Table of Contents
Title:
System dynamics of optimal commencement timing for
office building construction
Summary
Through office market forecasting consultancy jobs, a PhD dissertation and
several papers, I developed a system dynamics modd to simulate office market
oversupply cycles as a function of trend economic growth, an equilibium
vacancy rate and three supply response parameters. ( , 1997, 1999, 2002).
This paper proposes extensions of the previous work. Supply will be made
endogenous and “lumpy” by explicily using “building sized’ supply changes,
tather than assuming continuous supply responses. Prices will be explicitly
induded— previous versions used vacancy rates to signal market conditions.
Most importantly, forecast exors will be added to replace the deterministic
simulations of previous papers. These imovations in the modal should make it
more useful to decision makers for exploring optimal timing of major office
towers.
versus those that lose money (NPV <0) over the economic life of the building.
The decision problem is framed as a trade-off between two kinds of exrors
resulting from two different project commencement policies.
Many financial institutions adopt a “conservative” policy of waiting until cuent
market conditions (rent levels, oocupancy rates, space absorption, leasing pre-
commitments) justify construction. This creates a supply lag error due to
‘possible changes in market conditions over the 2:3 year period required for
construction. At the aggregate market level, the backlogs created by this policy
really increase risks by generating costly endogenous cycles (See “the problem’,
below).
A rational expectations commencement policy based on forecasting should.
eliminate cycles, replacing them with random walk enors. However if the
decision meker shifts to a just-in-time inventory decision policy based.on.
forecesting market conditions at a 2-3 year forecast horizon when the project will
be ready foroocupancy, there is a possihility of large forecast errors . Choosing
the optimum time for commencing a project depends on the relative size of these
two kinds of enors.
Decision makers need to understand the trade offs between possible forecast
eqors and market conditions changes (supply lag exors) in order to choosea
commencement timing policy that minimises the expected sum of the two
different types of enor.
Table 1: Comparing two decision policies under two states of nature
Forecast correct Forecast wrong
Stable market Both ok Current conditions better
Market change Forecasting better _| Depends
Table 1 shows that the best decision policy depends on what happens in the
market during the period between project commencement and completion of the
uilding as well as on the quality of forecests.
The problem
Office market cycles are ubiquitous under a wide vatiely of political and.
economic regimes, suggesting a more fundamental cause than local govemment
policies or market conditions. These cycles cause major damage to investor and.
financial intermediary cash flows and balance sheets and contribute to
mmaqcoeconomic problers.
Economists refer to the losses that result from allocating too much capital toa
particular asset class as “allocative inefficiency.” Such inefficiency reduces
aggregate retums on investments and condemns the economy to a Sower growth.
path. Large inefficiencies or misallocations can lead to economic instability,
especially where debt financing is used, as is often the case in real estate
(Minsky, 1974, Fisher, 1933). Major cycles are damaging enough to banks’
belance sheets that they can contzibute to financial crises and macroeconomic
recessions as in the US, (1975, late 1980s), Austialia (early 1990s), Japan (early
1900s-present) and Thailand, Indonesia and Korea (1997 present). The lessons
have not yet been leamed with the next property bubble likely to bustin China
(Australian Financial Review, 1/28/08).
The 1997 collapse of the Thai financial markets, followed by contagion to
Indonesia, Malaysia, Korea, Hong Kong and Taiwan, was triggered by a Thai
yproperty developer's three million dollar default on a payment due ona foreign.
quency (Euro) loan. The default signalled the extent of non-performing assets
‘throughout an overhult real estate sector. “When the Thai property market began
to unravd in 1996 and 1997, the property devdopers could no longer pay back
the cash that they had borrowed from financial institutions. This led to a
snowballing effect of non-performing loans...” (Hanipah, 2008). By 1999, about
half of all firms in Thailand, Indonesia, Malaysia and Korea were unable to meet
current debt repayments. (World Bank, quoted by Hanipah, 2008).
