Evaluating Real Estate Development

Using Real Options Analysis

Graeme Guthrie
Victoria University of Wellington
November 9, 2009
Abstract
This paper describes a simple method for using real options analysis to evaluate commercial
real estate projects when development can be suspended. The approach described can be
applied to projects at any stage of construction, which is especially useful when market
conditions are poor and suspension of many partly-completed projects is being considered.
All the calculations can be performed in a spreadsheet and only one parameterthe volatility
of the price of a completed projectneeds to be estimated in addition to those required for
static DCF analysis.
JEL Classication codes: C61, G13, G31, R33
Keywords: commercial real estate; real options; project evaluation
1 Introduction
Static DCF analysis remains the valuation tool of choice for many practitioners, despite its
failure to incorporate the eects of managerial exibility. This exibility takes many forms,
from the ability to delay beginning construction of a planned project to complicated sequencing
of large developments that may span several distinct stages spread over many years. Real
options analysis oers the potential to value projects in a way that recognizes the existence
of the exibility available to developers and, moreover, to determine the best way to exercise
this exibility. In contrast, static DCF analysis implicitly assumes that the entire sequence of
decisions regarding completion of a development is made before construction actually begins
and, no matter what might happen afterwards, is adhered to strictly. The main advantages of
static DCF analysis are that it is simple to implement and, because it is implemented in a fairly
systematic way, the results are easily communicated. In contrast, real options analysis is often
1
implemented using mathematical tools unavailable to many practitioners. Furthermore, many
dierent approaches are possible, and no single approach has emerged as standard practice.
Real options analysis has made substantial inroads into the academic study of real estate
markets. For example, it has been used to analyze the value of vacant land (Titman, 1985)
and the exibility embedded in typical leases (Grenadier, 1995); there is a large (and growing)
literature using real options analysis to analyze the equilibrium behavior of property prices
(Grenadier, 1996; Guthrie, 2009a). However, progress on implementing real options analysis in
practice has been much slower. While various case studies have appeared in the practitioner
literature, there seems to be no widely-used systematic approach to real options analysis applied
to real estate problems.
Several recent books attempt to make the insights of real options analysis available to a wider
range of practitioners. This paper explains how the approach presented in one of them, Guthrie
(2009b), can be applied to a wide variety of real estate applications. It avoids the dierential
equations and complicated simulations that are often used to carry out real options analysis,
instead using the familiar binomial-tree approach. The key step, and what distinguishes it from
standard derivative pricing situations, is the identication of a small number of discrete stages
of development and the allocation of the various items of capital expenditure amongst these
stages. This allocation reduces project evaluation to the analysis of a reasonably straightforward
compound option pricing problem. Construction lags add a slight complication to the analysis,
but they are easily handled in a systematic fashion.
The data requirements for this approach are not much more onerous than those for static
DCF analysis. In fact, the only additional inputs needed are the volatility of completed-building
prices and the expenditure required to suspend and resume construction. However, although
real options analysis uses the same inputs as static DCF analysis (augmented by volatility),
we will see that the importance of the various inputs changes when development exibility is
incorporated into the valuation procedure. For example, we nd that the market value of a
partly-completed project is much more sensitive to the level of the capitalization rate, and much
less sensitive to the price of completed buildings, than is suggested by static DCF valuation.
Thus, not only does static DCF analysis underestimate the market value of development projects,
but it also gives a misleading indication of the determinants of that market value. In short, it
tells practitioners to concentrate on building prices when they should perhaps be focussing more
on the capitalization rate and price volatility.
The particular example presented in this paper relates to a partly-completed project. The
developer has the option to suspend development, an option that many developers have exercised
in the 20072009 crisis period. In a world of volatile market conditions, the ability to slow down
or halt construction, and resume it at a later date, can be extremely valuable. Real options
analysis as described in this paper allows practitioners to value projects that have suspended
2
construction or have the option to do so in the future. Static DCF analysis, in contrast, does
not: it implicitly assumes that all future decisions are programmed in advance, which prevents
owners from resuming construction when market conditions improve and suspending it when
they deteriorate. Other forms of construction exibilitysuch as the option to change the scale
or scope of a project part way through constructioncan be handled using straightforward
extensions of the techniques described here.
1
We begin by describing our hypothetical example and then, in Section 3, we evaluate it
using static DCF analysis. Section 4 shows how to set up our valuation model and Section 5
then describes three dierent approaches to estimating the volatility parameter. The valuation
approach is presented in Section 6 and applied to our hypothetical project in Section 7.
2 Hypothetical development project
We will illustrate the real options approach using a hypothetical example of a partly completed
apartment complex in Las Vegas at the end of 2008.
2
The example involves a 24-story apartment
complex that, when completed, will have 650,000 sq ft of rentable space. Similar buildings
currently sell for $160 per sq ft. Construction, which will take 26 months if it is not interrupted,
is divided into ve distinct stages: construction of the substructure; construction of the lower
part of the buildings shell; construction of the upper part; the interior t out of the lower oors;
and the interior t out of the upper oors. The magnitude and timing of the required cash ows
are shown in the left-hand panel of Table 1 for the case when the projects construction is not
interrupted.
3
We assume that the level of all construction costs is risk free.
4
Construction can be interrupted between these stages (but not within a stage), but the
developer must pay $10,000 per month while construction is suspended to secure the site, retain
the services of key personnel, and so on. The entire project can be abandoned between stages
(but, again, not within a stage). The cost of cleaning up the site and shutting the project
down almost completely osets any residual land value, so that the project is worthless if it is
