Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of

Real Estate Ventures

by

Keith Chin‐Kee Leung

Bachelor of Commerce, Sauder School of Business
University of British Columbia, 2007

Submitted to the Program in Real Estate Development in
Conjunction with the Center for Real Estate
in Partial Fulfillment of the Requirements for the Degree of

Master of Science in Real Estate Development

at the

MASSACHUSETTS INSTITUTE OF TECHNOLOGY

February 2014
© 2014 Massachusetts Institute of Technology. All rights reserved.


Signature of Author__________________________________________________________________________________
MIT Center for Real Estate
January 30, 2014


Certified by___________________________________________________________________________________________
Richard de Neufville, PhD
Professor of Engineering Systems and of Civil and Environmental
Engineering
Thesis Co‐Supervisor


Certified by___________________________________________________________________________________________
David Geltner, PhD
Professor of Real Estate Finance and of Engineering Systems
Director of Research, Center for Real Estate
Thesis Co‐Supervisor


Accepted by___________________________________________________________________________________________
David Geltner, PhD
Chairman, Interdepartmental Degree Program in Real Estate Development
 Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of
Real Estate Ventures

by

Keith Chin‐Kee Leung

Submitted to the Program in Real Estate Development in
Conjunction with the Center for Real Estate
on January 30, 2014 in Partial Fulfillment of the
Requirements for the Degree of Master of Science in
Real Estate Development

ABSTRACT
This thesis introduces probabilistic valuation techniques and encourages their usage in the
real estate industry. Including uncertainty and real options into real estate financial models
is worthwhile, especially when there is an elevated level of unpredictability surrounding
the investment decision.

Incorporating uncertainty into real estate pro formas not only provides different results
over deterministic models, it changes the angle of attack to real estate valuation problems.
When uncertainty is taken into account, the focus shifts from simply maximizing financial
returns, to modeling and managing uncertainty to make better ex ante finance and design
decisions. The ability to add optionality in probabilistic financial modeling can enhance
returns by curtailing losses during downturns and taking advantage of upside conditions.
A step‐by‐step example is carefully crafted to demonstrate the simplicity with which
uncertainty, Monte Carlo Simulations and Real Options may be included into real estate pro
formas. The example is entirely Excel based and is separated into three parts with each
progressively increasing in complexity. SimpleCo Tower establishes the familiar
Discounted Cash Flow pro forma as a starting point. ModerateCo Tower describes how
uncertainty and Monte Carlo simulations can be incorporated into a pro forma while
illustrating the effect of non‐linearity on financial models. ChallengeCo Tower reveals how
real options can add value to an investment and how it should not be overlooked.

The case study illustrates how the techniques outlined in this thesis can add significant
value to real estate decisions without much added effort or investment in expensive
software. The case study also shows how the use of real world data to model uncertainty
can be put into practice.

Thesis Co‐Supervisor: Richard de Neufville, PhD

Title: Professor of Engineering Systems and Civil and of Environmental Engineering

Thesis Co‐Supervisor: David Geltner, PhD
Title: Professor of Real Estate Finance and of Engineering Systems
Director of Research, Center for Real Estate

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 2 
 
Table of Contents
ABSTRACT ................................................................................................................................................... 2 
Table of Contents ...................................................................................................................................... 3 
Table of Figures ......................................................................................................................................... 5 
Acknowledgements .................................................................................................................................. 6 
CHAPTER 1: Introduction ........................................................................................................................ 7 
1.1  Thesis Purpose ............................................................................................................................ 7 
1.2  Format of Presentation ............................................................................................................... 8 
1.3  Current Industry Practice: Excel, Argus and Discounted Cash Flow Analysis ........................ 9 
1.4  Reluctance to Adopt New Techniques and Reliance on Intuition ......................................... 10 
1.5  The Role of Modern Pedagogy in this Thesis .......................................................................... 12 
CHAPTER 2: SimpleCo Tower – A Deterministic Example .............................................................. 14 
2.1  Assumptions of a Deterministic Model .................................................................................... 14 
2.2  Projecting Cash Flows for SimpleCo Tower ............................................................................ 14 
2.3  Return Measures: NPV and IRR ................................................................................................ 15 
CHAPTER 3: ModerateCo Tower ‐ Incorporating Uncertainty into a Financial Model .............. 16 
3.1  Uncertainty in the Rent Growth Rate of ModerateCo Tower ................................................. 16 
3.2  Monte Carlo Simulations and Expected NPV ........................................................................... 17 
3.3  The Flaw of Averages and Jensen’s Inequality ........................................................................ 18 
3.4  A Different Approach to Real Estate Financial Analysis: Distributions and Risk Profiles ... 20 
3.5  Static Input Variables versus Random Walks ......................................................................... 22 
CHAPTER 4: ChallengeCo Tower ‐ Managing Uncertainty in Real Estate Projects ..................... 24 
4.1  Real Option Analysis in ChallengeCo using IF Statements ..................................................... 24 
4.2  ChallengeCo Tower’s Result with a Real Option ..................................................................... 25 
CHAPTER 5: Quantifying Uncertainty in the Real World ................................................................. 28 
5.1  Predictability in the Real Estate Market .................................................................................. 28 
5.2  Real Estate Economics and the Stock Flow Model for Office Properties .............................. 30 
5.3  Sources of Uncertainty in Real Estate ...................................................................................... 31 
CHAPTER 6: Managing Uncertainty in the Real World .................................................................... 37 
6.1  The Basics of Financial Options ................................................................................................ 37 
6.2  Sources of Value for Financial Options .................................................................................... 38 
6.3  The Valuation of Financial Options .......................................................................................... 40 

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 3 
 
 6.4  Real Options ............................................................................................................................... 41 
CHAPTER 7: Two World Trade Center Case Study ............................................................................ 44 
7.1  Scenario Background ................................................................................................................ 44 
7.3  Projecting Rents, Cap Rates, Construction Costs and Operating Expenses .......................... 46 
7.4  Pro Formas ................................................................................................................................. 51 
7.5  Real Option Triggers ................................................................................................................. 52 
7.6  Results ........................................................................................................................................ 53 
CHAPTER 8: Conclusion ......................................................................................................................... 55 
BIBLIOGRAPHY ........................................................................................................................................ 57 
Appendix A  Incorporating Uncertainty into a Financial Model ........................................... 60 
Appendix B  Performing Monte Carlo Simulations .................................................................. 62 
Appendix C  Creating Cumulative Distribution Functions (CDFs) in Excel ......................... 64 
Appendix D  Using IF Statements to Model Real Options for Real Estate Ventures .......... 66 

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 4 
 
Table of Figures
Figure 1: SimpleCo Tower Sketch ............................................................................................................ 14 
Figure 2: SimpleCo Tower Assumptions ................................................................................................. 14 
Figure 3: SimpleCo Tower Pro Forma ..................................................................................................... 15 
Figure 4: ModerateCo Tower Sketch ....................................................................................................... 16 
Figure 5: Normal Distribution Curve Used to Model Rent Growth ....................................................... 17 
Figure 6: Results from the Monte Carlo Simulation ENPV versus Deterministic NPV ........................ 18 
Figure 7: Non‐linearity in the Rent Growth Rate .................................................................................... 20 
Figure 8: ModerateCo Tower Cumulative Distribution Function .......................................................... 21 
Figure 9: Random Walk Illustration ........................................................................................................ 23 
Figure 10: Comparison of Returns between ModerateCo and SimpleCo .............................................. 23 
Figure 11: Three ChallengeCo Tower Options ........................................................................................ 25 
Figure 12: ChallengeCo Tower Expected NPVs ....................................................................................... 26 
Figure 13: World Trade Center Site Plan (PANYNJ, 2013) .................................................................... 44 
Figure 14: 2WTC Rendering (PANYNNJ, 2013) ...................................................................................... 45 
Figure 15: Real Estate Cycle Length ......................................................................................................... 49 
Figure 16: The Regular Sine Curve........................................................................................................... 49 
Figure 17: 2 WTC Sketch up of Alternatives ........................................................................................... 51 
Figure 18: 2 WTC Financial Model Results .............................................................................................. 53 
Figure 19: 2 WTC Distribution Function ................................................................................................. 54 

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 5 
 
Acknowledgements

Praises and thanks to God for the blessings throughout my time at MIT. My motivation and
drive comes from a sense of purpose that God provides me with.

I would like to thank the MIT‐SUTD International Design Center for funding this research. It
is my great hope that this thesis could, even in the tiniest way, help empower people to
undertake the impossible and design the unexpected.

I extend sincere appreciation to Professor Geltner and Professor de Neufville for their
wisdom and patience in developing this thesis. I am grateful and honored to have worked
with, not only two renowned intellectuals, but two great teachers. I’ll miss the laughs
during our weekly meetings in Dr. Geltner’s office. Thanks also to Professor Somerville at
UBC for giving me my first break in real estate ‐‐ I’ll never forget it.

The MIT CRE faculty and alumni are the best. Thank you for your inspiration. I look
forward to our next beer and insightful conversation about real estate and politics. Thanks
to the CRE class of 2013 for helping me in the trenches at MIT. It was a battle, but we all
made it, together. I can’t imagine life without the RECIII.

I am appreciative of my friends who always lift me up when I’m down, especially those at
CoL, APG, GCF, BSF, and VCBC. Thanks to AK who makes my day, every day.

Thank you to all of my superb former work colleagues for molding my professional
character and still providing encouragement even after my departure years ago.

Last but not least, a special acknowledgement goes to my family who support me no matter
what. We aren’t always a perfect bunch, but when it counts, we are there for each other.

Thanks again to all of you. My success is yours as much as it is mine.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 6 
 
CHAPTER 1: Introduction

“May the odds be ever in your favor.”

‐ Suzanne Collins, author of The Hunger Games

In the blockbuster science fiction novel and movie series, The Hunger Games, Collins
describes a dystopian society in which a handful of teenagers are engaged in an ultra‐
competitive battle to the death. This competitive environment could draw comparisons to
the arena of real estate investing, where deals are won or lost by razor‐thin margins. While
Collins’ quote suggests that nothing can be done about one’s odds in the world of her book,
this is not the case with real estate. With knowledge of probabilistic valuation methods,
real options, and economics, real estate professionals can effectively improve the odds in
their favor.

1.1 Thesis Purpose

The world of corporate finance was introduced to Monte Carlo methods approximately 50
years ago, significantly altering the valuation approach for derivatives. In contrast, there
has not been wide‐spread adoption of stochastic valuation techniques in real estate finance
despite the positive track record of Monte
Carlo Simulations in corporate finance Key Terms: Stochastic vs Deterministic

(Marshall & Kennedy, 1992). The benefits of A stochastic or probabilistic model

relies on probability to obtain its values
probabilistic valuation techniques for real
for future states of the system.
estate have been widely documented since
A deterministic model has no
the early 1990’s (Baroni, Barthélémy, & randomness involved in generating its
Mokrane, 2006; Farragher & Savage, 2008; future output values.

Louargand, 1992). Yet, the real estate

industry still relies on sensitivity analyses for their risk assessment of real estate
investments. Farragher and Savage’s 2005 survey of 32 intuitional investors and 156
developers showed that only 2% of these firms utilize Monte Carlo Simulation techniques.

The message from academia is not getting through to industry. The rejection of
probabilistic techniques by real estate professionals is due, in large part, to the inability of
academics to present a compelling argument for probabilistic financial modeling. Academic

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 7 
 
theses tend to be excellent at defining what concepts are, but have difficulty with coaching
the application of theoretical concepts to the real world. The fragmentation of the research,
which occurs because of the multi‐disciplinary nature of the subject matter, inhibits
acceptance of stochastic techniques because the true benefits are not realized together in a
sweeping overall view from the start of the process, all the way to the end. With roots in
engineering, mathematics, economics and finance, the concepts presented in this thesis
have never been presented together before.

This thesis advocates for the use of probabilistically‐based valuation in the real estate
industry by:

 organizing research from multiple disciplines,

 demonstrating the great numerical and strategic advantages of stochastic modeling,
 clarifying how little additional effort is required to achieve those advantages, and
 emphasizing the applicability of concepts described above to real world problems.

This thesis attempts to mend the disconnect between academia and industry by focusing
on the effective presentation of ideas and application of modern pedagogical theory.

1.2 Format of Presentation

This thesis is structured to appeal to a wide range of real estate professionals. The major,
big picture arguments for implementing probabilistic strategies may be of greater
importance to executives and managers, while an analyst may want to understand the finer
points of modeling uncertainty and real options in Excel. The chapters in this thesis vary in
their level of detail. Chapters 2, 3, and 4 walk through a simplified example that
incorporates elements of probabilistic valuation at a broad level to demonstrate the main
points of this thesis. Discussion in these chapters will tend to be more qualitative. For those
looking for a greater detail, the appendix describes how the ideas presented can be
implemented into Excel, step by step. Additionally, an Excel workbook of every example is
available for real estate practitioners to explore every cell. Chapters 5 and 6 show how the
probabilistic concepts translate to the real world, with a detailed case study of 2 World
Trade Center to bookend the thesis in chapter 7.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 8 
 
1.3 Current Industry Practice: Excel, Argus and Discounted Cash Flow Analysis

The tools of the trade for analyzing income producing properties are Microsoft Excel and
Argus. Over the last decade, the ability to work with Excel has become essential in the
business world, especially for graduates of business schools. Despite the prevalent usage of
Excel in the workplace, generally very few features of the program are used by
professionals. Excel users are largely unaware of the computing power available to them
and resort to using a handful of common finance calculator functions. However, this is not
the fault of professionals, as the user experience, beyond basic calculator functions,
becomes unintuitive and frustrating to those not familiar with computer programming.
Good coaching and constant practice is required to develop skills beyond basic calculator
functions in Excel and this thesis addresses this by providing easy‐to‐follow examples.

