N matrix the regression’ coefficients, 𝛼, a vector of size N and 𝜖, a 

vector of size T of the error terms. 



30 
 

Mean-variance spanning tests under conditional heteroskedasticity. 

Returns are assumed normally distributed for the tests described by Kandel  (1987). However, 

in many cases, the return of assets exhibits heteroskedasticity, in which case a different 

approach needs to be considered. 

Step-down mean-variance spanning tests under conditional heteroskedasticity. 

The distance between tangent portfolios (i.e., a rise in the slope of the tangent lines) is rather 

minor, and the difference between global minimum variance portfolios primarily drives the 

strength of the tests. In other words, even if the economic difference is not significant, a slight 

decrease in the global minimum variance portfolio's standard deviation may cause the null 

hypothesis to be rejected (Bélanger, 2020). 

On the other hand, even if there is a considerable economic difference between the tangent 

portfolios, this will not result in the null hypothesis being rejected. Since statistical significance 

does not necessarily equate to economic significance, it is advised to take a method where each 

component is dealt with separately. 

Therefore, in accordance with Kan et al.  (2012), we test the hypothesis H0: α = 0N first and 

then H0: δ= 0N if α= 0N. If the initial test is not accepted, a conclusion is made that the tangent 

portfolios differ statistically. The minimum variance portfolios are statistically different if the 

second test is not accepted. 

Sub-sample analysis 

- Mahalanobis distance 

The Mahalanobis distance has been utilized by Li (2010) to proxy financial instability. 

Financial volatility is described as periods when asset prices tend to exhibit greater upward and 

downward price movements when compared to their historical patterns. Both the volatility of 

individual assets and their correlation are considered by this measurement. Krutzman et al. 

(2010) define the turbulence index as: 

dt = (yt − 𝝁) Σ−1 (yt − 𝝁)′           (3) 

where, 

• dt = Turbulence during t (a scalar). 

• yt = Vector of asset returns for the period t. 



31 
 

• 𝝁=Vector of asset returns means. 

• 𝚺=Covariance matrix of asset returns. 

- Business cycles 

Separating the portfolios into sub-samples based on the economic cycle, as called by Statistics 

South Africa (STATSA) or any other official government agency, further supports the normal 

investor in understanding the hedging potential of REITs compared to other asset classes. 

Further to the portfolio analysis, we introduce cointegration, which identifies long-horizon 

relationships between the REITS and the different asset classes, which is especially relevant 

for investors with long investment horizons, as is mostly the case in the property industry. 

Importantly, assets that are cointegrated share risk profiles and move in tandem, providing low 

or no diversification benefits (Seiler,2001). Cointegration methods are also robust to 

correlation instability, addressing a major limitation of traditional modern portfolio-based 

methods. 

3.4 COINTEGRATION 
 

Cointegration is a statistical concept used to assess the long-term relationship between two or 

more time series variables. In this context of REITs, we use it to determine whether there is a 

stable long-term relationship between the returns of different REITs or between REIT returns 

and other financial market indices.  

After identifying and collecting the data as laid out above, we analyze for cointegration, we 

carry out the following: 

3.4.1 Unit Root Test 

Before performing a cointegration test, it is important to check whether the individual time 

series is stationary. We can use unit root tests such as the Augmented Dickey-Fuller (ADF) test 

or the Kwiatkowski-Phillips-Schmidt-Shin (KPSS) test to evaluate stationarity. If the data are 

not stationary, we may need to differentiate the series until they become stationary. 

3.4.2 Testing for Co-integration 

The most common test for co-integration is the Autoregressive distributed lag (ARDL) 

procedure (Peresan, 2001). In this approach, we shall continue as follows: 

Estimate linear regression. Regress one of the time series on the other time series (s). The 

stationarity of the resulting series of residuals must be tested using unit tests. 



32 
 

i. If the residual is stationary, this indicates cointegration. In this case, we can proceed to 

the second step of the Engle-Granger procedure compared to Cointegration Testing: 

ii. Use a cointegration test, such as the Johansen cointegration test, to formally test for 

cointegration. The Johansen test is used to determine the number of cointegrating 

vectors, which represents the number of cointegrating relationships between variables. 