abandoned.
We will use 2% as our estimate of the real risk-free interest rate, consistent with the yield
on 10-year Treasury ination-protected securities (TIPS) at the time, and 6% as our estimate of
1
See Guthrie (2009b) for many more examples of the types of situations that can be analyzed.
2
Las Vegas experienced a boom and subsequent bust in its commercial real estate market. For example,
according to the Current Employment Statistics produced by the Bureau of Labor Statistics, estimated total
employment in Las Vegas in building construction climbed from 11,400 in August 2002 to 19,200 in August 2007,
before falling to 12,600 in August 2009.
3
These cost estimates are broadly in line with those reported by RSMeans for similar projects in Las Vegas
during the rst quarter of 2009. See http://www.meanscostworks.com/.
4
It is straightforward to incorporate construction-cost risk, although this requires introducing a more compli-
cated valuation model built on a multinomial, rather than binomial, tree.
3
Table 1: Classifying cash ows needed to construct an apartment complex
Actual project Model project
Item Amount Date decided Date incurred Stage Duration Lump sum cash ow
($m) (month) (month) (periods) ($m)
Substructure
Excavation 0.1 0 0
Foundations 9.5 0 1 0.1 +
9.5
1.02
1/12
+
0.2
1.02
5/12
Slab on grade 0.2 0 5 I 6 = 9.78
Shell (1st stage)
Floor construction 3.3 6 6
Exterior walls 12.5 6 8
Floor construction 3.3 6 8
Exterior walls 12.5 6 10
Floor construction 3.3 6 10 3.3 +
15.8
1.02
2/12
+
15.8
1.02
4/12
+
12.5
1.02
6/12
Exterior walls 12.5 6 12 II 6 = 47.12
Shell (2nd stage)
Floor construction 3.3 12 12
Exterior walls 12.5 12 14
Floor construction 3.3 12 14
Exterior walls 12.5 12 16 3.3 +
15.8
1.02
2/12
+
12.9
1.02
4/12
Roof construction 0.4 12 16 III 6 = 31.86
Interior (1st stage)
Elevators and lifts 9.0 18 18
Interior t out 14.0 18 18 23 +
14
1.02
2/12
Interior t out 14.0 18 20 IV 4 = 36.95
Interior (2nd stage)
Interior t out 14.0 22 22 14 +
14
1.02
2/12
Interior t out 14.0 22 24 V 4 = 27.95
4
the capitalization rate, consistent with market evidence in the fourth quarter of 2008 (Poutasse,
2009).
5
In order to help us focus on the dierences between static DCF analysis and real options
analysis, we will ignore taxes.
3 Static DCF analysis
Static DCF analysis implicitly assumes that the developer does not exploit any of the exibility
embedded in the project. For the one described in Section 2, we would usually assume that
the developer constructs the building as quickly as possible. The estimated market value of the
project is the present value of its net cash ows, calculated assuming that construction follows
the timetable in the left-hand panel of Table 1. Completing the project is deemed desirable if
and only if this net present value is positive.
We could use a single discount rate to calculate the present value of all cash ows. However,
in keeping with the notion that discount rates should reect the systematic risk of the individual
cash ows, we will discount capital expenditure using the risk-free interest rate and discount
the expected value of the completed building using a rate that reects the systematic risk of
property prices.
If construction has not yet begun, then the present value of the capital expenditure in
Table 1 is $150.50m. To calculate the present value of the building we plan to construct, we
could rst compound the current market value of an equivalent building (0.65 160 = 104
million dollars) for 26 months using an estimate of the expected growth rate in building prices
and then discount this amount back to the present using the risk-adjusted discount rate. It
is simpler (and equivalent) to discount the current market value using the capitalization rate,
which compensates the owner for the fact that the building does not generate any operating
revenue for 26 months. Our capitalization rate of 6% implies a present value of $91.67m. Static
DCF analysis of the project would conclude that construction should not begin.
The approach is similar if the rst stage of the exterior shell (only) has already been com-
pleted. We should only consider the present value of the capital expenditure required to complete
the project, so that sunk expenditure is ignored and the other expenditure is discounted back to
month 12, when the rst stage of the exterior shell has just been completed. The present value
of the remaining capital expenditure is therefore
3.3 +
12.5 + 3.3
1.02
2/12
+
12.5 + 0.4
1.02
4/12
+
9.0 + 14.0
1.02
6/12
+
14.0
1.02
8/12
+
14.0
1.02
10/12
+
14.0
1.02
12/12
= 95.95
million dollars. Similarly, when calculating the present value of the completed building we need
to allow for the shorter time until completion, discounting from month 26 back to month 12.
5
The capitalization rate (b) is the ratio of net operating income to the market value of the hypothetical
completed project. It is related to the expected growth rate in property prices (g) and the risk-adjusted discount
rate via (1 +b)(1 +g) = 1 + RADR.
5
The present value of the completed building is
104
1.06
14/12
= 97.16
million dollars. In this case, static DCF analysis of the project concludes that construction
should continue; the implied valuation of the project equals the dierence between these two
present values, or $1.22m.
4 Modeling development exibility
There are several key decision points during construction at which the developer must commit to
particular future expenditure. For example, the developer must rst commit to the expenditure
required to excavate the site and complete the buildings foundations. The second commitment
is to begin construction of the lower part of the buildings shell. As shown in Table 1, this
commitment can be delayed until the substructure has been completed without construction
being interrupted. It can be delayed even longer, although in this case construction would have
to be suspended once the substructure is completed. Construction can resume only once the
commitment to the second group of cash ows is made. Similarly, the developer can delay
committing to the expenditure required to build the top part of the shell until the lower part
is completed without construction being interrupted. Longer delays will result in construction
being suspended with the shell only partly completed.
We capture this ability to delay committing to some future expenditure by reducing the
sequence of cash ows to a smaller set of equivalent lump sum cash ows, each incurred at
the time the commitment to incur the relevant future expenditure is made. For example, if
the developer decides to begin construction of the substructure at month 0 then the developer
commits to spend $0.1m at month 0, $9.5m at month 1, and $0.2m at month 5. This sequence
of cash ows is replaced by a single lump sum cash ow of
0.1 +
9.5
1.02
1/12
+
0.2
1.02
5/12
= 9.78
million dollars that is incurred at the date the decision is made, in this case month 0. Similarly,
at month 6 the developer commits to expenditure of $3.3m at month 6, $15.8m at month 8,
another $15.8m at month 10, and $12.5m at month 12. We replace this cash ow stream by
lump sum expenditure of
3.3 +
15.8
1.02
2/12
+
15.8
1.02
4/12
+
12.5
1.02
6/12
= 47.12
million dollars that is incurred at month 6. It is essential that these present values are calculated
as at the decision date, not as at month 0. Thus, the cash ow of $12.5m is discounted six months,