Argus is software designed to save real estate professionals time by allowing the input of
information through a graphical user interface (GUI). A pro forma is generated by Argus
once all the information is imputed. Argus, in particular, is useful for organizing lease
information and producing rent rolls, a task that is tedious when the analysis is performed
manually in Excel. Argus allows real estate analysts to assess the financial feasibility of a
deal quicker, enabling a firm to inspect a greater volume of deals. Unfortunately, Argus
does have a few drawbacks. While the pro forma is exportable to Excel, Argus does not
export the formulas which it uses to calculate its numbers, essentially making Argus a
“black box”; the inner workings and logic of the program cannot be inspected. Reliance on
the automation which Argus provides to real estate analysts could erode human
performance, as practice from working with the nuts and bolts of a real estate pro forma is
reduced. A similar argument is made over automation in aircraft cockpits, as reports, such
as Sarter & Woods (1994), express concerns over the ability of pilots to react to non‐
normal situations. Another issue is the inflexibility of Argus to adapt to a wide range of real
estate ventures. Argus is great at modeling “cookie‐cutter” projects, but its effectiveness is
reduced when it’s used to model complex real estate projects.

The main method of valuation for income producing real estate is the Discounted Cash
Flow (DCF) approach. While the direct capitalization method (using cap rates) is also

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 9 
 
widely used, the absolute reliance on one year’s net operating income relegates the direct
capitalization method to quick back‐of‐the‐napkin analyses. The DCF approach involves
projecting future years of cash flow and discounting them using a risk‐adjusted discount
rate to arrive at the Net Present Value of the project.

DCF pro formas are taught in introductory real estate and corporate finance courses in
universities around the world. There are slight variations in the way the DCF approach is
taught from school to school; this does little to deter the widespread usage of DCF pro
formas.

The deterministic DCF approach does possess limitations, however. First, the analysis of
uncertainty is very limited in DCF models. The discount rate reflects the level of risk in a
project, but this method oversimplifies risk by relying on single discount rate when there
are multiple sources of uncertainty. Also, the discount rate doesn’t take into account the
asymmetry between upside and downside risk – generally, downside events matter more
to investors than upside events. Thirdly, it ignores the effect of options or possible changes
which may occur to the real estate over the life of the investment as owners and managers
have flexibility to respond to changes in the economy by making decisions that affect future
cash flows. Despite its pitfalls, the DCF approach is well understood at all levels of
experience in the real estate industry which makes it a good starting point to discuss
probabilistic valuation techniques from. The basic DCF pro forma is highlighted in Chapter
2.

1.4 Reluctance to Adopt New Techniques and Reliance on Intuition

Why has the adoption of probabilistic valuation techniques, such as Monte Carlo
Simulations, not occurred in the real estate industry? Byrne (1996) suggests that both the
small teams and the entrepreneurial nature of the real estate industry prevents the full
acceptance of probabilistic methods in financial modeling. But shouldn’t the
entrepreneurial spirit of the industry translate into an insatiable appetite to find an edge to
get ahead of the competition?

Without a doubt real estate teams are small. Whether the teams are based in the largest
investment banks or in the largest multi‐national developers, only a few analysts and even

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 10 
 
fewer managers are involved in the decision‐making process in any given real estate
investment. The heavy workload on open deals could crowd out time available to spend on
improving processes, thus perpetuating the status quo. The notion that real estate firms are
not embracing stochastic valuation techniques because they are small should be rejected.
Real estate firms focus on efficiency and are likely to adopt new methods, processes, or
technology if the cost‐benefit rationale makes sense to them. Incorporating uncertainty
into the financial analysis of real estate ventures is a “low hanging fruit” and represents a
major improvement in analytics with very little effort or cost.

Real estate has always been perceived as less sophisticated compared to other asset classes
such as stocks or bonds. This perception was largely due to private nature of real estate
transactions and the lack of data available for economic analysis. While the market for
stocks has been developing since the 1600’s, real estate equity as a securitized asset only
began trading in the 1960’s. Without reliable data to guide finance decisions, real estate
professionals depended on their instincts and intuition to remain solvent during
recessions.

As any experienced professional knows, our instincts do fail us from time to time. Part of
the reason why uncertainty is overlooked is because it involves seeing financial losses as a
possibility. Negativity bias is a psychological phenomenon that may explain what happens
when we see losses or experience negative moments (Baumeister, Bratslavsky, Finkenauer,
& Vohs, 2001). A common example of this effect is the anti‐anticipation and stress of
receiving a large restaurant bill, which is further exacerbated if the actual bill amount is
unknown. Humans tend try to avoid these negative experiences that shake our confidence
even if great benefits are possible.

Previously published research advocating for the use of probabilistic valuation techniques
were missing a key component: data from a sufficient number of market cycles to describe
the behavior of market factors and uncertainty. With over 50 years of data available, the
time is ripe for real estate to explore scientific approaches. Appropriate usage of real estate
data from indices are discussed further in chapters 5 and 6.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 11 
 
The techniques described in this thesis will not eliminate the need for good instincts in real
estate, but rather, they will enhance decision‐making by providing different perspectives
on real estate problems.

1.5 The Role of Modern Pedagogy in this Thesis

The struggle of university researchers to connect with learners is well documented

(Seymour, 2008). The academic tenure system is cited as a major reason why teaching and
communication have taken a back seat, as professors are encouraged to push out
publications more than developing teaching skills. If professors have difficulty keeping
students in their classrooms engaged, what hope do they have in trying to engage readers
in a one‐way medium? Indeed, scholarly articles seem to be more effective in
communicating ideas to other academics, but what about the rest of society?

The main intent of this thesis is present probabilistic concepts to real estate professionals
with a high level of clarity. Often times, authors of scholarly articles enter into auto‐pilot
mode and deliver their ideas based on their own experience as learners or casual
observations. For this thesis, special attention is paid to pedagogy to prevent a researcher‐
centered teaching approach and move towards a learner‐centered approach.

As you might imagine, there is no scarcity of research on how adults learn. Described below
are two major theories in modern pedagogy which guide the manner of presentation for
concepts introduced in this thesis. The first theory is of mental models, or schemas. Child
psychologist Jean Piaget proposed a process in which children use their interactions with
the world to develop models of objects and patterns of action (Lang, 2008). It turns out
that what we know already about the world greatly influences how we encounter new
experiences; our existing models are under constant revision. When adults are met with
new experiences or ideas, they work to fit these new elements in to patterns which they
already understand. There are two learning processes which can occur when a person
encounters a new experience: assimilation and accommodation. Assimilation occurs when
a person takes in a new idea by making the idea fit to into their existing models. On the
other hand, accommodation occurs when the new idea does not fit into any pre‐existing
models and changes are made to a person’s existing models to take in the new information.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 12 
 
Awareness of both these processes is crucial to effective delivery of the ideas presented in
this thesis. In some scenarios, assimilation needs to occur which requires the presenter to
help the audience connect to pre‐existing knowledge. An example of this occurring in this
thesis is the use of the familiar Discounted Cash Flow pro forma as a starting point for more
complex feature additions. For scenarios in which accommodation is likely to occur, clarity
is vital to cease the perpetuation of common misconceptions and pitfalls. Clarity is
emphasized when presenting the Flaw of Averages in Chapter 3.

Bloom’s Taxonomy is the second pedagogical theory that is applied in this thesis. Bloom’s
Taxonomy is a framework developed by Benjamin Bloom in 1956 to categorize learning
objectives. The framework divides educational objectives into three domains: cognitive,
affective, and psychomotor (Krathwohl, 2002). Skills in the cognitive domain include those
of knowledge and critical thinking. The affective domain include skills relating to emotion,
while the psychomotor domain focuses on skills with physical tools, such as hammers. The
cognitive domain is most relevant for the concepts presented in this thesis. In the revised
Bloom’s Taxonomy, Krathwohl presents 6 levels of processes in the cognitive domain. From
lowest complexity to highest, they are: remember, understand, apply, analyze, evaluate and
create. If the goal is to teach professionals how to create their own simulations, the
corresponding discussions and examples should match that goal in detail and complexity.
Since chapters in this thesis vary in their objectives (some professionals might only want to
go up to ‘understand’ level, while other will want to ‘create’), careful attention is paid to
maintain consistency in cognitive levels. Mismatched objectives and discussions lead to
frustration for readers. The appendices and chapters 5, 6, and 7 cater to readers who want
to reach the ‘create’ level, while next 3 chapters reside at the ‘understand’ level. We begin
gently by walking through the deterministic discounted cash flow pro forma.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 13 
 
CHAPTER 2: SimpleCo Tower – A Deterministic Example

The first example, SimpleCo Tower, is a  

simplified discounted cash flow pro forma of  
the kind analysts typically use to financially
 
model commercial real estate transactions.
 
SimpleCo Tower is a 10 story office tower with
 
a floor plate of 17,000 sf. A financial model is
 
created to evaluate the purchase of the Figure 1: SimpleCo Tower Sketch
 
building for a price of $17 million. There are A visual representation of the 10‐
 story office tower, SimpleCo Tower.
three major sections to a pro forma: the
assumptions, the cash flow projections, and the outputs.

2.1 Assumptions of a Deterministic Model

The assumptions are a set of parameters with

SimpleCo Tower Assumptions
which the financial model most abide by.
Purchase Price $100 /gsf
Gross Floor Area 170,000 sf Some assumptions are physical (such as floor
Efficiency 90% area and efficiency), while others are
Office Rent $30 /sf
Rent Growth Rate 3% economic (such as rent and discount rate.)
Expense Growth  3% Estimating the assumptions accurately is
Stabilized Vacancy 5%
important because they drive all numbers in
Expenses $15 /sf
Capital Expenditures 10% of NOI the pro forma.
Terminal Cap Rate 11.00%
OCC/Discount Rate 12.50% 2.2 Projecting Cash Flows for SimpleCo
Figure 2: SimpleCo Tower Assumptions Tower
This chart can be viewed in the
The cash flow projection section of the pro
SimpleCo Excel file on the CD.
forma projects many line item several years
into the future. Variables in the formulas are often linked or referenced to the assumptions
on this page. The cash flow projection organizes the revenues and costs associated with a
particular property and calculates the net cash inflows/outflows for each year of property
ownership. In SimpleCo Tower, the Property before Tax Cash Flow (PBTCF) is calculated
without the effects of income tax or leverage.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 14 
 
  
(in 000's) Year 1 2 3 4 5 6 7 8 9 10 11
Potential Gross Income $4,590 $4,728 $4,870 $5,016 $5,166 $5,321 $5,481 $5,645 $5,814 $5,989 $6,169
 
Vacancy $230 $236 $243 $251 $258 $266 $274 $282 $291 $299 $308
Effective Gross Income $4,361 $4,491 $4,626 $4,765 $4,908 $5,055 $5,207 $5,363 $5,524 $5,689 $5,860
 
Operating Expenses $2,550 $2,627 $2,705 $2,786 $2,870 $2,956 $3,045 $3,136 $3,230 $3,327 $3,427
 
Net Operating Income $1,811 $1,865 $1,921 $1,978 $2,038 $2,099 $2,162 $2,227 $2,293 $2,362 $2,433
 
Capital Expenditures $181 $186 $192 $198 $204 $210 $216 $223 $229 $236
 
CF From Operations $1,629 $1,678 $1,729 $1,781 $1,834 $1,889 $1,946 $2,004 $2,064 $2,126
 
Reversion (Purchase and Sale) ‐$17,000 $22,120
PBTCF ‐$17,000 $1,629 $1,678 $1,729 $1,781 $1,834 $1,889 $1,946 $2,004 $2,064 $24,246
 
Figure 3: SimpleCo Tower Pro Forma
The cash flow projections are shown for SimpleCo Tower. This chart can be viewed in
the SimpleCo Excel file on the CD.

2.3 Return Measures: NPV and IRR

The output section of a DCF pro forma calculates the objective return measures. In the
world of finance, no return measure is as prevalent as Net Present Value (NPV), or its
sibling the Internal Rate of Return (IRR).

Net Present Value and IRR For SimpleCo Tower, our NPV at a 12.5%
The time value of money principle is the discount rate is ‐$135,000 and the IRR is
most fundamental in finance. Cash flow 12.37%.
today is worth more than cash flow in the
future because of interest earning SimpleCo Tower is a deterministic model.
potential. Future cash flows are
For each unique set of assumptions there is
discounted to arrive at an equivalent
value today called the Present Value (PV). one sole outcome. The output (NPV in this
Net Present Value is the sum of the PVs of case) is determined by the input
all future cash inflows and outflows of a assumptions to the exact cent. There is no
project.
uncertainty in the model because a set of
The Internal Rate of Return (IRR) is the
assumptions always lead to a sole output
discount rate which makes NPV equal 0.
return measure. Pressing the “F9” key
recalculates formulas in Excel, but doing so will never change the NPV in the SimpleCo
Tower pro forma.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 15 
 
The NPV can change if an assumption is manually altered. The effect on NPV of a change in
an assumption can be recorded in a sensitivity analysis. Utilizing a data table in Excel, the
change in NPV can be seen when one or two variables change (a sensitivity analysis is
performed on the rent growth rate in section 3.3). Unfortunately, this analysis is limited to
two variables and the real world usually doesn’t “hold all else constant”. What alternatives
are out there for financial modeling?

CHAPTER 3: ModerateCo Tower ‐ Incorporating Uncertainty into a Financial Model

Most real estate professionals are familiar

with the techniques described in the SimpleCo
Tower pro forma because the deterministic
DCF model is taught in many introductory
finance courses around the world.
ModerateCo Tower expands on the SimpleCo
Tower pro forma by adding uncertainty to one
of the assumptions, the rent growth rate.
Figure 4: ModerateCo Tower Sketch
A visual representation of ModerateCo Everything else about ModerateCo is the same
Tower, a 10‐story office tower that is
as SimpleCo. 
physically identical to SimpleCo Tower.