Property cycles have recurred at intervals for more than a century and in many
counties. Stenman’s system dynamics textbook includes a graph of 100 years of
Teal estate cycles and a description of the supply lags that generate endogenous
cydes (Stenman, 2000). A Melboume property crash during the 1880s led to one
of the worst economic collapses in Australian history, lasting a decade. The
Empire State building commenced construction in September 1929, weeks before
the October stock market crash. It was referred to as the “Empty State building’
and did not achieve full occupancy until after Wodd War II.
These cycles have not necessarily gotten less severe over time. The 1986 propaty
oversupply in over 200 U.S. markets caused some of the higgest losses in history
and were intertwined with the savings and loan problems that cost U.S. taxpayers
about $150 billion. Hendershott and Kane (1994) estimated losses (chiefly the
present value of uncollected rents on excess vacant space) at U.S. $130 hillion
during the 1980s U.S. office oversupply cycle that led to vacancy rates over 20%
in huncteds of U.S. cities.
Australia suffered a serious office oversupply cycle in the early 1990s that
contributed to the length and severity of a recession. Vacancy rates in the Perth,
Australia CBD reached 32% of the office stock, an astonishing figure considering
that demand growth had averaged less than 4% per year over the previous decade.
A single project in Perth destroyed one fifth of the capital of the state employees’
pension fund. Rents collapsed from about $350/n? in 1989 to a low of $65 in
1994, and building values throughout the market dropped to less than half of cost.
Australian banks wrote off $28 billion in bad loans in 1993-94 and an educated
guess would be that at least a third would have been due to empty office
buildings.
A lot of money tides on decision makers understanding their “windows af
opportunity” in timing the commencement of major office towers. These
windows tend to be brief because market conditions look similar to all of the
actors who are planning projects. Ina typical market, favourable market
conditions lead to a “race” between a number of competing projects (Grenadier,
1904).
A slighily tongue in cheek artide by two Califomia office market consultants
‘presented a two by two maizix showing developers reactions to good and bad
markets studies of favorable and unfavorable markets. In all four quadrants of
this table, even where a poor quality market study said not to build, the
devdopas’ condusion wes “build, build, build, build.” Devdopas, they
conduded, will build as many projects as they can finance. (Deloy and Rabin,
1974)
The hed mentality of fund managers and financial institution staff therefore,
becomes a sazious problem where all of the money sourves decide in a short time
that “now office projects will make money.” Of course, if they all go ahead with
jaojects the result will be that they all will lose money due to oversupply pushing
down rents and increasing vacancy rates. In Perth in 1994-5 all of thenew
yaojects were either put to lenders, sold for less than half of cost orhald with
impaired cash flows that lasted for a decade before the market recovered. The
‘State Employees pension fund cunrenily has their tower for sale for 45% of its
original real cost. However, projects leased in the mid-1980 showed very
atfractive reuums and projects bought at the trough of the cyde in 1998-96 gave
their owners outstanding capital gains. A Perth project bought for $16 million
sold for $35 million just 4 years later.
Obviously the timing of project commencement decisions is a key issue for office
markes, financial institutions and even in extreme cases for national economies.
Long tem leases “lock in” market conditions at the time of initial lease-up, so
even if the market recovers a building may sill have weak cash flows for 5-10
years or more. Discounting gives these early cash flows considerable weight in
detemining the overall results of the investment. As mle of thumh, if a project is
complded in a favourable leesing environment it will make money (NPV>0)
whileif itis competed in an unfavourable market the building will lose money
(NPV<0).
Identifying windows of opportunity
Office market cycles can be modeled as an inventory control strategic uncertainty
problem analogous to the bear game. Through interviews with leasing agents,
developers, project managers and researchers in the property industry, I collected
alist of overa dozen causes of cycles ( 1997) including prominently:
¢ Asymmmetic information. Some people make money even on projects that
fail, so there are incentives for them to advocate projects that are not
justified by market devand.
* “Prisoner's dilemma.” Strategic uncertainty where developers and
investors don’t know how many compditive projects will proceed means
indivictal rationality (this project will make money) becomes oollective
inationality (too many projects means the market collapses).
¢ — Saiially condated shocks in the macroeconomy. Economists still cannot
call “tums” from boom to recession very well— leading to hig mistakes in
plaming projects with long time lags.