6
Table 2: Five-stage representation of the apartment project
Stage Duration Expenditure
(months) ($m)
I 6 9.78
II 6 47.12
III 6 31.86
IV 4 36.95
V 4 27.95
from month 12 back to month 6, not all the way back to month 0. The results for the remaining
cash ows are shown in the right-hand panel of Table 1.
The apartment construction project is therefore reduced to the simpler ve-stage project
described in Table 2. For each stage the cash ows are collapsed into a single lump sum equal
to their present value. The timing of the individual cash ows is important only to the extent
that it determines their present value. However, we need to know the overall time required to
complete each stage because one stage cannot begin until the previous one has been completed.
For example, the developer cannot begin Stage II until at least six months have elapsed after
beginning Stage I.
All the information used up to this point is required for a static DCF analysis of the project.
The only additional information that we require in order to incorporate the various real options
into our project evaluation is the volatility of changes in the sales price of the completed building.
The next section shows how to estimate volatility from available market data.
5 Estimating volatility
We need to specify the volatility (that is, the standard deviation of the annual rate of return) of
the market value of the completed project. The hypothetical sale price of the building currently
under constructionif it had just been nishedis clearly unobservable, so that we must rely on
other, observable, variables to serve as proxies when estimating volatility. This section describes
three possible approaches to estimating the volatility of the completed projects market value.
The rst approach uses information contained in commercial real estate (CRE) indices, which
attempt to measure the behavior of diversied portfolios of such assets. Note, however, that
we are interested in the behavior of the (hypothetical) market value of a particular building in
a particular location. Even those indices that are disaggregated to the level of building classes
or geographic locations will feature some degree of diversication. Given that the changes in
prices of individual buildings will be less-than-perfectly correlated, this diversication means
that the volatility of an index will underestimate the volatility of our building price, perhaps to
7
Table 3: Estimates of volatility from various CRE price indices
Region Building Sample Standard deviation
Quarter Annual
NCREIF Property Index (NPI)
National All 1978:12008:4 0.0177 0.0354
National Apartments 1984:12008:4 0.0158 0.0317
Transaction-Based Index (TBI)
National All 1984:12008:4 0.0387 0.0773
National Apartments 1994:12008:4 0.0408 0.0817
Moodys/REAL CPPI
National Apartments 2001:12008:3 0.0444 0.0888
Western Apartments 2001:12008:3 0.0426 0.0852
a signicant extent.
CRE indices are frequently based on the appraised values of properties in selected portfolios.
For example, the National Council of Real Estate Investment Fiduciaries (NCREIF) publishes
the NCREIF Property Index (NPI) of investment properties held by pension funds in the United
States. The appraisal process tends to smooth out uctuations in the market value of properties,
so that the volatility of indices such as the NPI will tend to underestimate the volatility of the
particular diversied portfolio of properties. A better source for our purposes is an index such
as the Transaction-Based Index (TBI) produced by the MIT Center for Real Estate. The
TBI is calculated using the estimated relationship between transaction prices and property
characteristics for properties in the NCREIF portfolio. It will thus do a better job of capturing
the actual volatility in market prices than an appraisal-based index. Repeat-transaction indices,
such as the Moodys/REAL CPPI produced by the MIT Center for Real Estate and which
includes most transactions exceeding $2.5m, are also beginning to appear.
Table 3 reports volatility estimates derived from the NPI, TPI, and CPPI indices, where all
available data have been used for each series. The published index levels have been converted
into real terms using the CPI for all urban consumers (US city average) published by the
Bureau of Labor Statistics. The rst entry in each pair reports the standard deviation of
the quarterly change in the logarithm of the price-only index and the second entry gives its
annualized counterpart.
6
As expected, the appraisal-based NPI has very low volatility. For the
other two indices, volatility estimates range from 0.0773 (the national TPI, with all property
types incorporated in a single index) to 0.0888 (the national CPPI for apartments).
Historical data on real estate investment trust (REIT) share price returns provide a second
approach to estimating volatility. Because the share prices are true trading prices, the problems
6
The standard deviation is annualized by multiplying it by the square root of the number of observations per
year.
8
caused by the reliance of CRE indices on appraisals and infrequent transactions are avoided.
However, like the CRE indices, REIT prices reect the values of portfolios of properties. This
diversication means that REIT-based volatility is likely to underestimate the volatility for a
single building. Downing et al. (2007) have analyzed 18 rm-level REIT returns over the period
19952005. They obtain an average de-levered volatility estimate for multi-family properties of
0.107, with a standard deviation of 0.031.
The third approach that we consider uses observed prices of nancial securities with values
that are sensitive to the volatility of property prices. Given a suitable valuation model, the
so-called implied volatility is the level of volatility that sets the models predicted price equal
to the observed price.
7
Implied volatility is a forward-looking estimate, being based on the
markets expectations of future volatility. In contrast, because they are based on historical data,
the rst two approaches described above are backward-looking. Downing et al. (2007) calculate
the level of volatility implied by each of 4,032 CRE loans involving multi-family properties issued
during the period 19962005. The implied volatilities that they calculate have a mean of 0.198
and a standard deviation of 0.079.
8
Using loans on individual properties has the advantage that
the implied volatility reects the volatility of the price of a single property: the diversication
present in CRE indices and REIT returns is absent.
These three approaches produce volatility estimates ranging from approximately 0.08 (CRE
indices based on property portfolios) to 0.20 and above (volatility implied by loans on individual
properties). Since we require an estimate of the volatility for a single property, we set volatility
equal to = 0.20.
6 Using real options analysis to evaluate the project
We carry out our analysis of this problem using the framework of a binomial tree, which we use
to describe the behavior of the hypothetical price of a completed project. The T-year lifetime
of the development rights is divided into a sequence of N periods, each representing t = T/N
years. During each one of these periods the hypothetical price of a completed project changes
by a constant factor of either U = e