3.1 Uncertainty in the Rent Growth Rate of ModerateCo Tower

The SimpleCo Tower example assumed that the rent growth rate was 3% per year. Based
on the averaging of historic rent growth rates, 3% is a common assumption among real
estate professionals. When the rent growth rate is subject to uncertainty, it is
acknowledged that the true rent growth rate is unknown and varies within a range. Excel’s
random number function is used to simulate uncertain behavior. Appendix A goes through
step‐by‐step how uncertainty was built in to the ModerateCo Tower financial model.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 16 
 
 In ModerateCo, the
uncertainty is created
as a symmetrical
Normal distribution
around a mean. Using
3% as the mean for the
rent growth rate, there
should be an equal
chance for the growth
rate to appear above or
Figure 5: Normal Distribution Curve Used to Model Rent Growth below 3%.
Also known as a ‘Gaussian Distribution’, a random variable is
‘normalized’ according to this distribution. The RAND function
in Excel fetches a random number between 0 and 1 and is
centralized towards the mean. For example, if the number
comes out to be .159, it will be placed ‐1 standard deviation
from the mean. 68% of values (.159 to .841) will fall within 1
standard deviation of the mean.

3.2 Monte Carlo Simulations and Expected NPV

Once the financial model has input

Monte Carlo: What’s in a name?
assumptions that randomly change, the
As a favorite hangout of Ian Fleming’s
corresponding output NPVs can be fictional character James Bond, Monte
recorded many times using a Data Table in Carlo is often associated with luxurious,
mysterious and exotic living. Perhaps,
Excel. The process of running a model for a
Monte Carlo Simulations sound more
specified number of iterations is simply foreign then they actually are.
called a Monte Carlo Simulation. The NPV “Monte Carlo” Simulation just refers to a
will vary from simulation to simulation simulation where the number of
iterations are set by the user. For
because the input variables are always
example, we run 5,000 iterations for
changing. In the case of ModerateCo, the ModerateCo Tower, not one more nor one
input variable, Rent Growth Rate, changes. less.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 17 
 
The input variable could be 2% leading to a certain NPV value, and in the next iteration, the
rent growth rate could be 3.5%, leading to a higher NPV value.

After running 5,000 iterations of the model, ModerateCo Tower calculates the mean of the
5,000 NPVs to yield an Expected Net Present Value (ENPV). In this case, the mean of the
simulated NPV (ModerateCo) will be consistently greater than the deterministic NPV
(SimpleCo) even though 3% was used as the mean in ModerateCo. In other words, even if
multiple sets of 5,000 simulations were ran, the simulated ENPV of ModerateCo will
generally be significantly greater than the NPV of SimpleCo.

Figure 6: Results from the Monte Carlo Simulation ENPV versus Deterministic NPV
Interesting result! The simulated ENPV is an expected NPV because
it is just an average of all the results in a Monte Carlo Simulation. In
this case, ModerateCo’s NPV was recorded 5,000 times and
averaged to get an average of $375,575. The deterministic NPV is
taken directly from the SimpleCo pro forma. This result can be
viewed in the ModerateCo Excel file.

How could this difference occur? Shouldn’t the SimpleCo NPV and ModerateCo ENPV be the
same if we ran many simulations of ModerateCo?

Intuition may try to apply the Central Limit Theorem or Law of Large Numbers in this case.
As the number of iterations of a random independent variable becomes very large, the
variables will be normally distributed around the expected value (if using the NORM.INV
function). In fact, there should be close to an equal number of occurrences of rent growth
rate above and below the mean rent growth rate in ModerateCo since we are using a
symmetrical normal distribution to model the uncertainty in the rent growth rate. While
the input variable behaves this way with the expected value as its mean, this actually does
not extend to the output NPV. The Flaw of Averages explains why.

3.3 The Flaw of Averages and Jensen’s Inequality

First coined by Savage, Danziger, & Markowitz (2009), the Flaw of Averages is a major
error that occurs when using averages in deterministic models instead of proper stochastic
variables. De Neufville & Scholtes (2011) describe the Flaw of Averages as the widespread‐

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 18 
 
but‐mistaken assumption that evaluating a project around average conditions give a
correct result.

The simple math behind the Flaw of Averages concept is based on Jensen’s Inequality. In
1906, Danish mathematician Johan Jensen proved that:

Jensen’s Inequality
The average of all the possible outcomes associated with uncertain parameters is
generally not equal to the value obtained from using the average value of the
parameters.

Basically what happens is a symmetrically distributed input variable leads to an

asymmetric distribution of output values. When this occurs, the system or model is
described as non‐linear. The SimpleCo model is a perfect example because it’s pro forma
uses a 3% historic average for its rent growth. When the deterministic 3% is replaced with
an input random variable symmetrical distributed around 3%, the output NPV value ends
up significantly greater for ModerateCo over SimpleCo!

The source of non‐linearity in this case is annual compounding. The same effect that makes
compound interest (non‐linear) greater than simple interest (linear) at the same rate
generates the difference in returns between SimpleCo and ModerateCo.

For ModerateCo Tower, 3% is the mean growth rate, so a 2% growth rate and a 4% growth
rate should occur with equal probability. Because the curve is convex (due to
compounding), going up to 4% results in a greater upward NPV improvement [|2440‐(‐
135)|= 2,575] than the NPV erosion of going down to a 2% growth rate [|‐2,528‐(‐135)|
=2,393]. Systems behave asymmetrically when upside and downside effects are not equal.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 19 
 
 Figure 7: Non‐linearity in the Rent Growth Rate
Let’s say that we have two iterations of the model. In one iteration, the rent growth
rate is 2%, and the other is 4%. Leading to a NPV result set of ‐2,528 and 2,440. If we
average these two values, we get ‐44 which is higher than the result we would get at
3% of ‐$135! The difference becomes greater and greater as values further from the
mean are used. This sensitivity analysis of the rent growth rate can be found under
the SimpleCo pro forma, in the SimpleCo Excel file.

The Flaw of Average has a significant impact on NPV and could spell the difference between
winning a bid and losing a bid, as exemplified by the SimpleCo and ModerateCo
comparison.

3.4 A Different Approach to Real Estate Financial Analysis: Distributions and Risk
Profiles

Incorporating uncertainty into real estate pro formas not only gives a different result over
deterministic models (as per the Flaw of Averages), it changes the approach to real estate
valuation problems. In the deterministic SimpleCo Tower case, the strategy is to lock in a
set of ex ante assumptions based on the analyst’s best forecast, find the single best value
and hope for the best. When uncertainty is factored in to the analysis, the focus shifts to
modeling and managing the uncertainty to make better finance and design decisions today.
The single best expected value of NPV is no longer the sole objective in a stochastic model:
range and distribution of outcomes become relevant.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 20 
 
Introducing a realistic level of randomness into financial models changes the framing of the
valuation problem. Understanding the likelihood of losing or profiting become important
once we introduce uncertainty into the analysis. The cumulative distribution function
(CDF) of ModerateCo provides information on the probability of loss or profit scenarios.
ModerateCo has about a 50% probability of having negative NPV and a 50% probability of
a having a positive NPV. The downside probability is more limited than the upside
probability, as illustrated by the long tail towards the right (more positive NPVs).

This scenario is a typical observation for real estate projects. Ideally, an analyst will want to
manage the uncertainty by finding ways to limit the downside losses and accentuate the
upside profits.

Figure 8: ModerateCo Tower Cumulative Distribution Function

On the CDF, the likelihood of NPV outcomes is displayed as well as the NPV for the
deterministic SimpleCo Tower and the probabilistic expected NPV for ModerateCo
Tower. This chart and its corresponding Monte Carlo Simulation is included in the
ModerateCo Excel file, on the ‘ModerateCo Distribution” tab.

Other useful measures that come out of this analysis of distributions include Value at Risk
and Value at Gain. Value at Risk denotes how much loss could occur at a specified
probability over a time frame. In the ModerateCo example, the Value at Risk (V10 number)

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 21 
 
NPV is ‐$6 million. That is, there is a 10% probability that a negative $6 million NPV or
worse will be incurred over the 10 year life horizon of the investment. On the other side,
the “V90 NPV” is $7 million. This Value at Gain “V90” number can be read as: There is a
10% chance that the NPV for the project over the 10 year investment horizon will be over
$7 million.

3.5 Static Input Variables versus Random Walks

The rent growth rate’s behavior in the ModerateCo model is currently a static variable.
Once a rent growth rate is randomly generated for a scenario, it remains the same for the
life of the investment. Deterministic pro formas frequently model input assumptions as a
static variable because the basis for their assumptions are from historic averages of long‐
term annual rates. Economic conditions change over the life of a long‐lived investment and
deterministic financial models are poor at modeling this behavior. Since ModerateCo’s
input variables are randomly generated and do not rely on historic averages, a change over
time over can be modeled in to the annual rent growth rate.

Growth rates generally do not move independently from year‐to‐year with absolute
randomness; rates tend to vary around the results from the preceding period. Pearson
(1905) described this behavior as a “Random Walk”.

A random walk modeled into a pro forma will allow an investment’s profitability
performance to decline and recover over the investment horizon. This up and down
behavior is essential to the modeling of real options in the proceeding chapter.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 22 
 
 Random Walk Illustration for Rent Growth Rate Starting at 3%
Year 0 Year 1 Year 2 Year 3 Year 4 Year 5
6%
5% up 2% x
4% x up 1%
3% x dn 1% x
2% x dn 2%
1% dn 2%
0% x
Figure 9: Random Walk Illustration
A visual representation of a year‐to‐year random walk
evolution of the rent growth rate. This behavior can be
exhibited by many different variables.

The additional variability translates into greater volatility in the results of the Monte Carlo
simulation and amplifies effect of the Flaw of Averages.

The strong effect of the Flaw of Averages and Random Walk volatility should be enough
motive to start modeling real estate using probabilistic techniques. The thesis continues to
make the case for stochastic valuation of real estate in ChallengeCo Tower by using Real
Options.

Results ENPV St. Dev.

ModerateCo $192,043 $4,920,803
ModerateCo w/ Random Walk $1,042,254 $9,240,832
SimpleCo Deterministic NPV ($134,701)
Figure 10: Comparison of Returns between ModerateCo
and SimpleCo
Random‐walk behavior adds greater volatility to the
model which inflates the effect of the Flaw of
Averages even further. The Monte Carlo Simulations
and results can be viewed in the ModerateCo Excel
file.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 23 
 
CHAPTER 4: ChallengeCo Tower ‐ Managing Uncertainty in Real Estate Projects

With distributions, we gather information that will aid us in finance decision‐making. This
chapter focuses on how to use this information advantageously. Real option analysis is
utilized to explain the impact of adding flexibility into the design or financial models.

A Real Option is described specifically as a “right without an obligation”. Chapter 6

provides a detailed discussion on Real Options, but for now, a basic option to add more
floors in the future to the on‐going example of SimpleCo and ModerateCo towers will be
described.

ChallengeCo does not stray much from the enduring example. The subject building is still a
10 story office building. However, ChallengeCo Tower is now an investment in a 10 story
development project instead of a pre‐existing stabilized office tower. Rather than a
purchase price, we use a development cost to build the project. An option to build 10
additional floors in the future is examined further in the ChallengeCo Tower pro forma
provided in the ChallengeCo Excel file.

Almost all input assumptions are subject to uncertainty using the same NORM.INV function
described in ModerateCo. Additionally, the input assumptions will exhibit “random walk”
behavior, with the preceding year’s value used as the mean for next year’s value. Each input
assumptions will go through their own random walks, culminating into a specific NPV for a
unique 10 year unique state of the world.

4.1 Real Option Analysis in ChallengeCo using IF Statements

Using IF statements in Excel, real options can be modeled with ease. Two pieces of
information are required to model real options. Firstly, the “trigger” conditions need to be
specified: What conditions need to occur before the option is exercised? Secondly, the
exercise costs and other consequences of the option need to be identified: What is the effect
if the option is actually exercised? Once these two pieces of information are detailed, the
option can be modeled into the pro forma. The objective here is to model the option in such
a way that the consequences of an exercised option are automatically displayed in the pro
forma if the predetermined conditions occur. Then, a Monte Carlo Simulation examines the

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 24 
 
effects of the option on the NPV in comparison with an identical development without the
option. Appendix D, discusses the use of “if statements” to model real options in greater
detail.  

Figure 11: Three ChallengeCo Tower Options

The building on the left represents the Inflexible ChallengeCo project at 10 floors. It
is physically identical to ModerateCo and SimpleCo Towers. In the middle shows a
flexible design where an additional 10 floors can be built on top of the first phase of
10 floors. On the right is the inflexible 20 floor ChallengeCo Tower.

For ChallengeCo, three separate pro formas are created to show the difference in expected
NPV and distributions. One pro forma calculates the NPV for a development project with a
flexible design option built‐in to the model to construct an additional 10 floors at a later
date. The second pro forma calculates the NPV for a standard 10 story development with no
option built‐in. The third pro forma calculates the NPV for a 20 story development without
an option built‐in to the design.

4.2 ChallengeCo Tower’s Result with a Real Option

The results from the Monte Carlo Simulations show that design flexibility can have a
significant financial value. While the development with flexibility never dominates the two
option‐less alternatives (the flexible alternative distribution function is always to the left of
either the 10 story or 20 story distribution function), the results show how the flexible
alternative can be advantageous.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 25 
 
 Figure 12: ChallengeCo Tower Expected NPVs
The CDF shows how the flexible design (in red) uses the real option to take
advantage of upside conditions. At the low end, the flexible design does not exercise
its option to expand, so its CDF curve closely follows the curve of the 10 floor
inflexible design. If economic conditions are good, the flexible design begins to
deviate from the 10 floor inflexible design by exercising its option to expand and
follows closer to the 20 floor inflexible design curve to take advantage of the good
economy. The Monte Carlo Simulations and CDFs can be found in the ChallengeCo
Tower Excel file.

If economic conditions turn sour in the next 10 years, the option to expand is not exercised
and the distribution function of the flexible alternative “hugs” the 10 floor inflexible option.
In poor economic times, the flexible option will not perform as well as the 10 floor
inflexible development because some extra construction costs are “sunk” into the initial
construction cost of the flexible alternative (for example, constructing stronger columns to
take the load of a possible 10 floor addition). On the other hand, the flexible alternative
performs much better than the 20 floor inflexible development during a poor economy.

When economic conditions are good, the flexible alternative takes advantage of the upside
by exercising its option to build more space. This is illustrated when the 10 floor inflexible
alternative is compared with the flexible option above the $5 million NPV mark. The

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 26 
 
flexible alternative will deviate from the 10 floor inflexible alternative and capitalize on the
opportunity of great market conditions.