¢ System dynamics. Perhaps the most fundamental cause of endogenous
office market cycles, as demonstrated by a simple system dynamics
modd, is supply lags that lead to backlogs and overshooting of office
supply. This is, essentially, the same inventory control problemas “the
beergams” ina supply chain where mgjor office projects may take nearly
a decade to plan and complete ( , 1998, 1999, 2000).
The problen can be looked at as an “infonmation structure” or system “policies”
problem. Banks and otherinvestors, remembering how badly they were bumed in
the last cycle, tend to attempt to adopt “conservative” policies to avoid losses on
het are perceived (conedily) as highly risky major investments. Believing office
investments to be risky, investors and lenders tend to want to see them justified
by amment market conditions—rents, pre leasing commitments, overall market
space absorption and vacancy rates— before approving finance for construction of
new projects. But tenants are sddom willing to commit to long-term leases at
above market rents. Therefore, by the time rents and leasing pre-comnitments
from tenants rise to levels justifying new construction it is too late. Rents
continue their upwards spike (demand is relatively price inelastic during times of
robust economic growth) which then, based again on the “look at current market
conditions” mentality, in combination with strategic uncertainty and the
asymmehic information incentives arising from the profit centers in large
projects, tends to call forth excess supply.
Tamassuming, of couse, a purely speculative building, thatis, one that is
completed before it is rented. Itis possible to pre lease buildings, but that strategy
meray shifts risks to the tenants, without improving efficiency. Tenants vary in
their ability to undertake such risks. Major tenants (a national telecom ormajor
corporation) could eesily pre-commit to lock in a favourable rent. But then they
faoe the question of when to do so. The project commencement decision has been.
delegated to the tenant, but similar issues arise as to market efficiency and project
timing.
Tt should be pointed out that even if lenders were always prudent and insisted on
leasing commitnents before starting projects, the result oould be inefficient if
Jack of office space were a significant bottleneck restricting economic activity.
High rents and space shortages have negative effects throughout a local economy.
The efficient solution is always for supply and demand to he in balance and for
that to happen, supply lags have to be taken into account. Otherwise just-in-time
inventory is impossile ina growing economy simply because it takes time to
build major buildings. From the local economy’ s point of view, economic activity
and job growth ae probably enhanced by availability of plenty of office space
and lowerrents.
Staff in financial institutions find it very hard to bet large sums of money against
the current market reality. To “get itright” they would have to first say during the
oyde's recovery stage, “Hare is $300 million, go ahead and build a building that
would be worth $200 million at current market rents although there is not enough
demand to fill it” Only by making that kind of decision will projects be
complded early enough for just-in-time inventory, given the two-three years
required for construction. This “conservative” policy leads to a space shortage in
agrowing market during the time buildings are under constuction. As rents (and.
demand from tenants trying to get space quickly in a rising market) spike
‘upwads, it is undoubtedly perceived as an equally career threatening move to say
“No, you can't have $300 million to build a building, even though omment rents
say it will be worth $400 million, rents are continuing to tise sharply and alot of
tenants want space.” This would be the situation during a boom, before supply
hes caught up to demand. In these circumstances, lenders tend to ignore the
implications of cranes on the city skyline or “new supply in the pipdine”
Consavative lending policies designed to reduce tisk can therefore actually
create tisk by calling forth oversupply cycles in response to price volaiility in the
presence of supply lags. Too little supply, too late, sets the soane for oversupply a
few years in the future.
Froma system dynamics point of view, the problem can be solved by investors
and lenders adopting just-in-time inventory policies and changing the infomation
structure of the system. Instead of relying on current infonmation on market
Conditions, the decision should be based on forecasts of market conditions at time
of project compldiion. This implies a 2:3 year forecast horizon, the time required
from ground breaking to completion of major office construction projects and
taking acoount of the price elasticities due to impending new supply from
projects under construction.