t
> 1 or D = e

t
< 1. The starting value, denoted
X
0
, is the hypothetical market value the project would have today if it had just been completed.
In our example, X
0
= 104 million. By the time we get to date n, when there will have been
7
This approach is often used in derivative pricing applications, but it can also be used to estimate other
parameters of interest. For example, Boyle et al. (2009) use it to estimate the parameters needed to calculate
fair-value ground rentals.
8
Older studies obtain similar values. For example, Titman and Torous (1989) calculate an implied volatility
of building prices of 0.1550 from quoted rates on commercial mortgages during the period 19851987; Holland
et al. (2000) use mortgage rates reported by the American Council of Life Insurers during the period 19791992
to obtain estimates ranging from 0.15 to 0.25.
9
some random number i down moves and n i up moves, the hypothetical market value will
have changed to V
0
(i, n) = X
0
D
i
U
ni
. The top left-hand corner of the resulting binomial tree
is shown in the rst panel in Table 4, with each row corresponding to a dierent number (i)
of down moves and each column (n) to a dierent date.
9
This tree completely describes our
uncertainty about future property prices. For example, six periods from now, the hypothetical
price of the completed project could be as low as X
0
D
6
, as high as X
0
U
6
, or somewhere in
between. The possible prices comprise the column labeled by n = 6 in the rst panel in Table
4.
The binomial tree for V
0
is the basis of our valuation model. No matter what stage of
construction the project is at, its market value will depend on the hypothetical price of the
project if it had just been completed. Thus, we use V
m
(i, n) to denote the market value of the
project rights if m construction stages remain and a completed project couldhypothetically
currently be sold for V
0
(i, n). We use a separate table for each of the ve values of m, with the
entry in cell (i, n) of the mth of these tables giving the level of V
m
(i, n).
In order to understand how our modeling approach works we must understand how we move
through the trees as construction progresses and market conditions evolve. We start at node
(0, 0) of the bottom panel, which will give the estimated current market value of the project rights
if construction is yet to begin. If the developer decides to wait before beginning construction
thenbecause no construction is going to occur in the meantimewe stay in the same panel, but
move one column to the right each period, moving down one row each time a down move occurs
and otherwise staying in the current row. In contrast, if the developer decides to construct the
projects rst stage, we jump up to the panel directly above the bottom one, moving as many
columns to the right as it takes periods to complete that stage. Where exactly we emerge in
that column depends on how the market has behaved during construction. If every move during
construction of the rst stage has been up then we will emerge in the same row, but for each
down move that occurred during this period we will emerge one row lower. We move through the
panels like this, moving rightwards through each panel and occasionally jumping up to the next
one. Provided we do not permanently abandon the project along the way, we will eventually
arrive at either the top panel, indicating that the projects construction has been completed, or
the nal column of one of the other panels, indicating that the development rights have expired
before the project could be completed.
Now that we know how to navigate through the panels, we turn to the task of constructing
them. All the required calculations can be carried out in a spreadsheet containing six tables, one
for each of the project market values (V
m
). The rst table extends out as far as the developments
expiry dateso that the tables can be largebut each successive table is smaller than the one
9
Because there cannot be more down moves than periods, the part of the table below the diagonal does not
need to be lled in.
10
Table 4: Overview of solution technique
V0(i, n) 0 1 2 3 4 5 6
0 V0(0, 0) V0(0, 1) V0(0, 2) V0(0, 3) V0(0, 4) V0(0, 5) V0(0, 6)
1 V0(1, 1) V0(1, 2) V0(1, 3) V0(1, 4) V0(1, 5) V0(1, 6)
2 V0(2, 2) V0(2, 3) V0(2, 4) V0(2, 5) V0(2, 6)
3 V0(3, 3) V0(3, 4) V0(3, 5) V0(3, 6)
4 V0(4, 4) V0(4, 5) V0(4, 6)
5 V0(5, 5) V0(5, 6)
6 V0(6, 6)
V1(i, n) 0 1 2 3 4 5 6
0 V1(0, 0) V1(0, 1) V1(0, 2) V1(0, 3) V1(0, 4) V1(0, 5) V1(0, 6)
1 V1(1, 1) V1(1, 2) V1(1, 3) V1(1, 4) V1(1, 5) V1(1, 6)
2 V1(2, 2) V1(2, 3) V1(2, 4) V1(2, 5) V1(2, 6)
3 V1(3, 3) V1(3, 4) V1(3, 5) V1(3, 6)
4 V1(4, 4) V1(4, 5) V1(4, 6)
5 V1(5, 5) V1(5, 6)
6 V1(6, 6)
V2(i, n) 0 1 2 3 4 5 6
0 V2(0, 0) V2(0, 1) V2(0, 2) V2(0, 3) V2(0, 4) V2(0, 5) V2(0, 6)
1 V2(1, 1) V2(1, 2) V2(1, 3) V2(1, 4) V2(1, 5) V2(1, 6)
2 V2(2, 2) V2(2, 3) V2(2, 4) V2(2, 5) V2(2, 6)
3 V2(3, 3) V2(3, 4) V2(3, 5) V2(3, 6)
4 V2(4, 4) V2(4, 5) V2(4, 6)
5 V2(5, 5) V2(5, 6)
6 V2(6, 6)
V3(i, n) 0 1 2 3 4 5 6
0 V3(0, 0) V3(0, 1) V3(0, 2) V3(0, 3) V3(0, 4) V3(0, 5) V3(0, 6)
1 V3(1, 1) V3(1, 2) V3(1, 3) V3(1, 4) V3(1, 5) V3(1, 6)
2 V3(2, 2) V3(2, 3) V3(2, 4) V3(2, 5) V3(2, 6)
3 V3(3, 3) V3(3, 4) V3(3, 5) V3(3, 6)
4 V3(4, 4) V3(4, 5) V3(4, 6)
5 V3(5, 5) V3(5, 6)
6 V3(6, 6)
V4(i, n) 0 1 2 3 4 5 6
0 V4(0, 0) V4(0, 1) V4(0, 2) V4(0, 3) V4(0, 4) V4(0, 5) V4(0, 6)
1 V4(1, 1) V4(1, 2) V4(1, 3) V4(1, 4) V4(1, 5) V4(1, 6)
2 V4(2, 2) V4(2, 3) V4(2, 4) V4(2, 5) V4(2, 6)
3 V4(3, 3) V4(3, 4) V4(3, 5) V4(3, 6)
4 V4(4, 4) V4(4, 5) V4(4, 6)
5 V4(5, 5) V4(5, 6)
6 V4(6, 6)
V5(i, n) 0 1 2 3 4 5 6
0 V5(0, 0) V5(0, 1) V5(0, 2) V5(0, 3) V5(0, 4) V5(0, 5) V5(0, 6)
1 V5(1, 1) V5(1, 2) V5(1, 3) V5(1, 4) V5(1, 5) V5(1, 6)
2 V5(2, 2) V5(2, 3) V5(2, 4) V5(2, 5) V5(2, 6)
3 V5(3, 3) V5(3, 4) V5(3, 5) V5(3, 6)
4 V5(4, 4) V5(4, 5) V5(4, 6)
5 V5(5, 5) V5(5, 6)
6 V5(6, 6)