The expected NPV of the flexible alternative is the greatest among the three alternatives for
ChallengeCo Tower. For investors seeking to limit their downside exposure, while taking
advantage of the upside as much as possible, flexibility can be a major win. Flexibility in
design should not be overlooked when making investment decisions.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 27 
 
CHAPTER 5: Quantifying Uncertainty in the Real World

The example presented in Chapters 2 to 4 is simplified to underline the importance of

including uncertainty and managing the risk that exists in real estate investing. In this
chapter, the focus shifts to the execution of these techniques in the real world. To
implement stochastic techniques into a real world financial model, the inputs that are
subject to uncertainty must be quantified with a decent level of accuracy or the outputs
cannot be trusted – the computer science axiom, “Garbage In, Garbage out” is appropriate
here. Accuracy is important, but how precise or detailed should a financial model be? The
real estate industry has embraced the DCF method which, as demonstrated by the Flaw of
Averages in chapter 3, is generally less accurate and much less detailed than the simplest
stochastic models. From this casual observation, perhaps professionals put more stock in
accuracy (getting close to the actual number) rather than detail. There is a possibility of
increasing the complexity of a financial model so much that the major fundamentals of the
pro forma are washed out; losing sight of the “forest through the trees”. Thus, it is
important that increasing detail is not pursued at the expense of accuracy. The discussion
herein is science based, but the implementation of these concepts remains an ‘art’ requiring
real estate intuition.

5.1 Predictability in the Real Estate Market

In the 1960’s, a powerful theory emerged called the Efficient Market Hypothesis (EMH)
with contributions from economists such as Paul Samuelson and Eugene Fama (Malkiel &
Fama, 1970; Samuelson, 1965). Their work explains how the stock market is so efficient
and quick to adapt to new information that it is impossible to predict where the market is
going, since future information is unpredictable. By extension, the stock market should
behave as a complete random walk (Fama, 1995). Looking at historic trends is futile
because future information occurs independently from what has already happened.
Exchange traded funds (ETFs), which are funds which try to replicate the entire market,
were created to make use of the EMH mantra, “active management of funds won’t help”. In
fact, Fama & French (2010) show that 65% of actively managed high fee mutual funds did
not beat passively managed ETFs.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 28 
 
In recent years, the EMH has been challenged by research that argues for a level of
predictability existing in the markets. Lo & MacKinlay (1988) used new data to reject the
idea of the markets behaving as a random walk. Jegadeesh & Titman (1993) discovered
that momentum in the stock market can lead to above market returns; that is, relying on
the short term tendency for a stock’s price to go up if it was going up the last period. Shiller
(1990) proposed that a mean‐reverting behavior exists in the stock market due to investor
irrationality. Alas, some level of predictability exists which can be used advantageously
which keep financial analysts like the author of this thesis employed.

While weakened from modern empirics and the 2008 financial crisis, the EMH still affects
the way investors’ model future returns by cautioning market participants about how
difficult it is to earn above‐market returns. In addition, the EMH highlights the important
role which uncertainty plays in the market.

Do these theories primarily focused on the stock market translate over to real estate? Yes
and no. For variables in the office space market such as rent and vacancy, a level of
predictability does exist due to patterns in momentum and cyclicality which are discussed
in greater detail later in this chapter. In real estate asset markets, the EMH holds less
weight compared to stock exchanges, because the cash flows of real estate asset (which are
dependent on the space market) are fairly predictable. In general the EMH suffers from a
lack of applicability to real estate because of the heterogeneity of real estate, the lack of
public sales information, and time lags in the transaction process. On the other hand, Real
Estate Investment Trusts (REITS) can behave similarly to the rest of the public capital
markets. Generally, the more efficient a market is at integrating new information in asset
prices, the less predictable it is. The predictive nature could even become endogenous to
the price of an asset in a very efficient market. For example, when a new technique is
developed and proven to be capable of making above‐market risk adjusted returns,
everybody will immediately copy the technique which becomes the new standard. The
main takeaway from this section is that most real estate markets behave with both
predictability and randomness at the same time and this should be reflected in the way
financial models are created.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 29 
 
5.2 Real Estate Economics and the Stock Flow Model for Office Properties

The mechanics of how real estate prices move over time has been studied extensively and
can be described using a Stock Flow Model. A Stock Flow model is simply any model which
describes the process of how a durable stock of goods, such as real estate, increases and
decreases and interacts over time with the flow of usage (i.e. leasing) of that stock of goods.

To start off, the Stock Flow Model draws on the familiar concept of supply and demand,
with a few quirks, to correctly describe what occurs in real estate. Employment is the
driver for rent of office space on the demand side. On the Supply side, the stock of office
space is the main determining factor. The stock of office space is completely inelastic in the
short‐term because office buildings take time to build, When demand (employment)
increases, the rent must increase because it takes time for new stock to arrive in the form
of new construction. When employment fall, rents will fall by a greater percentage because
of the durability of real estate capital leading to complete inelasticity in supply in the short
run. New real estate stock is gradually introduced into the market to meet demand and
because of this, rents and prices react quickly to changes in the demand, but stock does not.
What triggers new construction? Asset prices – which are a function of rents and cap rates.
Di Pasquale & Wheaton (1996) illustrates these relationships between construction, asset
markets and space markets in the Four Quadrant model.

As the economy goes through its ups and downs, real estate prices and rents go up and
down because demand changes without a quick response from the supply side due to the
durability of real estate and lag to deliver new space. Eventually, increases in rents and
prices promote new construction which gradually alleviates pressure on rents as the new
space is delivered to market. Since there is a time lag in construction, it is rare that the
exact amount of completions comes online and perfectly meets demand; there will be
overbuilding and underbuilding which leads to real estate incurring its own cycle.

The variables required to create a stock flow model can be obtained by using linear
regression techniques on a large result set of reliable historic data. Multiple linear
regression attempts to quantify the relationship between a dependent variable and
multiple independent variables. For example, a regression can be ran between the square

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 30 
 
footage of occupied space and employment, a major driver of office demand. If the
numerical relationship can be predicted, forecasting employment (a source of demand) will
lead to a prediction for occupied space.

In terms of complexity, uncertainty can be incorporated at different levels in a financial

model. Some analysts will prefer to model uncertainty into rents directly, while others will
prefer to add uncertainty to more primal sources such as employment growth and run
these numbers through a stock flow model to arrive at rents. Please refer to Paradkar's
(2013) thesis for a model which incorporates uncertainty using the stock‐flow model.

5.3 Sources of Uncertainty in Real Estate

There are infinite possibilities when it comes to events or shocks that may influence real
estate asset values and rents. Uncertainty can be driven by changes in the macro‐economy
and local economy. Technological innovation such as hydrologic fracking come out of the
blue to effect office markets catering to the energy sector. Transportation infrastructure
changes give rise to winners and losers in real estate. The endless list of potential shocks
need to be simplified into a few sources before they can be quantified!

With the help of recent innovations in real estate indices, 7 important forms of uncertainty
can be quantified: long‐run market trend, long‐run market cycle, market volatility, short‐
run inertia, individual asset specific volatility, individual asset pricing noise, and ‘Black
Swans”.

Long‐run Market Trend: This is the straight line appreciation trend which prevails over the
long term in the real estate asset market. Research into residential real estate trends have
found that over the super‐long term (over the course of a century), prices appreciate close
to the rate of inflation (Eichholtz, 1997). Growth in commercial real estate over the long
haul has been found to be slightly less than inflation because of depreciation (Fisher,
Geltner, & Webb, 1994; Wheaton, Baranski, & Templeton, 2009). With this knowledge,
professionals may be tempted to input 1% or 2% because of the stability the Federal
Reserve Bank offers for inflation in the United States, but keep in mind that investment
horizons for commercial real estate tend to be shorter than 20 years.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 31 
 
Long‐run Market Cycle: The long‐run market cycle is the oscillating nature of real estate
prices that is easily observable on a time‐series graph. The peak‐to‐peak or trough‐to‐
trough timing has been between 15 and 20 years in last few commercial real estate cycles.
Wheaton (1999) explains how there are two ways in which the real estate market could
manifest itself. One view is that real estate developers are completely rational and forward‐
looking while the other view is that developers are backward‐looking (or ‘myopic’) when
forecasting future supply and demand. When agents are rational and forward looking, they
have a good understanding of how the market behaves with uncertainty, so prices reflect
the present value of future cash flows and the uncertainty surrounding the cash flows. A
practice which would classify as “myopic” behavior include extrapolating average historic
rates forward in financial models. In Wheaton’s simulations, he finds that both cases still
(myopic or forward‐looking) generate endogenous long‐run cycles within real estate as
developers struggle to forecast the exact amount of space to build.

Market Volatility: Zooming in a little bit from the Long‐run Market Cycle level, there is
volatility which exists month‐to‐month and year‐to‐year along the cycle preventing a
smooth oscillating curve. Events that can influence this type of uncertainty include natural
disasters or announcements by central banks. Any new discovery of information that
provides a shock that the market takes time to adjust to are uncertainties related to market
volatility.

Short‐run inertia: Also called momentum, this is the tendency for prices that are rising to
want to keep rising – or falling prices to keep falling. To measure inertia, auto‐regressive
techniques are employed which measures the level of influence a previous period’s price
movement has on the current period’s price change. If the relationship is high between the
prices for the two periods, it means that people are using the current period’s price as a
basis to forecast future periods. It is interesting to note that inertia is very weak in data
involving REITs because of how efficient securities markets are. The frictions in private real
estate markets, however, allow momentum to occur.

Individual Asset Volatility: If a financial model focuses in on a particular property, the model
will be subject to idiosyncratic asset volatility, that is, risk which is specific to an individual

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 32 
 
asset that doesn’t apply to the rest of the market. For example, a municipality might
announce a new rapid transit station to be constructed next to an office tower which
created an unexpected boost in the office property’s value.

Individual Asset Pricing Noise: In the private real estate market, every professional will have
differing opinions on the value of a property. If polled, the opinions of value for thousands
of real estate experts might be scattered around the ‘most likely’ value, with a greater
spread for more unique properties and less of a spread for properties with multiple
comparables. Noise is the effect that these differing opinions have on the pricing of real
estate. Appraisers take into account noise when they provide a range of prices that they
believe a specific property can sell for.

Black Swans: In defining what Black Swans events are, Taleb(2007) states: “first, it is
an outlier, as it lies outside the realm of regular expectations, because nothing in the past
can convincingly point to its possibility. Second, it carries an extreme impact. Third, in spite
of its outlier status, human nature makes us concoct explanations for its
occurrence after the fact, making it explainable and predictable.” Any event with major
impact on real estate values encompasses this risk. For example, a new renewable energy
source (making combustion engines obsolete) suddenly discovered in a lab at MIT could
have major “Black Swan” type ramifications for real estate.

Each type of uncertainty described above can be quantified on their own. Then a Monte
Carlo Simulation outputs the effect of uncertainty as a whole on a real estate project.
Modeling the effect of 7 types of uncertainty together without a Monte Carlo Simulation
would be practically impossible.

5.4 Real Estate Indices

Real Estate indices provide some of the data from which the 7 forms of uncertainty can be
extracted. There are many choices with regards to real estate indices, with each having
their own advantages and disadvantages. There are three major types of real estate indices
in the United States: appraisal‐based, transactions‐based, and stock market based (Geltner,
2014).

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 33 
 
Appraisal‐based indices use independent professional appraisals of properties to track real
estate markets. They were the earliest forms of indices in real estate, so they tend to have
longer histories. The major drawback with appraisal‐based indices is that they are
susceptible to a phenomenon called appraisal smoothing. Appraisers tend to use empirical
information such as sales comparable to determine the current value of properties which
develops a lag in their estimated value. This lag contributes to a smoothing effect on the
index which reduces the apparent systematic risk in the real estate returns.

Transactions‐based indices (TBI) use actual sales data of commercial real estate to track
the market. To accomplish this, many of these indices monitor pairs of sales on properties
to ensure that the changes reported are from an apples‐to‐apples comparison. TBIs are
relatively new with data only stretching back to 2000 but they hold great promise because
the underlying transaction price data not only quantifies market volatility reflected in the
indices themselves, but also quantify individual asset idiosyncratic uncertainty using the
residuals of the price regressions.

Stock market‐based indices track the movement of publicly traded real estate investment
trusts. Because each REIT generally specializes in one property type or another, they can be
a great source of data when looking at a particular geographical area or industry. Keep in
mind that REIT values do not perform the same as private real estate all the time. The
efficiency of the stock market eliminates much of the inertia that would exist in the private
market. In addition, there is evidence that the REIT market slightly leads the private real
estate market (Barkham & Geltner, 1995).

5.4 Translating Data on Uncertainty into a Pro Forma

Quantifying uncertainty can be done in many of ways on Excel, but an emphasis should be
placed on clarity and transparency as there are many moving parts to a stochastic pro
forma. Chapter 6 runs through a case study in which the research presented in this chapter
can be implemented in Excel. There are many modifications that can be made to pro forma
for the case study in Chapter 6 and some of these possible variations are discussed below.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 34 
 
In the ModerateCo example in Chapter 3, a normal distribution is used to disperse the
randomness around a mean, but there are other common methods for distributing
uncertainty.

Uniform Distributions: The RAND function in Excel functions as a uniform distribution on its
own. Every number between 0 and 1 has the same chance of appearing. If an analyst wants
to model a random change in price next year between ‐10% and +10%, with every value
having an equal chance of appearing, they can use the function = (RAND( )/5)‐0.1. The
divisor of 5 creates a 0% to 20% range, while subtracting 10% shifts the distribution down
to create a ‐10% and +10% bound.