3.4.3 Explanation of Results 

If the cointegration test confirms that there is a cointegrating relationship between the variables, 

this implies that they will evolve together in the long run; in this case, this will be REITs with 

other common stock and fixed-income assets. The coefficients from the regression equation in 

step a. can provide insight into the nature of this long-run relationship.   

3.4.4 Error correction model (ECM) 

Once the co-integration process is established, we can build an error correction model. This 

model allows analysis of short-term dynamics and adjustment towards long-term equilibrium 

when deviations occur. 

3.4.5 Forecasting and Analysis 

After confirming co-integration, we use the relationship to make long-term forecasts, perform 

a VECM to understand the direction of influence, and conduct further analysis and pseudo-

testing theory. Additionally, cointegrating relationships can change over time, so this analysis 

would need to be regularly updated to re-examine cointegrating relationships as needed. 

3.5 DATA AND DATA SOURCES 
 

To exclude geo-economic influence that could affect performance or distort risk, this paper 

will focus on REITS who have a minimum of 50% of their fixed-property and real estate equity 

investments in South Africa for portfolio construction. 

FTSE/JSE Real Estate Investment Trust Index, FTSE/JSE South Africa Listed Property Index, 

and the Government Bond Index with weekly total returns are analysed from 2014 to 2023  

(approximately 10 years). Weekly returns were used to smooth out intraday fluctuations and 

market noise, which would have tended to be present in daily returns. This allows the analysis 

to focus on more meaningful long-term trends, which the study is focused on, while retaining 

most short-term movements which would have otherwise been smoothened with monthly 

returns. Weekly returns also provide more data points, which increase statistical power 

analytically, which could be missed in the less frequent quarterly or monthly data, providing 

more robust results with more substantial relationships between variables. 



33 
 

The asset types we consider are South African government bonds, REITs, and common equity. 

The total return index for SA REITs is the FTSE/JSE J867. Most companies formerly 

recognized as Property Unit Trusts (PUTs) and Property Loan Stocks (PLSs) are now listed in 

the REIT index because they converted to this structure, so separating them is unnecessary. 

This Index includes non-SA REITs (i.e., companies identified as REITs in foreign markets). 

The stock series is the FTSE/JSE All Share Index (J203). The bond index used is the All-Bond 

Index (ALBI), which consists of the top twenty conventionally listed basic bonds; bonds with 

a term of less than one year are not included. 

3.6 PERIOD OF STUDY 
 

This paper will consider a period aligned with what real estate investors consider when 

investing in property, be it an acquisition or development, and when they dispose of assets. 

According to Wang (1993) taking, using over 1000 Canadian commercial real estate 

transactions, the average holding period is about 5–8 years, depending on property the type. 

Since the average holding period of property is 4 or 5 years, this study will consider two periods 

of this term for this research which is 10 years. Brown (2020) also finds that the average holding 

period for these properties is around five years. According to one of the largest REITs in their 

2020 interim results presentation, the holding period for the valuation of multi-tenanted 

properties is four years and single-tenanted properties ten years. REITs have also been around 

for almost ten years in South Africa, allowing us to get more accurate results. Brown (2020) 

states that a planned holding period is attained through the various financial models providing 

a present value and the internal rate of return. By tradition, many such projections are for a 

holding period of five years (Ziobrowski, 1999). 

 

 

  



34 
 

4 CHAPTER FOUR - DATA ANALYSIS 
 

Version 12 of the E-Views software was used to process the data due to its efficacy in 

examining time-series data. In interpreting the results, we consider the p-values associated with 

each coefficient. If a p-value is below a chosen significance level (0.05), the coefficient is 

statistically significant, and those coefficients that are statistically insignificant may be less 

reliable. Using MacKinnon (1996) one-sided p-values, the t statistic for a 5% level is  1.96, 

therefore; if this value is less than 1.96 then the variable lag is insignificant, and the opposite 

applies for a t statistic greater than 1.96. 

4.1 DATA  
 

The data used for graphic and descriptive analysis is presented in Table 3 below. 

Table 3 Outline of data used. 