Step 1: Fill in to
date N using
V0(i, n) = X0D
i
U
ni

Step 2a: Fill in date

N N1 + 1 with zeros

Step 2b: Fill in

earlier cols
using eq. (1)
with m = 1

Step 3a: Fill in date

N N1 N2 + 1
with zeros

Step 3b: Fill in

earlier cols
using eq. (1)
with m = 2

Step 4a: Fill in date

N N1 N2 N3 + 1
with zeros

Step 4b: Fill in

earlier cols
using eq. (1)
with m = 3

Step 5a: Fill in date

N N1 N4 + 1
with zeros

Step 5b: Fill in

earlier cols
using eq. (1)
with m = 4

Step 6a: Fill in date

N N1 N5 + 1
with zeros

Step 6b: Fill in

earlier cols
using eq. (1)
with m = 5
11
before it.
Construction of the rst panel in Table 4 is simplest since the entry at node (i, n) can be
calculated directly as V
0
(i, n) = X
0
D
i
U
ni
. However, construction of the other panels is more
complicated. We saw above that the system evolves starting at the bottom panel and then
jumping up one panel at a time, and moving from left-to-right within each panel. This direction
needs to be reversed when constructing the tables. That is, we actually work backwards through
time, starting with the table when there is just one construction stage remaining and lling it
in from right-to-left, before moving to the table when there are two stages remaining, and
repeating the procedure. We continue like this, until we eventually calculate the left-most cell
of the bottom table.
For each table, we begin by lling in the nal columncorresponding to immediately after
the latest date on which that stage of construction can beginwith zeros. Suppose that the next
construction stage takes N
m
periods to complete if there are m separate stages remaining. Then
if the developer has not begun the nal stage on or before date N N
1
, the rights are worthless
(since the project cannot be completed before the rights expire) and V
1
(i, N N
1
+ 1) = 0 for
all entries in that column. Similarly, if the developer has not begun the penultimate stage on or
before date NN
1
N
2
then the rights are also worthless, so that we make V
2
(i, NN
1
N
2
+1) =
0 for all entries in that column. We then ll in the remainder of the table, working from right
to left one column at a time, using the method that we now describe.
Consider what happens at an arbitrary node, (i, n), when there are still m stages remaining
to be completed. The developer has three choices: abandon the project permanently; suspend
construction (possibly temporarily); or construct the next stage immediately. The developer
will choose an action that maximizes the market value of the project rights, so that V
m
(i, n)
equals the maximum of the payos for these three actions. That is,
V
m
(i, n) = max{payo
abandon
(i, n), payo
suspend
(i, n), payo
develop
(i, n)}. (1)
This equation tells us how much the development rights are worth at this node andby checking
which of the three payos is largestindicates the optimal development policy. However, before
we can evaluate equation (1) we need to know how to calculate the three payos that it contains.
Abandonment The rst payo occurs when the developer abandons the project and receives
the salvage value A, for a payo of
payo
abandon
(i, n) = A.
Suspension Alternatively, the developer can pay H and temporarily suspend construction,
holding the project in its current state for one period. We already know that after one
period has passed we will be in the next column of the same table, in the same row if the
12
move is up and in the row below if it is down. That is, the project rights will be worth
V
m
(i, n + 1) if an up move occurs and V
m
(i +1, n + 1) if the move is down. In traditional
DCF analysis, we would calculate the market value of these possibilities by calculating
the expected value and then discounting it back one period using a risk-adjusted discount
rate. Real options analysis achieves exactly the same market-value estimate by altering
the expected cash ow and then using the risk-free interest rate to do the discounting.
Specically, the payo from suspension is
payo
suspend
(i, n) = H + (1 +r)
t

u
V
m
(i, n + 1) + (1
u
)V
m
(i + 1, n + 1)

