2nd Draft: AREUEA, January 4, 2013

The Volatility of Real Estate Markets:

a Decomposition
By

William C. Wheaton
Department of Economics
Center for Real Estate
MIT
Cambridge, Mass 02139
wheaton@mit.edu

The author is indebted to, the MIT Center for Real Estate. He remains responsible for all
results and conclusions derived there from.

JEL Code: R2. Keyword: Housing

ABSTRACT

This research examines how the volatility of real estate markets can be exactly
decomposed into supply and demand side factors that add up. The role of supply depends
keening on its timing relative to demand side shocks. When the correlation between the
two sides of the market is sufficiently positive, supply (historically) has helped to reduce
market volatility. Negative correlations imply that supply always contributes to market
volatility. We compare the overall volatility of 50 metropolitan markets across several
property types (offices, apartments, industrial, and hotels) and find very different
decompositions across the three types. With these decompositions, we then try to
characterize those kinds of markets (within each type of property) whose volatility is
driven more by demand rather than supply factors.

I. Introduction
This research examines the volatility of real estate markets across US metro areas.
The analysis is illustrated for 4 property types, full service hotels, office buildings,
apartments and industrial structures, but it is applicable to any type of real estate with
accurate data. The papers innovation is to use the volatility in vacancy rates (actually the
volatility in vacancy rate changes) rather than volatility in rents, prices or investment
return. There is a long literature linking rental movements to those in vacancy (RosenSmith, 1983), but by focusing directly on vacancy as our metric, market volatility can be
explicitly apportioned into supply and demand side shares. This decomposition can be
done exactly without any econometric application. This decomposition varies across
markets and is systematically different for each property type.
With this analysis, the paper next makes some initial attempt to examine the
correlates of both the overall variance of vacancy as well as the partition of this variance
into demand/supply shares. Smaller markets, and those that are faster growing, as well
as those with constrained land supply seem to have higher volatility in vacancy. In land
constrained markets the demand side also seems more responsible for volatility. The
impacts of a well known index of regulatory constraint are very mixed. Greater
development regulations seem associated with lower vacancy volatility and do not
generate a greater share of volatility coming from the supply side.
II. Literature Review
Real Estate volatility is most often discussed in the context of cycles. The brief
summary here will do only partial justice to the lengthy literature. The literature begins
with articles on the US construction cycle, and in particular that for the housing industry
that emerged during the Post World War II economic boom [Alberts (1962), GreblerBurns (1982), Topel-Rosen (1988)]. Much of this discussion focuses on the supply of
credit to the building industry as the US Federal Reserve Bank fought bouts of economic
inflation.
The discussion then shifts to the demand side of the market. Poterba (1982) argues
that some cyclic movement in real estate prices is inevitable due to building lags as
markets react to demand shocks with forward expectations. Mankiv (1988) provides a
3

counter argument that expectations are highly imperfect since the impact to the housing
market of the baby boom shock occurrs when such children enter the housing market
and not when they were born. Case and Shiller (1989) provide several additional
arguments supporting the view that irrational buyer expectations contribute to market
fluctuations. Stein (1995) suggests that institutional credit constraints inhibit and
encourage buyers rather than irrational expectations.
The discussion then switches to commercial real estate, where a number of authors
note the presence of longer-term building cycles that seem to be unrelated to the broader
economic or credit cycle [Wheaton (1987), Voith and Krone (1988), King and McCue
(1987)]. In a series of papers, Grenadier (1995, 1996) shows that the exercise of
development options could result in building cascades shifting the blame for real
estate oscillations back again to the supply side. Macro and financial economists joined
the debate [Kiotaki-Moore (1996), Childs-Ott-Riddough (1996)] demonstrating that with
collateral constraints, development credit can in fact exhibit herd behavior. The emphasis
on supply as a source of market rigidities and oscillations continues with recent empirical
work to identify housing supply elasticities and link them to natural or institutional
constraints [Harter-Dreiman (2004), Saiz (2010), Gyourko-Saiz-Summers (2008)].
Most recently there are a series of more empirical papers that examine long term
housing price volatility in relation to price growth or return, as well as to rent
fundamentals [ Capozza-Hendershott-Mack (2004), Campbell-Davis-Gallin-Martin
(2009), Cannon-Miller-Pandher (2006).
Two observations seem in order in reviewing this literature. First, the focus has
mostly been on the volatility of asset prices and construction. Just a few papers have
addressed the role of vacancy and its cyclic movements. Secondly, most work has focused
exclusively on one side of the market or the other. There really has been no empirical
attempt to partition or assign responsibility for volatility to both sides of the market. This
paper will begin to fill this gap.