 Indicator Description 

J805TR INDEX FTSE/JSE SA REIT Index 

JALSH INDEX JSE All Share Index 

IGOV TR INDEX FTSE/JSE Inflation-Linked Government index 

CP507394 Corp Republic of South Africa International Bond 1 

EI879470 Corp Republic of South Africa International Bond 2 

EI595689 Corp Republic of South Africa International Bond 3 

NPN  Naspers Ltd. 

FSR  First Rand Ltd. 

SBK  Standard Bank Ltd. 

AMS  Anglo American Platinum Ltd. 

GRT  Growthpoint Properties Ltd. 

VKE  Vukile Property Fund Ltd. 

RDF  Redefine Properties Ltd. 

 

4.2 DESCRIPTIVE STATISTICS 

 

Tables 4 and 5 present the descriptive statistics of the various assets for the full sample and 

during the COVID-19 pandemic periods, respectively. For the ten years, average weekly 

returns for the REIT indexes vary between 0% and 0,001%, whilst bonds all have negative 



35 
 

returns and common stock is all positive, with the highest being 0,005%. The volatility of all 

indexes during this period was also observed, with bonds being the lowest, followed by 

common equity and REITs being in a similar range. This is based on the standard deviation of 

bonds between 0.01% and 0.02% and common equity at 0.02% and 0.07%. In contrast, REITs 

are between 0.04% and 0.07%, which would imply the impact of having all three in one 

portfolio, giving a wider mean variance spread. When comparing these measures against 

REITs, common equity appears to provide better risk-adjusted returns, showing the highest 

return at a moderate risk level. In contrast, bonds provide more conservative returns at the least 

level of risk. REITs appear to have greater maximums than the rest, which means there have 

been periods where it would have been an incremental value to have them in a portfolio.  

 

Table 4 Descriptive Statistics for March 2013 to March 2023 

The table shows the descriptive statistics for the weekly return data of a portfolio with a 

combination of REITS, Common stock, and Government Bonds for the 10-year period                                                                                                                                     

Assets 

  AMS  

CP50739

4  E1595689  E1879470  FSR  GRT  RDF  NPN  SBK  VKE 

Mean 0,0006 -0,0004 -0,0006 -0,0002 0,0008 -0,0003 0,0008 0,0045 0,0058 0,0011 

Median -0,0001 0,0000 -0,0002 -0,0001 -0,0017 0,0000 0,0000 0,0032 0,0035 0,0000 

Maximum 0,1707 0,0684 0,1018 0,0388 0219083 0,3635 0,4828 0,4155 0,3122 0,2036 

Minimum -0,1003 -0,1065 -0,1646 -0,0733 -0,3446 -0,2499 -0,3467 -0,3129 -0,2850 -0,2500 

Std.Dev. 0,0298 0,0117 0,0225 0,0102 0,0521 0,0482 0,0731 0,0657 0,0742 0,0355 

Skewness 0,3164 -1,6313 -0,8374 -1,4130 -0,7085 1,0385 1,4480 0,3474 0,3545 0,3486 

Kurtosis 

562446

2 21,4040 11,7224 13,9073 9,6845 16,3217 14,9621 7,2702 4,9097 1425940 

Jarque-Bera 157,30 7540,16 1702,61 2740,09 >9999 >9999 >9999 403,97 89,56 2746,70 

Probability 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 

Sum 0,3073 -0,1946 -0,3023 -0,0988 0,3922 -0,1472 0,3959 2,3551 2,9991 0,5538 

Sum Sq. Dev 0,4586 0,0707 0,2615 0,0535 1,4038 1,1996 2,7620 2 232271 2,8492 0,6507 

Observations 518 518 518 518 518 518 518 518 518 518 

 

During the national state of disaster announced by the national government for the COVID-19 

pandemic, bonds exhibited the lowest average returns, followed by common equity, while 

REITs had the highest returns. REITs also had the second lowest variance for the period, 

between bonds and common equity, which implies that they would have reasonable Sharpe 

ratios relative to the other asset classes whilst also showing the largest maximum.  