,
where

u
=

1+r
1+b

t
e

t
e

t
e

t
,
r is the risk-free interest rate, b is the capitalization rate, and is the volatility of the
value of the completed project.
10
The term in large brackets is the expected value of the
rights after one period (calculated using the risk-neutral probability
u
), which is then
discounted back one period using the risk-free interest rate.
Construction The third payo is the most complicated to evaluate. If the developer decides to
construct the next stage of the project, then he commits to a series of cash outows with
present value I
m
. Since this stage takes N
m
periods to complete, there are many possible
outcomes to consider, one for each number of down moves that occur during construction.
For example, if there are j down moves (and so N
m
j up moves), the project rights will
be worth V
m1
(i +j, n+N
m
) when this stage of construction is completed. The subscript
indicates that there will be just m 1 stages remaining, the row number indicates that
there have been i +j down moves since date 0 in total, while the column number indicates
that we will be at date n +N
m
. When we take all possibilities into account, we nd that
the payo from completing the projects next stage is
payo
develop
(i, n) = I
m
+ (1 +r)
Nmt
Nm

j=0
(j, N
m
) V
m1
(i +j, n +N
m
),
where (j, N
m
) is the probability that exactly j of the next N
m
moves are down, where each
move is down with probability 1
u
.
11
The rst term is the present value of the required
capital expenditure, while the second term is the expected payo (calculated using the
risk-neutral probability), discounted back to the present using the risk-free interest rate.
10
Derivations of this and other results can be found in Chapter 3 of Guthrie (2009b).
11
It can be calculated in Excel using (j, Nm) = BINOMDIST(j, Nm, 1 u, FALSE). Alternatively, it can
be calculated directly using
(j, Nm) =
Nm
j!(Nm j)!

Nmj
u
(1 u)
j
.
13
Table 5: Summary of solution technique
Step Instruction
1 Fill in the tree for V0 out as far as date N using V0(i, n) = X0D
i
U
ni
2a Set all values of V1 at date N N1 + 1 equal to zero
2b Moving one column left at a time, ll in the remaining columns of the tree for
V1 using equation (1) with m = 1
3a Set all values of V2 at date N N1 N2 + 1 equal to zero
3b Fill in the remaining columns of the tree for V2 using equation (1) with m = 2
4a Set all values of V3 at date N N1 N2 N3 + 1 equal to zero
4b Fill in the remaining columns of the tree for V3 using equation (1) with m = 3
5a Set all values of V4 at date N N1 N2 N3 N4 + 1 equal to zero
5b Fill in the remaining columns of the tree for V4 using equation (1) with m = 4
6a Set all values of V5 at date N N1 N2 N3 N4 N5 + 1 equal to zero
6b Fill in the remaining columns of the tree for V5 using equation (1) with m = 5
Construction of the suspension and construction payos for the special case where m = 2,
i = 1, and n = 1 is illustrated in Table 4. We are currently at the node indicated by the
boxed cell, which shows V
2
(1, 1). The payo from suspension depends on the two shaded cells
in the same table since the value of the development rights changes to either V
2
(1, 2) or V
2
(2, 2)
after one period elapses. We take the weighted average of these two entries (where the weights
are the risk-neutral probabilities), discount back one period using the risk-free interest rate,
and then subtract one periods worth of suspension expenditure. In contrast, the payo from
developing this stage of the project depends on the shaded cells in the second table. Assuming
monthly time steps, construction will take four periods, so that the rights will be worth one
of V
1
(1, 5), . . . , V
1
(5, 5) as soon as construction of the next stage is complete. To calculate the
development payo, we take the weighted average of these ve entries (where the weights are
the risk-neutral probabilities given by the s), discount back four periods using the risk-free
interest rate, and then subtract the lump sum development expenditure.
The entire solution process is summarized in Table 5.
7 Results
Our parameter values are reported in Table 6. Using monthly time steps (that is, t = 1/12)
implies up and down moves of U = 1.0594 and D = 0.9439, respectively, and a risk-neutral
probability of an up move of
u
= 0.4579.
Table 7 shows an extract from the spreadsheet constructed using the approach described
in Section 6, corresponding to the rst six months of each table. The entry in each cell gives
the market value of the development rights for the indicated number of remaining stages, with
14
Table 6: Parameter values
Parameter Symbol Value
Needed for static DCF analysis
Current sale price X0 104
Capitalization rate b 0.06
Risk-free interest rate r 0.02
Capital expenditure I5 9.78
I4 47.12
I3 31.86
I2 36.95
I1 27.95
Construction times

N5 6
N4 6
N3 6
N2 4
N1 4
Not needed for static DCF analysis
Price volatility 0.20
Salvage payo A 0
Suspension expenditure