Normal (Gaussian) Distributions: In the ModerateCo and ChallengeCo examples, a normal

distribution was used to disperse random variables. Normal distributions are familiar with
most professionals with college degrees because these distributions are common place in
introductory statistics courses. Normal distributions are often observed and nature and
play an important role in science and business. In the standard normal distribution,
sometimes referred to as a “bell curve”, only two unknowns are required to create a curve:
mean and standard deviation. The mean is the average value of all the random numbers in
the distribution, and in a symmetrical normal distribution, the mean will lie directly in the
middle of the curve. The standard deviation (denoted by σ) is a measure of the spread of
the distribution. The larger the standard deviation, the wider the range of numbers will be.
In a standard normal distribution an empirical rule exists that states: 68% of values fall
within one standard deviation from the mean, 95% within 2 standard deviations, and
99.7% within 3 standard deviations. This rule can also be referred to as the 68‐85‐99 rule
or the 3 sigma rule. When setting up a normal distribution of random variables, keeping the
3 sigma rule in mind will help “fit” a distribution to observed volatility in an index. Excel
has numerous function which relate to normal distributions but for modeling uncertainty,
NORM.INV is the most used. This handy function fetches the number at a certain
cumulative probability of a standard normal curve with a mean and standard deviation that
a user can specify. For example, if the normal curve mean was 3% with a standard
deviation of 1, a probability of 50% in the NORM.INV function will fetch a 3%. In the

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 35 
 
probability turned out to be 15.9%, the NORM.INV function will fetch a 2 (since a
cumulative probability of 15.9% ends up a ‐1 σ).

Triangular Distributions: Sometimes the range and most likely number for a random
variable is known. For these problems where there is not much information available, a
triangular distribution can be used. A major benefit to triangular distributions is that the
bounds are limited, compared to the theoretically infinite bounds of normal distributions.
Triangular distributions are commonly used in business because not much information is
required, yet allows for a “best guess” as the most likely number (or the mode). The mode
does not need to be at the median between the two bounds, but if isn’t, it becomes more
difficult to model in Excel. For cases where the triangular distribution is not symmetrical, it
is suggested to use an Excel add‐in, such as @RISK software, to simplify formulas using
their framework. As for symmetrical triangular distribution, imputing =rand()+rand()‐1 in
to excel will model a triangular distribution between ‐1 and 1, centered around 0. To move
the center and mode of the distribution, add or subtract values. For example,
=[rand()+rand()]+19 will model a distribution between 19 and 21, with 20 in the middle.
To expand the bounds, multiply the RAND+RAND expression with a desired factor. For
example, =4*[rand()+rand()] will yield a distribution between 0 and 8 centered around 4.

Other distributions are also possible, but it is suggested that a program such as @RISK
software is used to keep formulas from becoming untidy. Long, elaborate formulas
decrease transparency and increase difficulties when troubleshooting.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 36 
 
CHAPTER 6: Managing Uncertainty in the Real World

Quantifying the uncertainty in the inputs of a financial model was the focus in chapter 5.
Now the focus will shift to a discussion on methods that manage uncertainty. The word
“option” is frequently used to describe an alternative or a choice in everyday speech. The
definition for an option used in this thesis is more specifically defined as “a right, but not
the obligation, to buy (or sell) an asset under specified terms” (Luenberger, 1998). This
chapter sheds light on how financial options generate value and discusses the applicability
of financial option analysis to improve the financial performance of real estate projects.

6.1 The Basics of Financial Options

Options are a class of financial instruments widely known as derivatives. Derivatives are
aptly named because their values are derived from other assets. Options can be conceived
for stocks, bonds, commodities, foreign currency and other assets. In essence, options are
contracts acquired at a cost that allow a party the right to purchase or sell an asset, without
obligation, usually at a specified time and at a predetermined price. Just like the assets
which these “contracts” are dependent on, options can be traded in private or on public
derivative markets, such as the Chicago Board Options Exchange.

Two major types of financial options exist: call options and put options. Call options offer a
party the right to purchase an asset for a predetermined price. Put options offer a party the
right to sell an asset for a predetermined price. This predetermined price is called the strike
price or exercise price.

Here is a scenario that demonstrates how a call option works:

Prospero Mining Company’s stock price is $100 today and is undergoing an important
geological study at one of their prospective mining sites in Canada. Portia, the rich savvy
investor, only wants to invest in Prospero if the geological study finds gold; it would be
disastrous for the company’s stock price if gold is not found. However, Portia is also afraid
that she might lose out if gold is found because the stock price of the Prospero Mining
Company has the potential to double or triple! Portia’s solution is to purchase a call option
for the Prospero stock for $5 (known as the option premium) at an exercise price of $110.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 37 
 
For $5, Portia gains a right to buy Prospero Mining Company’s stock at $110 sometime in
the future. If Prospero Mining Company’s geological study turns out positive, the stock
price doubles, but Portia maintains the right to buy the stock at $110.

The mechanism for put options works the same as call options, except it is now a right to
sell instead of buy. For example:

Antonio is a wheat farmer and he’s worried that the market price of wheat may fall from its
current price of $10 a bushel. If the price falls below $8, Antonio will not have enough
income to get by this year and would need to sell his farm. Antonio buys put options from
Claudius for an option premium of $1 per bushel with a strike price of $9. No matter what
happens to the wheat price, Antonio will be able to survive because he has hedged his
downside risk. If the price of wheat increases, Antonio will not exercise the option. If the
price of wheat falls dramatically, Antonio is safe because of the put option.

To simplify the scenarios, the duration that an option is valid for was not discussed in
Portia’s or Antonio’s example above. In reality, options vary in their exercise terms and
expiration dates. The two most common types of options are American and European
options. In typical American options, the holder of the option can exercise the option at any
time before the expiration date; if an option is good for a year, the option holder can
exercise it anytime within a year. In European options, the option holder can only exercise
the option at the expiration date. Thus, the option holder of a European option has much
less flexibility in exercising the option. Other exotic option types exist, but the vast majority
of options are sold in an American or European style.

6.2 Sources of Value for Financial Options

Options provide risk mitigation by effectively operating as insurance for more costly assets.
In the section 6.1 examples, Portia and Antonio were able to change their exposure to risk
by purchasing options. The future may lead to positive or negative outcomes, but options
allow investors to hedge against the risk of negative outcomes. There is no doubt that
options can be very valuable, but what actually generates this value in options?
Fundamentally, there are two drivers of value for an option: time and uncertainty.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 38 
 
In American options, the value increases as the expiration date is further into the future
because it provides the option buyer more flexibility in the exercise timing. The longer the
duration of the option, the greater the chances of the option present in a “in the money”
state which increases the value of the option.

The value of a European option depends on the market prediction of what the environment
will be like at the exercise date, since European options can only be exercised at one
forward date and not before. All else being equal, the further into the future an exercise
date is for a European option, the greater the uncertainty; leading to a higher option
premium.

Options are only relevant because of uncertainty. In a world void of uncertainty, no one
would buy or sell options because market participants would simply chose to buy or not
buy the underlying asset with complete knowledge of what their investment return would
be. An option’s function is to protect an investor from risk, but with no risk to hedge
against, there is no need for options. It is interesting to note that, as the level of uncertainty
increases, the value of an option increases as well. Here, the uncertainty can be thought of
in two components: the possibility of loss, and the magnitude of the potential loss.

The higher the probability of an option being exercised, the higher the option premium will
be priced at by option writers. If there is near certainty that an option will be exercised,
option writers will price their option very close to the strike price to ensure that there is a
fair deal for both the buyer and seller in a competitive market.

The magnitude of loss relates to the volatility of an underlying asset’s value. Assets with
large price swings will not only have a higher possibility of being exercised, but also have
the potential to overshoot the exercise price by a greater margin creating a larger loss for
the option seller and a larger gain for the option buyer. While the purchasers of options are
protected with a right but not an obligation, the writers of options are obligated to deliver
on their contracts, exposing themselves to risk. In a competitive market (see the discussion
on efficient markets in section 5.1), this risk will be priced ex ante into an option with
available information, leaving little room for above‐average risk adjusted returns.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 39 
 
6.3 The Valuation of Financial Options

Like other assets traded on a public exchanges, options are subject to competitive market
dynamics, where the market collectively determines the price. How do market participants
value options? There are three major techniques used to value options: Black‐Scholes
(mathematical) models, Binomial Pricing Model Models, and Monte Carlo Simulations.

Black and Scholes (1973) introduced the famous Black‐Scholes formula to calculate the
price of European Style Options using 5 known variables: strike price, stock price, time,
volatility, and risk free rate. The model works on the assumption that options will be
priced correctly in the market ‐‐ arbitrage opportunities (using replicating portfolios)
quickly bring option prices back in line. In essence, Black and Scholes used this assumption
to derive an equation for valuing European options. Unfortunately, the Black‐Scholes
formula is tremendously cumbersome to work with for American Options because of the
possibility of exercise before the expiration date of an option.

First proposed by Cox, Ross, & Rubinstein (1979), Binomial Tree Models quickly became a
favorite among analysts trying to model American Options because of the model’s intuitive
simplicity. The Binomial Tree model is created by formulating different scenarios which
could occur to an underlying asset over time and recording them into a lattice structure.
Each level of the tree represents a period of time, with two routes (up or down routes)
available for the asset’s price to follow at each node (Veronesi, 2010). Probabilities are
assigned to each path, but normally, analysts define each branch as having a 50% chance of
realization to simplify the model. A new asset price is assigned for each branch in the tree,
leaving a visual representation of a varying price of an underlying asset over time. Based on
the forecasted values (at discrete time intervals), the option premium is calculated starting
from the end values and gradually computing the option values all the way back to the
present.

A major limitation of the Binomial Tree Pricing Model is the inability to incorporate
multiple sources of uncertainty. All uncertainty must factored into the price and
probabilities which the model operator employs at each node. To overcome these
restrictions, analysts have started to harness the power of computing technology to

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 40 
 
simulate thousands of scenarios in a matter of seconds, dwarfing the limited number of
possibilities which can be modeled in a Binomial Tree. Using Monte Carlo Simulations to
model the value of options allows for vast customization of uncertainty. To model option
premiums using Monte Carlo method, an analyst first sets up a model in which an
underlying asset’s price is subject to uncertainty over time. IF statements (see Appendix D)
are used to mimic the logic in option exercise decisions made based on the simulated asset
value. Once the model is in place, many iterations are performed, with the final effect of the
option in each scenario being recorded and analyzed.

A Monte Carlo Simulation acts as a “black‐box” or a short‐cut where difficult partial

difference equations is normally required find an option’s value. Like a stew cooking in a
large pot, an analyst can continue throwing different factors of uncertainty, pay offs,
probabilities and other nuances into the financial model. The computer does all the heavy
lifting with Monte Carlo Simulations! It is no secret that this thesis advocates for the use of
Monte Carlo Methods because of all of its advantages for very little effort.

6.4 Real Options

Like their financial counterparts, a real option is defined simply as a right without
obligation. While financial options are tied to securities such as stocks or bonds, real
options pertain to business decisions and are often, but not always, associated with
tangible assets such as real estate or machinery. Incorporating design flexibility into a real
estate project is an example of real option. The ability to alter the design of a structure to
meet future conditions is a choice which can be made in the future. Yet, there is no
obligation to exercise the option if conditions do not support a change to the structure.

Many principles of financial options translate over to real option analysis, but there are a
few key differences. Real options are more valuable when there is greater uncertainty
looming over business decisions and they tend to run in perpetuity with American style
option exercise terms. Real options are not sold on public options exchanges, so arbitrage
opportunities to correct prices of real options do not exist. Furthermore, real options may
not be derived from anything at all, making each real option very unique to their situation.
With no asset to serve as a basis for its value, real options are not really derivatives; they

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 41 
 
are more like business decisions which can be made in the future. With financial options,
the underlying asset’s price effects the value of the option, but for real options, the factors
that determines value are not always tradable commodities. These factors could be
intangibles such as demand‐based measures like rent or vacancy.

In a real option, the cost to implement flexibility is often independent of economic

conditions. To put it in another way, financial options have the expectations of the future
priced into it by the market, making it difficult to make a profit. Typically, this is not usually
the case with real options because no price correction mechanism exists. For example, let’s
say a small‐cap development firm, Lear Developments, is planning to build a three story
parking garage and wants to embed the flexibility to expand the garage at a future date by
building another three stories on top of the existing structure. The option premium in this
case is the extra construction cost to add the flexibility and the cost to build the three
additional stories. Capulet Construction builds the parking garage and sends Lear an
invoice based on the current labor and building material prices. Capulet Construction does
not care about what profit Lear will earn; they just want to build the parking garage, collect
their fee, and move on to another construction project. There is a great opportunity for
Lear to make a positive NPV on the project by exploiting the disconnect between the
construction cost and the economy; the real option premium is not related to the potential
payoff!

Financial options tend to have simple terms, but real options can involve many interrelated
qualities. Rather than a simple exercise price, real options could have a grand criteria that
needs to be satisfied before it is exercised. In these situations, the Black‐Scholes Model and
Binomial Tree model cannot quantify the value of a real option. The level of complexity
required by real options analysis makes the Monte Carlo Simulation the tool of choice when
dealing with real options.

Financial options are valued by directly inputting unknowns, such as strike price and
exercise period, into equations to arrive at the option premium. Real options are not as
straight forward. Because of their intricacy, it can be near impossible to directly compute
the value of a real option. Monte Carlo simulations offer a work‐a‐round solution by

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 42 
 
allowing analysts to find the NPV of a project without an option and comparing it to the
NPV of the same project with an option in place. Conceptually, the option value would be
the difference between the two NPVs.

The ability to model different real options into a real estate financial model is an option in
itself! The NPV values which are calculated from a real options analysis are never binding.
Real option modeling enjoys asymmetric outcomes; all of the real options that were not
worthwhile are not undertaken, while “home run” options are pursued further. There is
nothing to lose when modeling options, but potentially a lot to gain.

Sometimes, real options are already free to implement. The freedom to walk away from a
property with an ‘underwater’ loan prevents further losses, effectively capping downside
outcomes. In a financial model which incorporates uncertainty, this real option of walking
away from an underwater loan improves expected NPV, yet costs nothing. Therefore,
modeling this real option into a financial model can provide that extra edge that is
difference from winning and losing a competitive land bid.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 43 
 
CHAPTER 7: Two World Trade Center Case Study

We developed a short case study to connect and apply the concepts presented in this thesis to
a realistic situation. The scenario created for this thesis is inspired by an actual case study
done on the World Trade Center site in New York City (Queenan, 2013). While the story is
fictitious, the assumptions are made to be as realistic as possible with basis from legitimate
sources.