36 
 

In summary, the risk-return relationship gives the impression that REITs are attractive 

investment opportunities. However, the relatively high kurtosis could exert an impact on their 

investment performance when the whole distribution is considered rather than only the first 

two moments. 

 

Table 5 Descriptive Statistics during the national state of disaster between the 27th of March 2020 to 

the 8th of April 2022.  

The table shows the descriptive statistics for the weekly return data of a portfolio with a combination 

of REITS, Common stock, and Government Bonds during the peak of COVID-19 in South Africa. 

 

 

4.3 CORRELATIONS 
 

Tables 6 and 7 present the correlation matrix between REITs and other asset class returns for 

the full sample and during the Covid-19 pandemic. 

For the full period, higher correlations are observed between the bonds of 0.62 to 0.85 but show 

very low and negative correlations with REITs. Common equity has very low correlations with 

each other, which could be a factor of the spread across the various asset classes with different 

qualities, but also have low and negative correlations with REITs. The correlation between 

common equity and bonds is also low extremely low. This suggests that REITs could have 

  AMS CP507394 E1595689 E1879470 FSR  GRT  NPN RDF  SBK  VKE  

Mean 0,0024 0,0009 0,0023 0,0008 0,0140 0,0104 0,0099 0,0121 0,0098 0,0102 

Median 0,0020 -0,0005 -9,8781 -42371983 0,0048 0 0,0065 0 0,0154 0 

Max 0,0602 0,0393 0,1018 0,0313 0,2190 0,3635 0,1625 0,4828 0,3122 0,2036 

Min -0,0598 -0,0189 -0,0439 -0,0215 -0,1894 -0,1898 -0,1823 -0,2695 -0,1667 -0,1491 

Std. Dev. 0,0252 0,0090 0,0197 0,0069 0,0612 0,0642 0,0627 0,1090 0,0753 0,0573 

Skewness 0,0120 2,0027 1,4685 1,4108 0,2757 1,9545 -0,1314 0,9910 0,4755 0,8240 

Kurtosis 2,6631 9,8800 9,2348 8,9271 4,6431 12,2472 3,6701 6,1351 4,3854 5,0904 

Jarque-Bera 0,5086 282,5619 211,7671 192,1163 13,3932 449,3544 2,3099 61,3347 12,5887 31,5913 

Prob. 0,7754 4,3901 1,0360 1,9163 0,0012 2,6542 0,3151 4,8010 0,0018 1,3805 

Sum 0,2610 0,0914 0,2418 0,0817 1,5007 1,1078 1,0592 1,2900 1,0487 1,0885 

Sum of 

 Sq. Dev. 0,0671 0,0085 0,0411 0,0051 0,3965 0,4372 0,4163 1,2590 0,6010 0,3485 

Observations 107 107 107 107 107 107 107 107 107 107 



37 
 

diversification benefits in the long term if included in a portfolio with the other two asset 

classes. During the Covid-19 pandemic, the correlations were relatively similar for all the asset 

classes. These were lower across the board.  

Table 6 Correlation Matrix between March 2013 to March 2023 

 

 

 

 

 

 

Assets 

Assets 
 

CP507394 Corp EI879470 Corp EI595689 Corp NPN S FSR  SBK  AMS  GRT  VKE  RDF  

CP507394 Corp 1,00 
         

EI879470 Corp 0,65 1,00 
        

EI595689 Corp 0,62 0,85 1,00 
       

NPN  -0,11 -0,09 -0,11 1,00 
      

FSR  -0,07 -0,01 -0,05 0,32 1,00 
     

SBK  -0,04 0,00 0,03 0,44 0,04 1,00 
    

AMS  -0,01 -0,01 0,01 0,07 0,34 0,05 1,00 
   

GRT  0,03 0,01 0,01 -0,01 -0,16 -0,01 -0,09 1,00 
  

VKE  0,02 0,02 -0,04 0,08 0,10 -0,03 -0,08 0,17 1,00 
 

RDF  -0,08 -0,06 -0,07 -0,01 -0,06 -0,05 0,02 0,06 0,04 1,00 

 

Table 7 Correlation Matrix during national state of disaster between 27th of March 2021 to the 8th of 
April 2022 