H 0.01
Lifetime of development rights

N 120

Assumes monthly time steps

shaded cells indicating situations when the next stage should be constructed immediately. For
example, if the project had just been completed it would be worth $104m (the top left-hand
entry of the rst table). It would be worth just $74.05m if a single stage remained (the top left-
hand entry of the second table), in which case the remaining stage should be built immediately.
Similarly, the project would be worth $35.31m if two stages remain and $1.67m if no construction
at all has occurred.
Table 8 compares the results of our real options analysis with those of static DCF analysis, as
described in Section 3. The rst two entries in each row give the market value of the development
rights if the building price is $160 per sq ft, calculated using the static DCF approach described
in Section 3, and the break-even level of the building price (that is, the level of the building price
that makes the present value of the remaining capital expenditure equal the present value of the
completed building, both calculated assuming that development is not interrupted). The next
two entries report analogous information calculated using the real options approach described
in Section 6. That is, the market values are the values of V
m
(0, 0) calculated in the spreadsheet
summarized in Table 7 and the development thresholds are the lowest building prices for which
15
Table 7: Spreadsheet extract
V0(i, n) 0 1 2 3 4 5 6
0 104.00 110.18 116.73 123.67 131.02 138.80 147.05
1 98.17 104.00 110.18 116.73 123.67 131.02
2 92.66 98.17 104.00 110.18 116.73
3 87.46 92.66 98.17 104.00
4 82.55 87.46 92.66
5 77.92 82.55
6 73.55
V1(i, n) 0 1 2 3 4 5 6
0 74.05 80.11 86.53 93.33 100.54 108.18 116.27
1 68.32 74.05 80.11 86.53 93.33 100.54
2 62.92 68.32 74.05 80.11 86.53
3 57.82 62.92 68.32 74.05
4 53.01 57.82 62.92
5 48.47 53.01
6 44.18
V2(i, n) 0 1 2 3 4 5 6
0 35.31 41.26 47.56 54.23 61.30 68.79 76.73
1 29.70 35.31 41.26 47.56 54.23 61.30
2 24.40 29.70 35.31 41.26 47.56
3 19.80 24.40 29.70 35.31
4 16.00 19.80 24.40
5 12.85 15.99
6 10.26
V3(i, n) 0 1 2 3 4 5 6
0 11.43 14.33 17.87 22.19 27.42 33.77 41.44
1 9.03 11.40 14.30 17.85 22.17 27.41
2 7.07 9.00 11.37 14.28 17.83
3 5.48 7.04 8.97 11.34
4 4.20 5.45 7.01
5 3.17 4.17
6 2.35
V4(i, n) 0 1 2 3 4 5 6
0 2.55 3.41 4.51 5.89 7.62 9.78 12.44
1 1.84 2.52 3.38 4.47 5.85 7.57
2 1.30 1.81 2.48 3.34 4.42
3 0.88 1.27 1.79 2.45
4 0.58 0.86 1.25
5 0.35 0.56
6 0.20
V5(i, n) 0 1 2 3 4 5 6
0 1.67 2.31 3.12 4.16 5.48 7.14 9.21
1 1.16 1.65 2.27 3.08 4.12 5.43
2 0.78 1.14 1.62 2.24 3.04
3 0.50 0.76 1.11 1.58
4 0.29 0.48 0.74
5 0.15 0.28
6 0.07
16
Table 8: Comparison of static DCF and real options analysis
No. stages Static DCF analysis Real options analysis
remaining Market value Price threshold Market value Price threshold
m ($m) ($ per sq ft) ($m) ($ per sq ft)
5 58.84 263 1.67 341
4 47.75 241 2.55 328
3 1.22 158 11.43 215
2 35.31 104 35.31 140
1 74.05 44 74.05 59
it is optimal to immediately construct the next stage of the project at date 0.
12
If construction of the project has not begun, then static DCF analysis severely underestimates
the market value of the project, giving it a negative value whendue to the option to delay
construction until market conditions improvethe project rights actually have a positive value.
Moreover, it is optimal to begin development only when the building price is much higher than
the level predicted by static DCF analysis ($341 per sq ft rather than $263 per sq ft). The
higher building price is needed to compensate the developer for the loss of the delay option
that is destroyed when construction begins. This is also the case when the rst stage has been
completed.
With three stages remaining, static DCF analysis would incorrectly suggest that construction
should be resumed immediately, since the current building price of $160 per sq ft exceeds the
static-DCF threshold. However, Table 8 shows that development should actually be delayed
until the building price rises to at least $215 per sq ft. It also shows that the development
rights are worth almost ten times the value predicted by static DCF analysis. Thus, using static
DCF analysis to evaluate projects for which the development option is at-the-money (that is,
the building price is currently close to the break-even threshold) undervalues the projects and
leads to too much development.
When there are only one or two stages remaining, static DCF analysis performs well for both
valuation and decision making, because in this case the value of being able to delay exercising
the development option is very low. As Table 8 shows, in this case the current building price of
$160 per sq ft is substantially above the optimal development threshold, so that the development
option is deep in-the-money. This behavior suggests that when analyzing a project, the exibility
to suspend development in the early stageswhen the development threshold is high and the
development option is less likely to be deep in-the-moneyis more valuable than exibility to
delay completing the nal few stages. Thus, when simplifying a project into a small number
12
When using Excel, the optimal threshold can be found by using the Solver function to nd the value of X0
for which payo
develop
(0, 0) = payo
wait
(0, 0).
17
Figure 1: Sensitivity analysis with three stages remaining: building price and capitalization rate
100 150 200 250
20
0
20
40
60
80
b = 0.03
b = 0.06
b = 0.09
$ per sq ft
$
m
of stages as in Section 4 it is important to focus on separating out various stages early in the
development process; later stages can be grouped together into a small number of stages without
greatly compromising our analysis.
The approach described in this paper assumes that the underlying parameters are known
with certainty. However, in practice, most parameters are estimated with uncertainty. One
possible response to this parameter uncertainty is to carry out sensitivity analysis, starting
with a distribution of possible parameter values and nishing with the implied distribution of
project values. Because the approach in this paper can be implemented within a spreadsheet,
construction of the distribution can be carried out in the same way as for static DCF analysis.
Rather than repeat that here, we restrict ourselves to examining the eect of changes in three
important parametersbuilding price, capitalization rate, and volatilityon the market value
of the development rights.