7.1 Scenario Background

Wall Street has seen better days. 5 years after one of the deepest recessions in US history,
the blame game is in full flight. Recovery has been excruciatingly slow and the popular
thing for politicians to do is to pick apart Wall Street. No one seems sure about the future
state of the financial sector.

After much difficulty with leasing One World Trade Center and Three World Trade Center
in the thick of the financial
crisis, Goldstein Properties
and The Port Commission are
hoping to unload the
development rights to the 2.3
million square feet office piece
of Two World Trade Center.
Arden Forest Properties is
interested in developing the
blue chip office tower and
brought in Timon Capital to be
the money partner.

Figure 13: World Trade Center Site Plan (PANYNJ, 2013) Not surprisingly, Timon is

Two World Trade Center sits on the most North very concerned about the
Eastern parcel in the site. One WTC, 3 WTC, and 4 future of the office market in
WTC are all office towers.
Manhattan. In an effort to put

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 44 
 
their business partner at ease, Arden Forest creates a pro forma to see how the
development would perform with uncertainty in the market and address the downside
possibilities. Michael Cassio, a senior analyst at Arden Forest, wonders if this is a good
opportunity to use his internship experience last summer at a derivatives desk in Chicago.
Although Michael has only seen the value created by financial option for an investor, he
suggests implementing an option in this development to see what would happen.

Two World Trade Center is part of a

major mixed‐use development consisting
of 5 office skyscrapers plus retail,
cultural, and transportation
infrastructure. Sitting at the North East
corner of the site, Two World Trade
Center will be the second tallest building
of the World Trade Center complex. As
with all the other WTC sites, The Port
Commission will look after the
construction of the foundations because
of the intricate network of tunnels below
connecting to new WTC transportation
hub south of the 2WTC site.

Goldstein has already committed to

developing the retail podium at the foot
of 2WTC, but now they going through a
bid process for the development rights
to the office portion of the property.
Figure 14: 2WTC Rendering (PANYNNJ, 2013)
A rendering of 2WTC as viewed from the Michael Cassio knows that this will be a
9/11 Memorial. very competitive bid for a rare world‐
class property. The dollars at stake are much higher with a landmark skyscraper in the
world’s most prominent financial center. The slightest miscalculation on the pro forma can

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 45 
 
cost Arden Forest billions. Michael hopes that his secret weapon, real options analysis, will
help Arden Forest win the site without exposing them to risk without compensation.

7.2 Creating a Detailed Stochastic Pro Forma for 2 World Trade Center
Michael begins with a big picture strategy for the pro forma: quantify uncertainty in the
office market, create a real options model, and evaluate the pro forma by performing a
Monte Carlo Simulation. Michael understands that office employment growth is the main
driver of real estate demand, but there is simply not enough time to compile the data he
needs for the bid that needs to be submitted in just a few days. Plus, he’s not comfortable
creating a real estate stock‐flow model because he didn’t read section 5.2 of this thesis or
Sarwesh Paradkar’s award winning MSRED thesis. Mr. Paradkar shows how the stock flow
model can be implemented with a little extra data on employment, vacancy, and real estate
stock (Paradkar, 2013). As far as uncertainty goes, Michael decides to directly forecast
office rents.

7.3 Projecting Rents, Cap Rates, Construction Costs and Operating Expenses

Initial Rent

First off, Michael needs to know what the average office lease would go for in 2 WTC if it
were leasing today. Luckily, there are many lease comparables within the same complex!
One World Trade Center and 4 World Trade Center are asking for $75 per square for in
gross rent for space despite an abundance of vacant space in the Downtown submarket
(Levitt, 2013). Not all is lost as the entrance of One World Trade Center to the market has
gradually increases office rents downtown to an average of $60 per square foot (Kozel,
2013). Michael decides that $65 per square foot is a reasonable rent for 2 WTC, since it will
be a brand new Class A building. On the other hand, Michael doesn’t want to put rents near
$75 per square foot because 2 WTC needs to entice tenants away from other competing
WTC office towers.

Leading with $65 per square foot, we can follow along with Michael’s work in the
‘Projections’ tab of the 2 World Trade Center pro forma. We will gradually add layers of
uncertainty to our rent forecasting.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 46 
 
Long‐Run Trend

This was an easy one for Michael. The long‐run trend for office properties has been well
researched in terms of asset price and rents. Office properties don’t beat the rate of
inflation over the long‐run (Eichholtz, 1997; Fisher et al., 1994; Wheaton et al., 2009). In
fact, real estate will do slightly worse because of depreciation in the property. Since the US
Federal Bank has a stated goal to keep inflations at around 2%, Michael chooses to use 2%
as the long‐run prevailing trend and varies it using a normal distribution. By setting a half‐
range of 1%, Michael uses 1% divided by 3 as the standard deviation in the NORM.INV
function, which makes it a 99.7% probability that the long‐run trend will lie between 1%
and 3%. The initial rate now exhibits long‐run trending by adding the percentage increase
year after year.

Market Volatility

For market volatility, Michael uses gross rent data to find the historic volatility. In the
assumptions Excel file, there is a sheet called volatility which details how to find the
volatility of rents. In this case, we have quarterly data of rents and we first covert the rents
into a percentage change from quarter to quarter. Next, the standard deviation is found on
the quarterly changes using the ST.DEV function. Lastly, we translate the quarterly
volatility to an annualized number but multiplying the quarterly volatility by the square
root of the number of periods. In this case, we multiple the quarterly standard deviation by
the square root of 4 to get an annualized number. Of course, the volatility could be positive
or negative in any given year, so Michael uses the NORM.S.INV function. This function uses
a cumulative distribution function and translates a random variable to go either positive or
negative around a normal distribution. A random number of .5 will make the factor 0, while
and cumulative probability of 16% (roughly at ‐1 standard deviation) will end up with a
factor of ‐1 and so on. Calculating the standard deviation on the projected volatilities
should return a number close to the historical 7%.

Inertia
For momentum, we use the gross rent data and take the first difference of it, which is the
rate of change from year to year. A linear regression was performed with a lagged

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 47 
 
percentage change in rents to arrive at an inertia factor. Please refer to the linear
regression performed in the AR tab of the “2 WTC assumptions” to see how the rate of
33.5% for inertia was retrieved. If the market volatility was negative last term, the current
period will have some downward pressure from momentum.

Market Cyclicality

Peak‐to‐peak, real estate cycles have lasted between 15 and 20 years (Geltner, 2013).
While not always synced, the real estate and asset and space markets appear to have
similar durations. To model this cyclical effect, Michael uses a sine curve to create a factor
for rents each year. The coefficient in front of a sine curve affects the amplitude, or the
height of the waves, and the numbers inside the sine function affect the duration and
position of the curve. The Sine curve on its own has a cycle duration of 2π and amplitude
range of 1 to ‐1. So, Michael translates the Sine curve to work in the spreadsheet by using
the formula:
2
sin 1
2
Where: Maximum amplitude in %
Number of years since start year, with start year = 0
Cycle Starting Position (years after upward mid‐point)
Duration of one full cycles in years

The +1 at the end of the function shifts the entire sine cover up to have a mid‐point of 1
instead of 0.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 48 
 
 For amplitude,
Michael does a quick
analysis of NYC office
rents between 1988
and 2011. First, he
converts the nominal
rents to real rents,
then he observes the
difference between
lowest and highest
values with the total
change from peak to
Figure 15: Real Estate Cycle Length
The peak‐to‐peak and trough‐to‐trough duration is between trough observed to be
15 and 20 years (Geltner, 2013). close to 40%. Lastly,
Michael looks at the Moody’s/RCA CPPI TBI to see where the real estate environment is in
the cycle. It seems like it should be around half way on to the peak in 2013, but Michael
decides to retard the cycle in the analysis a bit because economic recovery has been slower
than usual.

With all the

parameters accounted
for, the Sine wave is
converted to a factor
with the current year
as a factor of 1.

Figure 16: The Regular Sine Curve

When not transformed, the sine curve has a cycle duration of
2π and an amplitude of 1. The cycle’s y‐intercept is 0.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 49 
 
Noise
Noise is taken in to account symmetrically with a 10% range normal distribution. 10% is a
typical range used by appraisers when providing their opinion of value.

Black Swan

Black swans are by definition, impossible to predict but we can still simulate the effect.
Michael gives each year a 5% chance to occur, which gives about a 40% chance of a black
swan event occurring every 10 years. The magnitude of impact is set at ‐25%, more than
half of the cycle effect occurring one year seems appropriate for a meaning event that will
affect the investment.

Now that rents are modeled, Michael turns to modeling other items which need to be
projected forward.

Cap Rate

Projecting future cap rates is important because it greatly affects our terminal value
calculation and option trigger. Again, looking at RCA data, the cap rate seems to fluctuate
between 8.5% and 5%, giving a midpoint of 6.75%. Since 2 WTC will be a modern blue chip
office tower, Michael decides to set the mean cap rate a little lower at 6.5% with the same
half range of 1.75%. The asset market cycle tends to lead the space market cycle but they
can be out of sync at times. Going for the rule rather than the exception, Michael sets the
position of the cap rate cycle one year in front of the space market cycle.

Operating Expenses and Construction Costs

Michael uses the same method to account for uncertainty in expenses and construction
costs as the ModerateCo example in chapter 3 of this thesis. Operating expenses are set to
grow at around the targeted rate of inflation by the US Federal Reserve Bank, 2%.

For Construction Costs, RSMeans data says that office building construction costs in New
York have risen about 5% in the last year (Carrick, 2013). The hard cost quoted by
RSMeans is $223 per square foot. Assuming that 2 WTC will be a high quality, expensive

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 50 
 
building, Michael uses $300 per square for hard costs. Adding a rule of thumb (30% of hard
cost) for soft costs, brings the total construction cost number to $390 per square foot.

7.4 Pro Formas

Now with our parameters projected in

to the future, Michael is ready to set
up pro formas to test out his real
options hypothesis. He will create 3
separate pro formas and compare
their financial performance under
uncertainty. The first pro forma will
be a 60 floor inflexible case. Another
inflexible case will be modeled, but for
30 floors instead of 60 floors. The last
one will have a real option imbedded
into the design by starting with 30

Figure 17: 2 WTC Sketch up of Alternatives floors and allowing for the flexibility
The building on the left is a model of a 30 to build another 30 floors on top in
floor inflexible design. The middle building
the future.
represents a flexible design option while the
building on the right is the inflexible 60 floor
office tower design. The construction time should be
longer for the 60 floor tower, so
Michael makes sure that the 60 floor tower takes 4 years to build versus the 3 that the 30
floor towers need. The construction costs are discounted at an OCC of 1.6%, with 50 basis
points from the risk premium of construction cash flows and 110 basis points for the risk
free rate. The risk premium for construction cash flows reflect the low systematic risk
involved. Construction costs are usually locked in with a contact and do not move with
capital markets.

The risk free rate is 1.1%, using the 10 year Treasury bill rate, minus 150 basis points to
account for the yield curve effect (Bloomberg Markets, 2014). The cash flow from the tower
during its fully leased, stabilized phase is discounted at 5.1%, reflecting the average risk

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 51 
 
premium of 4.5% for all NCREIF commercial real estate between 1970 and 2010 (Geltner,
2014). Michael haircut the risk premium by 50 basis points because this office tower will
be a blue chip property in one of the most desirable locations in the world. Lastly, the
stabilized NPV will be discounted in the development phase by 7.1% to account for the risk
involved in leasing up a primarily speculative development. Geltner (2014) states that the
risk premium for development phase cash flows typically have a 50‐200 basis point
premium over equivalent properties in the stabilized phase.

Many developers may be tempted to use a single discount rate for the entire project.
However, it is important that each phase with a different level of risk have a different
opportunity cost of capital. These discount rates should also be justifiable through market
evidence because it is the capital markets that determine these rates.

The pro forma models 5 and 10 year leases with rent escalations using if statements. This is
key to our analysis because lease rates will be locked in by their lease terms while the
market can move either way during the lease. Since Michael doesn’t know what length the
leases will be, he has used the RAND function to determine if the leases will be 5 year or 10
year leases, the most common lease lengths in commercial real estate.

The inflexible 30 and 60 floor pro formas do not stray much from a standard pro forma. The
flexible 30 floor pro forma will need to model the real option though.

7.5 Real Option Triggers

To model a real option, two things need to be defined. First off, the trigger conditions need
to be modeled. For the flexible case, the trigger is pulled whenever the addition becomes
profitable. To measure profitability, Michael uses the NPV investment decision rule:

1. Maximize the NPV across all mutually exclusive alternatives;

2. Never choose an alternative that has NPV < 0.

The acquisition cost of the land does not factor into this decision because it is what is called
a “sunk cost”. Sunk costs are costs incurred in the past that cannot be recovered regardless
of future outcomes. So, they should not be considered when making future decisions.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 52 
 
The NPV for each year is easy to calculate using project cap rates and construction costs.
Michael uses an IF statement to act as a switch to signify if construction begins on the
addition or not. Since the addition can only take place once, another IF statement is used to
ensure that no addition has started previously before beginning construction in that year.

The other piece that needs to be defined for a real option is the consequence. In this case,
what happens is that an extra 1.3 million square feet is added two years (to account for
construction) after the year of the construction trigger. New leases have to kick in, and a
10% construction cost premium is added to the construction cost. The NPV is then
calculated the same way as the inflexible pro formas.