                                                                                                                             Assets 

 

 

 

 

 

 

 

 

 

   

Assets 

 
CP507394 Corp EI879470 Corp EI595689 Corp NPN  FSR  SBK  AMS GRT  VKE  RDF  

CP507394 Corp 1,00                   

EI879470 Corp 0,73 1,00                 

EI595689 Corp 0,71 0,92 1,00               

NPN  -0,20 -0,18 -0,28 1,00             

FSR  -0,27 -0,20 -0,26 0,18 1,00           

SBK  -0,15 -0,10 -0,09 0,37 -0,12 1,00         

AMS  -0,03 0,04 0,03 0,03 0,44 0,07 1,00       

GRT  0,17 0,07 0,13 -0,09 -0,20 0,00 -0,05 1,00     

VKE  0,17 0,13 0,07 0,07 0,06 0,04 0,02 0,23 1,00   

RDF  0,04 -0,01 -0,10 0,07 -0,07 -0,10 0,06 0,03 0,08 1,00 

 

For the full 10-year horizon, REIT total returns were negatively correlated with stocks and 

bonds, i.e., FSR vs GRT at -0.11%, GRT vs SBK at -0.09%, and VKE vs EI879470 at -0.08%. 

This could be ideal for diversification because it means that the returns move in opposite 

directions, which helps reduce overall portfolio risk. After all, when one asset's value decreases, 

the other asset tends to offset some losses by increasing in value. The pair showed strong 

correlations between each other, with a similar trend observed for the two during the national 

state of disaster. REITs, however, appear to have a stronger relationship with the rest of the 

market during this period. This could be a factor of all having exposure to the market spread 

headwind brought by the pandemic, leaving many economies in distress. 

Tables 8 and 9 present the correlation matrix between REIT, common equity, and government 

bond index returns for the full sample and during the Covid-19 pandemic. The tables below are 

more straightforward due to the number of variables and, therefore, easier to interpret. 



38 
 

Table 8 Correlation matrix for indices from 2013 to 2023 

 J8055TR JALSH_INDEX IGOV_TR_ 

J805TR 1   
JALSH -0,3126 1  
IGOVTR -0,3772 0,9199 1 

 

Table 9 Correlation matrix for indexes during national state of disaster 

 J8055TR_IN JALSH_INDEX IGOV_TR_I 

J805T 1   
JALSH_ -0,9539 1  
IGOV_ -0,9250 0,9395 1 

 

4.4 PORTFOLIO OPTIMIZATION 
 

In this section, two portfolio sets are re-introduced, which are as follows: portfolio 1: portfolio 

containing common equity, bonds as well and REITs, and portfolio 2: portfolio with common 

equity and bonds without REITs. 

Below is a comparison of four Sharpe ratios for the portfolios, which show different scenarios 

to study how portfolio 1 and portfolio 2 perform in the full period and during the Covid-19 

pandemic. The risk-free rate is from the South African Reserve Bank. 

Table 10 Sharpe ratios of the different portfolios during COVID-19 and the full sample period 
 

10 Years          Covid-19 

Portfolio 1 0,03 0,38 

Portfolio 2 0,09 0,20 

 

Figures 3 to 6 present the portfolio opportunity sets for the full sample and, during Covid-19 

while figures 7 to 10 present their respective efficiency frontiers. From the various efficient 

frontiers below and the Sharpe ratios, there are several different outcomes to consider. Firstly, 

REITs appear to have had a negative impact during the longer period because the portfolio 

Sharpe ratio when the asset class is included during the pandemic is higher than that of the 

portfolio without REITs; however, when looking at the Sharpe ratios on the 10-year horizon, 

the Sharpe ratio lower than the Sharpe ratios are both are lower than 0.1%. When comparing 

the portfolio comprising all three asset classes during the pandemic vs the portfolio with only 

stocks and bonds, having REITs in a mixed portfolio would have provided a benefit. 