Figure 1 plots the market value (in $m) of the development rights when three stages remain,
as a function of the current building price (in $ per sq ft). The three solid curves show the
real-options valuations corresponding to dierent capitalization rates, 3% (top), 6% (middle),
and 9% (bottom); the dashed curves show the corresponding static DCF valuations. All other
parameters take the values in Table 6. For each combination of building price and capitalization
rate, the height of the curve equals V
3
(0, 0). The dot on each curve indicates the lowest level of
the building price at which it is optimal to immediately begin construction of the third-to-last
stage of the project. Real options analysis yields price thresholds ranging from $201 per sq ft
when the capitalization rate is 9% to $261 per sq ft when it is 3%; static DCF analysis gives
respective thresholds of $153 per sq ft and $163 per sq ft. Two points stand out. Firstly, static
DCF analysis gives development thresholds that are far too low. Secondly, it gives a misleading
view of the sensitivity of the optimal development threshold to the level of the capitalization
rate.
18
Figure 2: Sensitivity analysis with three stages remaining: building price and volatility
100 150 200 250
20
0
20
40
60
80
= 0.25
= 0.20
= 0.15
$ per sq ft
$
m
Figure 2 repeats the format of the preceding one, but this time the three solid curves show
the real-options valuations corresponding to dierent volatilities, 0.15 (bottom), 0.20 (middle),
and 0.25 (top). Because volatility does not aect static DCF analysis, there is just one dashed
curve showing the corresponding static DCF valuations. All other parameters take the values in
Table 6. Real options analysis yields price thresholds ranging from $192 per sq ft when volatility
is 0.15 to $243 per sq ft when it is 0.25. Both the value of the project rights and the optimal
development threshold are sensitive to the level of price volatility.
The relative importance of these three parameters can be measured by calculating the change
in each parameterwhile holding all other parameters at their baseline levelsrequired to
increase the project rights market value by 10% (that is, from $11.43m to $12.56m). Numerical
search routines (such as Solver in Excel) can be used to calculate that the building price needs
to rise by 2.4% (from $160.00 per sq ft to $163.79 per sq ft), or price volatility needs to rise by
7% (from 0.200 to 0.214), or the capitalization rate needs to fall by 10% (from 0.060 to 0.054). In
contrast, the static DCF valuation of the project rights increases by the same (absolute) amount
of $1.14m if the building price rises by 1.2% (to $161.88 per sq ft) or the capitalization rate
falls by 18% (to 0.049); changes in price volatility have no eect on the static DCF valuation.
Thus, static DCF valuation overstates the importance to valuation of the building price and
understates the importance of the capitalization rate; it entirely misses the importance of price
volatility.
8 Conclusion
This paper has demonstrated a straightforward method for incorporating construction exibility
in the evaluation of commercial real estate developments. All the calculations can be performed
in a spreadsheet and only one parameterthe volatility of the price of a completed project
needs to be estimated in addition to those required for static DCF analysis. The key step is to
19
allocate individual items of capital expenditure to distinct stages of development. The method
allows the developer to suspend construction between stages, thus capturing much of the value
of the real options embedded in typical development projects. Reducing the project to a small
number of stages means that we need to build only a few tables in the spreadsheet: one for the
market value of the completed project and one for each construction stage.
To make the calculations as simple as possible, we suppose that the developers decision con-
cerns only the timing of construction. Other possibilities, such as varying the rate of construction
according to market conditions and deciding to change the projects design mid-construction, can
also be analyzed using minor modications of the techniques described in this paper (Guthrie,
2009b, Chapter 9). Other extensions are possible, but substantially complicate the analysis. For
example, it is possible to incorporate uncertainty surrounding the technological requirements of
a project. A closely-related issue is the uncertainty surrounding future levels of some parame-
ters. That is, even if we know the level of the interest rate now, we do not know its value one
year from now with certainty. Sensitivity analysis can be used here, butbecause it varies the
level of a parameter while assuming that it will be constant at this levelit does not capture
all of the consequences of parameter volatility. A better approach is to explicitly include the
relevant parameter as an additional state variable. Again, this can be incorporated, but only if
we are willing to use multinomial trees rather than the simple binomial trees used in this paper.
References
Boyle, G. W., Guthrie, G., and Quigley, N. (2009). Estimating implied valuation parameters:
Extension and application to ground lease rentals. Accounting and Finance, 49(3):465479.
Downing, C., Stanton, R., and Wallace, N. E. (2007). Volatility, mortgage default, and CMBS
subordination. Technical report, Rice University.
Grenadier, S. R. (1995). Valuing lease contracts: A real-options approach. Journal of Financial
Economics, 38:297331.
Grenadier, S. R. (1996). The strategic exercise of options: Development cascades and overbuild-
ing in real estate markets. Journal of Finance, 51(5):16531679.
Guthrie, G. (2009a). House prices, development costs, and the value of waiting. Technical
report, Victoria University of Wellington.
Guthrie, G. (2009b). Real Options in Theory and Practice. Oxford University Press, New York,
New York.
20
Holland, A. S., Ott, S. H., and Riddiough, T. J. (2000). The role of uncertainty in investment:
An examination of competing investment models using commercial real estate data. Real
Estate Economics, 28(1):3364.
Poutasse, D. (2009). 2008Q4: A quarter like no other. National Council of Real Estate Invest-
ment Fiduciaries.
Titman, S. (1985). Urban land prices under uncertainty. American Economic Review, 75(3):505
514.
Titman, S. and Torous, W. (1989). Valuing commercial mortgages: An empirical investigation
of the contingent-claims approach to pricing risky debt. Journal of Finance, 44(2):345373.
21