7.6 Results

in Millions Inflex 30 Flex 30 Flex 40 Flex 50 Inflex 60 Michael

Expected NPV ($191) $14 $9 $4 $14
compares the
Median ($251) ($145) ($137) ($128) ($95)
Mode ($300) ($500) ($500) ($300) ($300) returns with
Std Deviation $371 $710 $714 $719 $731
Percentiles distributions
Value 5% ($686) ($792) ($847) ($901) ($962)
At 10% ($604) ($708) ($742) ($770) ($808) after performing
Risk 25% ($455) ($543) ($535) ($527) ($502)
a Monte Carlo
Median 50% ($251) ($145) ($137) ($128) ($95)
Value 75% $7 $400 $397 $388 $414 Simulation and
At 90% $295 $979 $963 $948 $977
Gain 95% $509 $1,383 $1,372 $1,349 $1,380 the results are
Figure 18: 2 WTC Financial Model Results intriguing. The
The flexible 30 floor design and the inflexible 60 floor design
outperform the other buildings. Note that the flexible 30 floor design first thing that
has the lowest potential losses, yet maintains good gains when the pops out is how
economy is good.
the inflexible 30
floor model’s financial performance is terrible. Also unfortunate, the real option did not
have as much of an effect on expected NPV as Michael hoped. The Flexible 30 floor option
has the same expected NPV as building the entire 60 floors outright. However, not all is lost
because the flexible design is doing its job, protecting Arden Forest Properties when the
economy is doing poorly. The cumulative distribution shows how the flexible design
behaves like the 30 floor inflexible project at the low end of possibilities, but then starts to

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 53 
 
behave like the 60 floor inflexible project at the higher end. The flexible 30 floor option is
definitely the most preferable option.

If there was no
construction cost
premium, the
flexible 30 floor
option actually out
performs even at
the top end in the
same environmental
conditions as the
inflexible 60 floor
option. During these Figure 19: 2 WTC Distribution Function
conditions, the real The flexible 30 floor design’s CDF performs like the inflexible
60 floor design, except at the low end of NPVs where the
option is almost downside is minimized.
always exercised
right away in the first year possible, 2017. Despite taking an extra year to construct 60
floors, the flexible option comes on out above the inflexible option because of the lease
timings. The economy reaches the peak of the cycle in 2019, right when the leases from the
new addition come online, while all of the leases of the inflexible schemes are stuck at
lower rates signed 2 years previously. Those poor performing leases will also be renewed
at another low point in the cycle 10 years afterwards in the lowest point of the real estate
cycle.

While the flexible 30 floor model outperforms the other schemes, it does so under specific
conditions modeled by the Analyst. Regardless, Michael is convinced that real option
analysis will help Arden Forest win the bid, arming the company with knowledge of
distributions will help the company focus on managing uncertainty rather than relying on
guesswork and gut feelings in deterministic financial modeling of Real Estate.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 54 
 
CHAPTER 8: Conclusion

“The more you know, the more you know that you do not know” is a commonly used
maxim about Socratic Ignorance. It seems that uncertainty persists more today than ever
before. Perhaps the human race is just learning more about uncertainty through humbling
events such as the mega‐recession in 2008. One thing that we can be assured of is that
uncertainty will continue to play a role in real estate investing whether professionals
acknowledge uncertainty or not.

There is little doubt that real estate firms and investors would like to incorporate the
techniques presented in this thesis. The unfamiliarity with stochastic methods is the main
stumbling block and is understandable when millions of dollars are on the line with each
real estate project. At the same time, it is the fact that a lot of dollars are at stake which
makes these techniques important. Engineers and scientists have relied on Monte Carlo
Simulations to build atomic bombs, fly to the moon, and save lives during pandemics for
almost a century now. Surely stochastic methods will add value in real estate if used
properly.

In this thesis, it was shown how simple it can be to add uncertainty into otherwise
deterministic pro formas which every real estate professional is familiar with. The
unpleasant effect of Jensen’s inequality on return measures in deterministic models is
exposed. Monte Carlo Simulations were demystified in under 5 minutes using a few
keystrokes in Excel. The value of distributions over point estimate was displayed, which
gave an entirely new perspective on financial returns. Real Options was the tool presented
that enables a real estate venture take advantage of uncertainty.

In Chapters 5 and 6, these concepts were further detailed with solid theory and empirical
evidence. Uncertainty in real estate was broken down and explained, using new data tools
and indices to quantify volatility and risk. The academic underpinning of real options was
presented with theory borrowed from financial options. Then we revealed the mechanics
of real options in the context of real estate.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 55 
 
To help translate theory in to practice, a modern example of 2 World Trade Center was
presented. For simplicity, only one out of a variety of options was employed in the analysis,
but the power of real options in real estate was clearly on display. While it is not a
guarantee that real options in a real estate venture will increase value monetarily, the act of
analyzing real options is a valuable option in its own right for real estate where irreversible
investment decisions are made frequently.

Perhaps the greatest advantage of understanding these concepts is the change in mindset
when it comes to approaching real estate problems. A new grand avenue is opened up
when uncertainty becomes part of the analysis. There are endless possibilities with real
option analyses and creative problem solvers will be the greatest benefactors.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 56 
 
BIBLIOGRAPHY

Barkham, R., & Geltner, D. (1995). Price Discovery in American and British Property
Markets. Real Estate Economics, 23(1), 21–44. doi:10.1111/1540‐6229.00656

Baroni, M., Barthélémy, F., & Mokrane, M. (2006). Monte Carlo Simulations Versus DCF in
Real Estate Portfolio Valuation. Groupe ESSEC. Retrieved from
http://books.google.com/books?id=QU2emgEACAAJ

Baumeister, R. F., Bratslavsky, E., Finkenauer, C., & Vohs, K. D. (2001). Bad is stronger than
good. Review of General Psychology, 5(4), 323.

Black, F., & Scholes, M. (1973). The Pricing of Options and Corporate Liabilities. Journal of
Political Economy, 81(3), 637–654. doi:10.2307/1831029

Bloomberg Markets. (2014). United States Government Bonds. Retrieved January 30, 2014,
from http://www.bloomberg.com/markets/rates‐bonds/government‐bonds/us/

Byrne, P. (1996). Risk, Uncertainty and Decision‐Making in Property Development. Spon.

Retrieved from http://books.google.com/books?id=DpMhmQEACAAJ

Carrick, A. (2013). RSMeans’ dollar‐per‐square‐foot construction costs: four types of office

buildings. RSMeans. Retrieved January 25, 2014, from
http://www.reedconstructiondata.com/market‐intelligence/articles/rsmeans‐dollar‐
per‐square‐foot‐construction‐costs‐four‐types‐of‐office‐buil

Cox, J. C., Ross, S. A., & Rubinstein, M. (1979). Option pricing: A simplified approach. Journal
of Financial Economics, 7(3), 229–263.

De Neufville, R., & Scholtes, S. (2011). Flexibility in Engineering Design. MIT Press. Retrieved
from http://books.google.com/books?id=pKjnnqilr3EC

Di Pasquale, D., & Wheaton, W. C. (1996). Urban economics and real estate markets. Prentice
Hall PTR. Retrieved from http://books.google.com/books?id=bIFPAAAAMAAJ

Eichholtz, P. M. A. (1997). A Long Run House Price Index: The Herengracht Index, 1628–
1973. Real Estate Economics, 25(2), 175–192. doi:10.1111/1540‐6229.00711

Fama, E. F. (1995). Random walks in stock market prices. Financial Analysts Journal, 75–80.

Fama, E. F., & French, K. (2010). Luck versus Skill in the Cross‐Section of Mutual Fund
Returns. The Journal of Finance, 65(5), 1915–1947. doi:10.1111/j.1540‐
6261.2010.01598.x

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 57 
 
Farragher, E., & Savage, A. (2008). An Investigation of Real Estate Investment Decision‐
Making Practices. Journal of Real Estate Practice and Education, 11(1), 29–40.
Retrieved from http://ares.metapress.com/content/G26M82G770TK8L72

Fisher, J., Geltner, D., & Webb, R. B. (1994). Value indices of commercial real estate: A
comparison of index construction methods. The Journal of Real Estate Finance and
Economics, 9(2), 137–164. doi:10.1007/BF01099972

Geltner, D. (2013). Quantifying Uncertainty in Real Assets: What the new data tells us, & a
framework for applying it….

Geltner, D. (2014). Commercial real estate analysis & investments / David M. Geltner ... [et
al.]. Mason, OH : Thomson, c2007. Retrieved from
http://libproxy.mit.edu/login?url=http://search.ebscohost.com/login.aspx?direct=tru
e&db=cat00916a&AN=mit.001400326&site=eds‐live

Jegadeesh, N., & Titman, S. (1993). Returns to Buying Winners and Selling Losers:
Implications for Stock Market Efficiency. The Journal of Finance, 48(1), 65–91.
doi:10.2307/2328882

Kozel, P. P. (2013). Colliers Manhattan Office Report Q4 2013. Retrieved from

http://www.colliers.com/‐/media/E8D7EE501D8E419DA06C7EC4AF6F9E0B.ashx

Krathwohl, D. R. (2002). A revision of Bloom’s taxonomy: An overview. Theory into Practice,

41(4), 212–218.

Lang, J. (2008). On Course: a week‐by‐week guide to your first semester of college teaching.
Harvard University Press.

Levitt, D. M. (2013). NYC’s World Trade Tower Opens 40% Empty in Revival.
Bloomberg.com. Retrieved January 30, 2014, from
http://www.bloomberg.com/news/2013‐11‐12/world‐trade‐center‐tower‐debuts‐in‐
manhattan‐leasing‐test.html

Lo, A. W., & MacKinlay, A. C. (1988). Stock Market Prices do not Follow Random Walks:
Evidence from a Simple Specification Test. The Review of Financial Studies, 1(1), 41–66.
doi:10.2307/2962126

Louargand, M. . (1992). A Survey of Pension Fund Real Estate Portfolio Risk Management
Practices. The Journal of Real Estate Research, 7(4), 361–373.

Luenberger, D. (1998). Investment Science.

Malkiel, B. G., & Fama, E. F. (1970). Efficient Capital Markets: A Review Of Theory And
Empirical Work*. The Journal of Finance, 25(2), 383–417.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 58 
 
Marshall, K., & Kennedy, C. (1992). Development Valuation Techniques. Journal of Property.
Valuation and Investment, 4(11), 57–66. Retrieved from
http://books.google.com/books?id=BP1eAAAACAAJ

PANYNJ. (2013). WTC Site Plan. panynj.gov. Retrieved January 30, 2014, from
http://www.panynj.gov/wtcprogress/wtc‐site‐plan.html

Paradkar, S. (2013). Risk adjusted asset valuation using a probabilistic approach with
optimized asking rents and resale timing options. Massachusetts Institute of
Technology.

Pearson, K. (1905). The Problem of the Random Walk. Nature, 72(294).

Queenan, C. (2013). World Trade Center Phasing Case Study.

Samuelson, P. A. (1965). Proof that properly anticipated prices fluctuate randomly.

Industrial Management Review, 6(2).

Sarter, N. B., & Woods, D. D. (1994). Pilot Interaction With Cockpit Automation II: An
Experimental Study of Pilots’ Model and Awareness of the Flight Management System.
The International Journal of Aviation Psychology, 4(1), 1–28.
doi:10.1207/s15327108ijap0401_1

Savage, S. L., Danziger, J., & Markowitz, H. M. (2009). The Flaw of Averages: Why We
Underestimate Risk in the Face of Uncertainty. Wiley. Retrieved from
http://books.google.com/books?id=2lsLAQi0LlcC

Seymour, E. (2008). Talking About Leaving: Why Undergraduates Leave The Sciences.
Westview Press. Retrieved from http://books.google.com/books?id=e3h48xnglrUC

Shiller, R. J. (1990). Market Volatility and Investor Behavior. The American Economic
Review, 80(2), 58–62 CR – Copyright &#169; 1990 American Economi.
doi:10.2307/2006543

Taleb, N. N. (2007). “The Black Swan: The Impact of the Highly Improbable.” New York
Times. Retrieved from http://www.nytimes.com/2007/04/22/books/chapters/0422‐
1st‐tale.html?_r=0

Veronesi, P. (2010). Fixed Income Securities: Valuation, Risk, and Risk Management. Wiley.
Retrieved from http://books.google.com/books?id=vXCc4oWg_ssC

Wheaton, W. C. (1999). Real estate “cycles”: some fundamentals. Real Estate Economics,
27(2), 209–230.

Wheaton, W. C., Baranski, M. S., & Templeton, C. A. (2009). 100 years of commercial real
estate prices in Manhattan. Real Estate Economics, 37(1), 69–83.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 59 
 
APPENDIX 
Appendix A Incorporating Uncertainty into a Financial Model

The RAND function in Excel randomly generates a number between 0 and 1. Each time a
recalculation occurs, the RAND function generates a new number. This function is the
source of uncertainty in the ModerateCo model. Each number in the RAND function has the
same probability of occurrence. For example, 0.1 has the same probability of occurring as
0.9.

To translate this random number to a working rent growth rate another function must be
used in conjunction with the RAND function.

Chapter 5 is dedicated to modeling uncertainty in the real world. For now, a very simplified
method is used to translate the random numbers generated using the RAND function in to
rent growth rates. Instead of every number in the RAND function occurring with equal
chance, extreme numbers or outliers should occur with less probability than numbers in
the “center” or near a long‐run mean.

The NORM.INV function takes in a random probability and translates the number using a
normal distribution function. Recall that 68% of values occur within one standard
deviation from the mean. 95% and 99% of value occur within two and three standard
deviations of the mean respectively.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 60 
 
Nesting a RAND function as the probability input for the NORM.INV function allows the
output value to be normally distributed instead of evenly distributed.

The probability accepted by the NORM.INV function is taken in as a cumulative probability

of a normal distribution. If a 0.5 is given by the RAND function, the NORM.INV function will
output the mean. If 0.023 is given by the RAND function, the NORM.INV function will output
a number two standard deviations below the mean. Any probability between 0.16 and 0.84
will return a value within one standard deviation from the mean. The mean and standard
deviation must be specified to make the NORM.INV function work.

The deterministic SimpleCo example used a 3% rent growth rate. To maintain consistency,
3% is also used as the mean in the ModerateCo rent growth rate formula. As an
assumption, 2% is used was the standard deviation for rent growth rate. Using 2% as our
assumed standard deviation means that growth rate should be within ±2% of the mean (or
between 1% and 5% in our example) 68% of the time and should be within ±4% of the
mean 95% of the time.