39 
 

Figure 3 Portfolio 1 Opportunity Set  Figure 4 Portfolio 1 Opportunity Set  during Covid-19

  

Figure 5 Portfolio 2 Opportunity Set                 Figure 6 Portfolio 2 Opportunity Set during Covid-19

  

 

 

 

Figure 7 Portfolio 1 Efficient Frontier     Figure 8 Portfolio 2 Efficient Frontier 

      
          

-0.05%

0.00%

0.05%

0.10%

0.15%

0.20%

0.25%

0.30%

0.35%

0.00% 2.00% 4.00%

R
e

tu
rn

Risk
-0.20%

0.00%

0.20%

0.40%

0.60%

0.80%

0.00% 1.00% 2.00% 3.00% 4.00%

R
e

tu
rn

Risk

0.00%

0.05%

0.10%

0.15%

0.20%

0.25%

0.30%

0.35%

0.40%

0.45%

0.00% 2.00% 4.00% 6.00%

R
e

tu
rn

Risk

0.00%

0.20%

0.40%

0.60%

0.80%

1.00%

0.00% 1.00% 2.00% 3.00% 4.00% 5.00% 6.00%

R
e

tu
rn

Risk

0.0%

0.5%

1.0%

1.5%

2.0%

0.0% 0.2% 0.4% 0.6%

R
is

k

Return

0

0.005

0.01

0.015

0.02

0.025

0.03

0 0.002 0.004 0.006

R
is

k

Return



40 
 

Figure 9 Portfolio 1 Efficient Frontier during         Figure 10 Portfolio Efficient Frontier during COVID-19 

Covid-19 

   

 

4.5 MEAN-VARIANCE SPANNING TEST 
 

A Mean-Variance Spanning Test is a method used to evaluate whether one investment portfolio 

dominates another in terms of mean returns and variance of returns. The test is typically used 

in the context of Modern Portfolio Theory (MPT). The method of mean-variance spanning, 

initially introduced by Kandel (1987), statistically tests the impact that the introduction of 

additional N risky assets, referred to as test assets, has on the efficient frontier of an investment 

opportunity set of K benchmark assets.  

The test is made from the perspective of a South African investor seeking diversified exposure 

on JSE with REITs as a proxy. If adding more assets does not increase the efficient frontier 

that a set of assets generates, then the set of assets is said to span the mean variance. Among 

other things, mean-variance spanning has been used to test linear-factor asset pricing models, 

analyse mutual fund performance, evaluate the advantages of international diversification, and 

assess the viability of alternative asset classes. 

 

4.5.1 Mean-variance Spanning Tests under normality. 

Having established that this dataset exhibits non-normal behaviour, the standard mean-variance 

spanning tests is invalidated as this assumes that the returns of assets are normally distributed. 

If the actual distribution of returns deviates significantly from normality, the results of the tests 

may be biased or unreliable. 

0.0%

0.5%

1.0%

1.5%

2.0%

0.0% 1.0% 2.0% 3.0% 4.0%

R
is

k

Return

0.0%

0.5%

1.0%

1.5%

2.0%

2.5%

3.0%

3.5%

4.0%

0.0% 0.5% 1.0% 1.5%

R
is

k

Return



41 
 

Table 11 Normality tests for Portfolio 1 and 2 

 Portfolio 1 Portfolio 2 

Jarque-Bera 895.8841 5418383 

Probability 0.000*** 0.000*** 

 

4.5.2 Mean-variance spanning tests under conditional heteroskedasticity. 

Generalized Method of Moments (GMMs) spanning tests are performed to control for the non-

normality of returns and the heteroscedasticity of error terms. With this, we try to establish 

whether the portfolio of REITs spans the common equity portfolio, and as a proxy, we also 

consider the effect of bonds. The dependent variable, portfolio 1, is only common equity, while 

the independent variable, portfolio 2, is REITs, and portfolio 3 is bonds. The below provides 

further insight as to whether adding REITs to the portfolio would have provided a 

diversification benefit. At more than 5% (p >0,05), we reject the serial correlation null 

hypothesis, implies that the selected REITs portfolio has no significant effect on the existing 

portfolio. Only bonds in this model prove to be significant (p < 0.05). with a negative 

coefficient which means that bonds have a negative impact on the variable the existing portfolio 

of common equity. 