Trading off market change versus forecasting risks
Under the “ournent conditions justify project commencement” policy, investors
are exposed to the exor arising from changes in market conditions (other projects
commencing, onset of recession, collapse of a telecom and dot.com boom, etc.)
during the 2:3 year construction period. On the other hand, if they justifysupoly
decisions on the basis of forecasts of conditions ata 2:3 years horizon, they are
subject to forecast enors. In that period of time, the economy could “tum” from
boom to recession, a majorrisk exposure. Forecasting an increase in demand over
a period when demand actually contracts could lead to very serious enors.
From an individual project perspective, the conservative “wait for omrent
conditions” policy is probably rational. But collectively, because it generates
cydes, itis probably indional.
My modd demonstrates that the “current conditions” policy generates cycles.
Cydles are exacerbated by higher trend economic growth, quicker responses to
demand (bunching project commencements), a tendency for the industry to build
more than is justified by demand and the length of the supply lag. Longer lags
lead to longer cycles with higher amplitude and these are explosive cycles even.
when economic growthis a steady, moderate linear trend. The model showed that
cydes moderated dramatically as the lag decreased from 18 months to one year.
This implies that office markets could mitigate cydes and move towards a
yandom walk by using rational expectations and forecasts— just-in-time policies.
A sinple mule that commenced a few projects every year would be more efficient
‘than the current bunching of project commencements. This is a classic
counteriniuitive system dynamics result
But, all of this assumes perfect foresight. My previous papers have explored the
dynamics of systems without incorporating the forecast exors problem. Adding
forecasting exors to the model will make the simulation more realistic. These
simulations will provide important insights on efficient policies for timing of
‘project commencements under uncertainty about future market conditions. If
there are saially condated and large changes in supply and demand conditions
during construction and the forecast gets these wrong, then it is possible that the
“use forecesting for just-in-time inventory” policy would give worse results
(laqger financial losses, more of a mismatch of supply and demand) than the
“wait until justified by current conditions” project commencement policy. The
yandom welk of rational expectations could have larger exors than the cyclical
oversupply exors generated by current conditions policies. However, this seems
unlikely, given the size of the cydles generated by the current conditions policy at
‘the aggregate market levd. Rational expectations might expose individual
jarojects to large erors. Similar forecasting exors by anumber of finns could
‘throw the market out of equilibrium and lead to inefficiency, but itis doubtful
that it would generate cycles.
Depending on assumptions about forecast exrors and changes in market
conditions, one could imagine a range of policies being correct under differing
circumstances. Because forecast exors get much larger at a three year horizon
‘than they would be at a six month horizon, investors might optimise by
forecasting at some shorter horizon than the full 2-3 year constuction horizon, in
effect compromising between “current conditions” and “rely on forecasts”
‘policies by choosing an intermediate forecast horizon. The contribution of system.
dynamics modes can be to quantify this trade-off and help identify optimal
intermediate project timing policies.
The macroeconomy, a fundamental driver of office demand, can be forecast fairly
accurately one quarter ahead, but much less accurately eight quarters (two years)
out Probably by 5 years, the best office demand forecast could not do much,
better than to use the mean of the past time series, so the forecast exor becomes
the variance of the series.
Forecast exrors can be estimated from past time series and from literature on
errors in forecasting the macroecononry. Since exrors would undoubtedly be
highly variable across markets and time patiods, it is perhaps more important to
understand how the errors operate to create changes in optimal project timing
than to wony about whether forecast exors are estimated acamaidy. This would.
best be addressed on a case by case basis. The focus in this paperis on system
dynamics, given particular pattems of forecast exors.
Tt seems likely that if forecast exors increase as a function of the forecast
horizon, and are relatively srvall for the first few quarters, the most efficient
estimates could be based on a forecast horizon somewhere between the
construction horizon and zero. That is, it might be sensible to build the project
besed on “current conditions” expected at anintermediate forecast horizon, say
six months or one year in the future. That would allow a project to “get a jump”
on compaiitors that are cautiously waiting for rising rents to justify construction
‘besed on current conditions, without taking undue risks that demand forecasts
would be very far wrong.
The mocklling exercise therefore, is to explore market efficiency (size of “exors”
changes and forecast exors. The result will be decision polices for optimal
yaoject timing, based on expected market changes and forecast exors.
Back to the Top
CC BY-NC-SA 4.0