(in 000's) Year 1 2 3 4 5 6 7 8 9 10 11
Potential Gross Revenue $4,590 $4,783 $4,985 $5,195 $5,414 $5,642 $5,880 $6,128 $6,386 $6,656 $6,936
Vacancy $230 $239 $249 $260 $271 $282 $294 $306 $319 $333 $347
Effective Gross Revenue $4,361 $4,544 $4,736 $4,935 $5,144 $5,360 $5,586 $5,822 $6,067 $6,323 $6,589

Operating Expenses $2,550 $2,627 $2,705 $2,786 $2,870 $2,956 $3,045 $3,136 $3,230 $3,327 $3,427

Net Operating Income $1,811 $1,918 $2,031 $2,149 $2,273 $2,404 $2,541 $2,686 $2,837 $2,996 $3,162

Capital Expenditures $181 $192 $203 $215 $227 $240 $254 $269 $284 $300

CF From Operations $1,629 $1,726 $1,828 $1,934 $2,046 $2,164 $2,287 $2,417 $2,553 $2,696
Reversion (Purchase and Sale) ‐$17,000 $28,749
PBTCF ‐$17,000 $1,629 $1,726 $1,828 $1,934 $2,046 $2,164 $2,287 $2,417 $2,553 $31,445

NPV $3,019

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 61 
 
Appendix B Performing Monte Carlo Simulations

Our probabilistic ModerateCo pro forma looks exactly the same as the deterministic
SimpleCo pro forma, except that our NPV now changes when we recalculate formulas by
hitting F9. Each recalculation represents a different scenario under uncertainty. While it is
interesting to see the NPV jump around, it would be useful if there was a way to record the
NPV values for many iteration/simulation runs.

This pro forma is now set up for Monte Carlo simulations. Many iterations of the model are
ran and the output values (in our case, NPV) are recorded into a table.

To set up the 2 column simulation table, reference the output value (NPV) in the top row of
the right column.

The next step is to select the entire table and bring up the Data table window from Data ‐>
What if Analysis ‐> Data Table. For the column input select, just select any blank cell in the
spreadsheet and it should populate the rest of the simulations.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 62 
 


In this simplistic example, we ran 10 simulations. The number of simulations which can be
ran is only limited by the processing power of computers.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 63 
 
Appendix C Creating Cumulative Distribution Functions (CDFs) in Excel

Creating a cumulative distribution function of the results from the Monte Carlo Simulation
is simple to do in Excel. The CDF is the chief output result from the pro forma and provides
much more information than a single NPV value.

Once a data table is created to record the results from the simulations, a new table can be
created, sorting the results from smallest to largest using the SMALL function.

The SMALL function selects a number from the result corresponding to the rank specified
as “k”. Thus, rank 1 would refer to the smallest number in the result set, and rank 2 would
refer to the second smallest number in the result set. The sorted NPV values will be the x‐
axis values for the cumulative distribution function. For the y‐axis values, 1/(number of
iterations) is given to each result. Effectively, in 5,000 runs of the model, each result
accounts for a 1/5000 or (0.02%) slice of the distribution. As a cumulative distribution
function, 0.02% is added to each successive result. Graph the table as a scatter plot and
format as necessary.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 64 
 


A few observations can be made about the CDF function of ModerateCo Tower. The greater
the slope of the graph, the greater the likelihood of the corresponding NPVs. In
ModerateCo, there is about a 50% chance of positive NPV and 50% chance of negative NPV,
but the loss and returns are not symmetrical. The loss at the worst 10% of scenarios was at
least ‐$6 million, which the gain at the best 10% of scenarios was at least $7 million. These
numbers are called the Value‐at‐risk (@ 10%) and Value‐at‐gain (@ 90%) numbers.
Another observation one could make is that there is about a 30% chance that the final NPV
will be between ‐$2,000 and $2,000.

In ChallengeCo Tower, multiple CDFs are plotted on the same graph to compare and
analyze the probabilistic outcomes across multiple alternatives.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 65 
 
Appendix D Using IF Statements to Model Real Options for Real Estate Ventures

As presented in section 5.3, The Binomial Lattice method is commonly used to value real
options, but the focus of this appendix will be the Monte Carlo method used in conjunction
with IF statements to model the behavior of Real Options because of the simulation
method’s ability to model several different sources of uncertainty, In the hands of a creative
analyst, a plethora of different situations and circumstances can be modeled with the
flexibility that the Monte Carlo Simulation method offers.

The IF statement is an important logic function in Excel that allows the program to make
decisions automatically for you (because manually making 5,000 switches is madness).
Basically, the IF statement can be used as an automatic switch ‐ in the context of Real
Options, the IF statement allows the spreadsheet to make its own decision on whether to
exercise an option or not.

For ChallengeCo, the option to build 10 more floors sometime in the future needs to be
modeled. When will construction begin for the 10 additional floors? Construction will only
commence if the economy does well; few developers would want to exercise this option
when rents are low and vacancy is high! The first step in modeling real options is to specify
the ‘trigger’ conditions.

ChallengeCo Parameters Year 0 1 2 3
Total Development Cost $100 /gsf
Efficiency 90%
Gross Floor Area (Addition Incl) 170,000 sf 0.00 0.00 0.00
Option Exercise Trigger: Rent $35 /sf
Flex Development Cost $105 /gsf $         108.15 $         110.30 $     111.37
Office Rent $30 /sf $           30.90 $           31.51 $       31.82
Rent Growth Rate 3% 1.99% 0.97% 0.79%
Expenses $15 /sf $           15.45 $           16.00 $       16.59
Expense Growth Rate 3% 3.56% 3.70% 7.11%
Vacancy 10% 19.38% 15.20% 26.38%
Capital Expenditures 10% of NOI 10.16% 10.84% 10.97%
Terminal Cap Rate 11.00% 11.18% 11.84% 11.57%
OCC/Discount Rate 12.50%

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 66 
 
In ChallengeCo, the input assumptions are given a healthy dose of random walk, plus two
new assumptions appear: total development cost and flex development costs. Total
development cost will be the initial hard and soft costs of constructing the office building.
The flex development cost represents the cost to construct an addition, inflated by the
same rate as the rent growth rate. The option exercise trigger cannot be missed with a thick
red border.

The goal here is to model the construction of an additional 10 floors sometime in the future
when the economy improves. In this case, the rent is initially $30 per square foot, so the
option should be exercised when the rent rises up. Let us see what happens if we trigger
construction of the additional 10 floors when the rents reach $35 per square foot.

ChallengeCo Parameters Year 0 1 2 3 4 5 6
Total Development Cost $100 /gsf
Efficiency 90%
Gross Floor Area (Addition Incl) 170,000 sf 0.00 0.00 0.00 0.00 0.00 170,000 sf
Option Exercise Trigger: Rent $35 /sf
Flex Development Cost $105 /gsf $         108.15 $         108.46 $     110.78 $     113.79 $     116.91 $     123.14
Office Rent $30 /sf $           30.90 $           30.99 $       31.65 $        32.51 $        33.40 $        35.18
Rent Growth Rate 3% 0.29% 2.13% 2.73% 2.74% 5.33% 0.43%
Expenses $15 /sf $           15.45 $           16.05 $       16.62 $        17.12 $        17.73 $        18.59
Expense Growth Rate 3% 3.89% 3.55% 2.97% 3.61% 4.84% 0.86%
Vacancy 10% 4.79% 12.95% 10.16% 9.99% 4.88% 0.00%
Capital Expenditures 10% of NOI 9.85% 9.76% 9.44% 10.04% 9.72% 9.51%
Terminal Cap Rate 11.00% 10.83% 11.31% 11.08% 11.60% 11.92% 11.49%
OCC/Discount Rate 12.50%

In the 6th year of the pro forma, the rent climbs above $35 per square and immediately, an
additional 170,000 square feet (10 floors) is added through by using IF statements. The IF
statement is used as a switch between adding 170,000 square feet and not adding more
floor area.
value if true
At its core, the IF Statement has three parts.
logical test
1) Logical test

2) Value if true
value if false

3) Value if false

Organized in this format: =IF ( logical test , value if true , value if false )

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 67 
 
Think of the IF statement as a "fork in the road" with the logical test as the input. Value if
true and value if false are outputs.

The logical test is an equation that tells the computer what to do when it encounters the
fork in the road. Verbally, the logical test will read: “if the rent is greater than $35…” In
excel it would be: “E$8$>B5” with B8 being the rent in the current year and B5 as the
trigger rent.

‘Value if true’ is the outcome which will occur when the logical test is true, with the
opposite being true for the “value if false”.

Build 170k sf
If Rent > Trigger

IF ( Rent > Trigger, 170000 , 0 )

Don’t build

For each year that the possibility exists to construct another 170,000 sf, an IF statement is
required.

New Problem: In current form, the IF statements we created will automatically construct
an additional 170,000 sf without memory of what happened in the previous years. Thus,
there could be 170,000 sf of construction every year even though the real option that is
being modeled can only occur one time.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 68 
 
To solve this, an IF statement is nested inside another one to create a switch with more
than two outcomes. Here are the components:

1) Logical test value if true

logical test #1
2) Value if true
value if true #2
3) Value if false...another IF statement
logical test #2
(value if false)
3a) Value if true
value if false #2
3b) Value if false

In the nested IF statement, the first logical test is set up to see if 170,000 square feet was
constructed beforehand. If there has been another 170,000 square feet built beforehand,
then no construction takes place (effectively, ignoring anything rent does in that year). If
the addition has not been constructed yet, then proceed to the second IF Statement level.

On the second level, the previous logical test and outcomes are the same.

IF ( SUM( previous years sf )>0 , 0, IF (Rent > Trigger, 170000 , 0 ))

Now, the Real Option of building an additional 10 floors sometime in the future is modeled
and is ready for the Monte Carlo Simulation.

Has construction  Don’t build
already begun?
Build 170k sf

Is Rent > Trigger?

Don’t build

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 69 
 
Rarely does rent make up a trigger on its own for real estate, vacancy is important factor
also. Adding vacancy as another trigger is simple with a third nested IF statement. Notice
that the decision is to only build if the vacancy rate falls below the trigger. The logic can be
followed by the tree below:

Has construction  Don’t build
already begun?
Don’t build

Is Vacancy > Trigger?
Build 170k sf

Is Rent > Trigger?

Don’t build

IF ( SUM( previous years sf )>0 , 0, IF ( Vac > Trigger, 0 , IF ( Rent > Trigger, 170000 , 0 ) )

IF statements become messy very quickly, but with good organization and patience, even
the most the complex decision rules in Real Options can be modeled. Once there is
confidence that the additional 170,000 square feet is constructed when the decision rules
are satisfied, the parameters can be tied in to the rest of the pro forma to eventually
calculate down to the NPV.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 70 
 
ChallengeCo Parameters Year 0 1 2 3 4 5 6
Total Development Cost $100 /gsf
Efficiency 90%
Gross Floor Area (Addition Incl) 170,000 sf 0.00 0.00 0.00 0.00 170,000 sf 0.00
Option Exercise Trigger: Rent $35 /sf
Option Exercise Trigger: Vacancy 5.00%
Flex Development Cost $105 /gsf $         108.15 $         111.67 $     117.06 $     119.86 $     128.57 $     139.67
Office Rent $30 /sf $           30.90 $           31.91 $       33.45 $        34.24 $        36.73 $        39.91
Rent Growth Rate 3% 3.26% 4.82% 2.39% 7.27% 8.64% 6.85%
Expenses $15 /sf $           15.45 $           15.71 $       15.41 $        15.29 $        14.74 $        14.91
Expense Growth Rate 3% 1.69% ‐1.92% ‐0.76% ‐3.59% 1.12% 0.74%
Vacancy 10% 1.61% 4.01% 4.74% 0.00% 0.00% 1.13%
Capital Expenditures 10% of NOI 10.18% 10.23% 10.32% 10.40% 9.53% 8.20%
Terminal Cap Rate 11.00% 10.96% 11.27% 11.85% 12.06% 12.12% 11.56%
OCC/Discount Rate 12.50%

ChallengeCo Flexibility
(in 000's) Year 1 2 3 4 5 6
Potential Gross Income $4,728 $4,882 $5,117 $5,239 $5,620 $12,212
Vacancy $76 $196 $243 $ $ $138
Effective Gross Income $4,652 $4,686 $4,874 $5,239 $5,620 $12,073

Operating Expenses $2,364 $2,404 $2,358 $2,340 $2,256 $4,562

Net Operating Income $2,288 $2,282 $2,517 $2,900 $3,364 $7,511

Applying the same Monte Carlo Simulation as in ModerateCo Tower, the impact of the Real
Option is evident the CDF is compared to the CDFs of models without flexibility built in.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 71 
 
When compared to an otherwise identical 10 floor officer tower, ChallengeCo Tower with a
Real Option of building an additional 10 floors in the future is slightly worse off when
economic conditions turn out to be poor. The flexible office tower’s CDF is slightly shifted
to the left of the 10 floor office tower’s CDF because of the slightly higher construction costs
we modeled in for flexibility ($105/sf vs $100/sf). On the other hand, if economic
conditions turn out excellent, the flexible ChallengeCo Tower dominates the 10 floor office
tower because the real option to expand is exercised and allows the flexible building to take
advantage of favorable conditions.

Now contrast the flexible ChallengeCo Tower with the 20 floor non‐flexible building. The
20 floor building is exposed to much more risk as the upside is good, but the downside is
absolutely disastrous. This high level of risk in the 20 floor office building occurs because
of the high operation leverage created from the large construction cost.

There is no standard way of modeling Real Options into your financial model. The level of
complexity is only limited by the creativity and patience of the analyst creating the pro
forma. Armed with knowledge of a handful of functions in Excel, any real estate
professional can easily model uncertainty and real options in to pro formas to provide
valuable insights for their multi‐million dollar projects.

As illustrated in this example, incorporating design and/or financial flexibility into real
estate investments can have a significant impact in not only Expected Net Present Value,
but risk profiles as well. Understanding the effects of under certainty on real estate
ventures can lead a tremendous competitive advantage.

Beyond DCF Analysis in Real Estate Financial Modeling: Probabilistic Evaluation of Real Estate Ventures 72