Table 12 GMM of Portfolio 1 and 3 during full sample period 

Variable Coefficient Std. Error t-Statistic Prob.   

Constant -0.000 0.000 -1.501 0.133 

POTFOLIO2 0.000 0.000 5.398 0.000*** 

     

R-squared:  0.002 

Adjusted R-squared: 0.000 

 

4.5.3 Sub-sample analysis 

As a sub-sample analysis for different business cycles, the Covid-19 period is selected as an 

element of focus for this research. As with the full sample, the dependent variable, portfolio 1, 

is only common equity, while independent variable portfolios 2 and 3 are REITs and bonds, 

respectively. This is important for the investor interested in the behaviour of REITs in a 

portfolio during an economic downturn, particularly the pandemic. This also allows us to better 

understand the performance of REITs and their resilience to such economic conditions. For 

this, we continue with the period selected above, which is 27th of March 2020 to the 8th of April 

2022. The same period was used for all portfolios. In this instance, the independent variable 

portfolios had an insignificant effect on the portfolio of common equity. 



42 
 

Table 13 GMM of Portfolio 1 and Portfolio 2 during the COVID-19 period 

 

Variable Coefficient Std. Error t-Statistic Prob.   

Constant 0.006 0.007309 0.833 0.406 

PORTFOLIO2 -0.000 2.19E-05 -7.547 0.000*** 

 

R-squared  0.014126 

Adjusted R-squared 0.004737 

 

4.6 COINTEGRATION 
 

This section follows  (Octavia et al.2020) to analyze the cointegration and causality relationship 

among the selected market indexes: South African REIT, common equity, and bond markets. 

4.6.1 Unit Root Test 

Initially, the stationarity of the three indexes is examined using a unit root test and visual 

observations of the trend over time. A unit root is a characteristic of a time series that 

indicates it has a stochastic or random trend, making it non-stationary. This is because 

cointegration refers to a linear combination of non-stationary indices. Specifically, the 

augmented dickey fuller test Dickey (1981) was selected for this purpose. The Null Hypothesis 

for this test (H0) is that the series has a unit root which implies non-stationarity or the variable 

not reverting to the mean overtime.  

Figure 11 All Share Index Graph for the 10 year period 

 

With reference to the JSE All Share Index trend under observation, the variable shows a 

gradual increase from 2014 to 2017, some oscillation tendency in 2018-2019, and a clear 

increase until 2023. The REIT Index has seen greater movements from inception, with a large 

30,000

40,000

50,000

60,000

70,000

80,000

90,000

3/
10

/1
4

12
/1

2/
14

6/
3/

15

11
/2

0/
15

5/
8/

16

11
/1

1/
16

4/
21

/1
7

10
/2

/1
7

4/
20

/1
8

9/
14

/1
8

3/
5/

19

8/
30

/1
9

2/
28

/2
0

8/
5/

20

1/
29

/2
1

8/
1/

21

1/
4/

22

7/
1/

22

12
/1

6/
22

10
/3

/2
3

JALSH INDEX



43 
 

drop in 2020 (which is clearer observed in the pandemic test), then returning to the initial 

position. Lastly, the bond index shows consistent growth from 2014 until 2023.   

Table 14 All Share Index Unit Root Test 

H0: D (JALSH_INDEX) has a unit root                                                   

    t-Statistic Prob.* 

Augmented Dicker-Fuller test statistic   -15,173 0.00 

Test critical values: 1% level -3,447   

  5% level -2,868   

  10% level -2,570   

 

Figure 12 REIT Index Graph for the 10-year period 

 

 

Table 15 REIT Index Unit Root Test 

H0:  D (J805TR_INDEX) has a unit root                                                   

    t-Statistic Prob.* 

Augmented Dicker-Fuller test statistic   -18,711 0.000*** 

 1% level -3,447   

Test critical values: 5% level -2,868   

  10% level -2,570   
 

 

 

 

 

4,000

6,000

8,000

10,000

12,000

14,000

16,000

18,000

3
/1

0
/1

4

1
2

/1
2

/1
4

6
/3

/1
5

1
1

/2
0

/1
5

5
/8

/1
6

1
1

/1
1

/1
6

4
/2

1
/1

7

1
0

/2
/1

7

4
/2

0
/1

8

9
/1

4
/1

8

3
/5

/1
9

8
/3

0
/1

9

2
/2

8
/2

0

8
/5

/2
0

1
/2

9
/2

1

8
/1

/2
1

1
/4

/2
2

7
/1

/2
2

1
2

/1
6

/2
2

1
0

/3
/2

3

J805TR INDEX



44 
 

Figure 13 Bond Index Graph for the 10 year period 

 

 

Table 16 Government Bond Index Unit Root Test 

H0: D (IGOV_TR_INDEX) has a unit 
  

    t-Statistic 

Augmented Dicker-Fuller test statistic   -15,882 

 1% level -3,447 

Test critical values: 5% level -2,868 

  10% level -2,570 
   

   
However, after applying the first differences to all the variables, the p-value is less than 5%, 

which means the null hypothesis can be rejected. With the unit root now absent, it can be 

concluded that they are stationary at first difference. 

4.6.2 Lag Length Selection 

Before proceeding, determining the optimal lag length is necessary, providing the optimal 

trade-off between model fit and complexity. To determine the optimal lag length of the time 

series, different information criteria were selected and employed, namely LR (sequential 

modified LR test statistic, each at 5% level), FPE (Final Prediction Error Criterion), AIC 

(Akaike Information Criterion), SC (Schwarz Criterion) and HQ (Hannan-Quinn Criterion) 

with lag orders ranging from 1 to 4. 

Furthermore, when testing for cointegration between two or more time series variables, the 

significance of the trend and intercept terms in the cointegration equation can also be crucial. 

A cointegrating relationship implies a long-term equilibrium association between the variables; 

the trend and intercept terms in the cointegration equation capture this long-run relationship. 

Testing the significance of these terms helps in determining whether cointegration is present. 

200

220

240

260

280

300

320

340

3/
10

/1
4

12
/1

2/
14

6/
3/

15

11
/2

0/
15

5/
8/

16

11
/1

1/
16

4/
21

/1
7

10
/2

/1
7

4/
20

/1
8

9/
14

/1
8

3/
5/

19

8/
30

/1
9

2/
28

/2
0

8/
5/

20

1/
29

/2
1

8/
1/

21

1/
4/

22

7/
1/

22

12
/1

6/
22

10
/3

/2
3

IGOV TR INVEX



45 
 

Table 17 Lag Length Selection 

Lag LogL LR FPE AIC SC HQ 

0 -7810.224 NA   3.14e+14  41.89396  41.92550  41.90649 

1 -7695.439  227.1091  1.78e+14  41.32675  41.45291  41.37685 

2 -7623.012  142.1352  1.27e+14  40.98666   41.20744*  41.07433 

3 -7597.281  50.08089  1.16e+14  40.89695  41.21236   41.02220* 

4 -7582.515  28.50367   1.12e+14*   40.86603*  41.27606  41.02885 

5 -7578.268  8.128708  1.15e+14  40.89152  41.39617  41.09191 

6 -7572.613  10.73441  1.17e+14  40.90945  41.50873  41.14742 

7 -7559.890   23.94497*  1.15e+14  40.88949  41.58339  41.16503 

8 -7551.652  15.37248  1.16e+14  40.89358  41.68210  41.20669 
Where AIC: Akaike information criterion, SC: Schwarz Criteria and HQ: Hannan–Quinn information criterion 

 

Table 18 Trend and Intercept Test 

Variable 

Coefficient Std. 

Error 

t-Statistic Prob. 

Constant 847,367 241,672 3,5062 0,000*** 

  @TREND -4,448 1,097 -4,0513 0,000*** 

     

R-squared  0,041 

Adjusted R-squared 0,038 

 

From Table 18, it is clear from the P-value that the end and intercept are statistically significant 

which means they should be retained for the unit root test. 

 

Table 19 Augmented Dickey-Fuller test for Error Term 

Null Hypothesis: has a unit root                                                                          

    t-Statistic Prob.* 

Augmented Dicker-Fuller test statist