III. Decomposing the Volatility in Vacancy

The analysis begins with a set of definitions of market variables in (1) below.
Using these definitions the last identity (1) is well known.
Vt : vacancy rate
S t : stock of space
OS t : occupied space (demand ex post)

(1)

(1 Vt ) OS t / S t
It is possible to linearly decompose the occupancy rate (1-minus the vacancy rate)
into demand and supply components, but this requires taking the log of the ratio of space
consumption to that of total space supply. The variances of log variables can be quite
different from the variance of the original variable.1 To work with the variance of linear
variables one can use the differences of the variables in (1) and then examine the flows of
supply and demand and their respective impacts on the changes in vacancy. This is done
in (2) by identifying and linking the change in the stock of space, the change in space
consumption (ex post demand growth) and the change in vacancy rate.
Ct : space completions
A t : net space absorption (OS t - OS t 1 )
S t S t 1 Ct

(2)

At Ct (Vt Vt 1 ) S t
Vt Vt 1 Ct / S t At / S t
In the calculations above, net absorption (the growth in ex post demand or
occupied space) is defined as completions minus the change in vacancy in the full stock
(including the new completions). This definition is slightly different from one often used
in the profession the change in total occupied space.2 With the definition above the
bottom identity in (2) holds exactly. Hence there is a simple linear relationship between
the change in the vacancy rate and the new supply rate minus the demand growth rate.

Thelevelofoccupancydecomposesas:log(1Vt)=log(OSt)log(St).

InthisalternativedefinitionAt=Ct(1Vt1)(VtVt1)St.Withthisdefinitionthesubsequent
decompositionofvacancychangeisveryclosebutnotquiteexact.

Figure 1 illustrates the movement in vacancy changes as opposed to vacancy levels for the
San Francisco office market. It is the former that will be decomposed rather than the latter.

SanFranciscoOfficeVacancy:Levels/Changes
25
20
15
10
5
0
5
10 1
9
8
7
.4

1
9
8
8
.4

1
9
8
9
.4

1
9
9
0
.4

1
9
9
1
.4

1
9
9
2
.4

1
9
9
3
.4

1
9
9
4
.4

1
9
9
5
.4

1
9
9
6
.4

1
9
9
7
.4

SFVac

1
9
9
8
.4

1
9
9
9
.4

2
0
0
0
.4

2
0
0
1
.4

2
0
0
2
.4

2
0
0
3
.4

2
0
0
4
.4

2
0
0
5
.4

2
0
0
6
.4

2
0
0
7
.4

2
0
0
8
.4

2
0
0
9
.4

SFVacchange

With the last identity in (2) equation (3) below exactly breaks apart the volatility in
vacancy (or rather vacancy changes) into that due to demand growth (net absorption), that
due to supply growth (completions), and the third (equally important) component the
correlation or covariance between completions and absorption.

2 (Vt Vt 1 ) 2 ( At / St 1 ) 2 (Ct / St 1 ) 2 cov( At / St 1 , Ct / St 1 )

2 ( At / St 1 ) 2 (Ct / St 1 ) 2 ( At / St 1 ) (Ct / St 1 )corr( At / St 1 , Ct / St 1 )

(3)

The third component the covariance between the two sides of the market place
plays a pivotal role. If supply and demand are highly correlated contemporaneously then
there will be little movement in a markets vacancy rate. A change in vacancy results
when the two sides of the market move at different periods in effect a difference in
timing. To better see this, the second expression in (3) is examined in more detail and
then rearranged into (4). Here the variance in vacancy can be divided up into just two
terms: one for demand (absorption) and the other for supply (completions adjusted for
timing).

These two terms can be thought of as the contributory share (to vacancy

variance) from each side of the market.

6

 2 (Vt Vt 1 ) 2 ( At / S t 1 ) 2 (Ct / S t 1 )[1 2corr ( At / S t 1 , Ct / S t 1 )

Demand

( At / S t 1 )
]
(Ct / S t 1 )

(4)

... Supply...

Taking the derivative of the variance in vacancy movements with respect to the
variance in supply, we get expression (5). This is the contribution of a unit increase in
supply variance to the overall variance in vacancy changes or market volatility.
2 (Vt Vt 1 )
( At / S t 1 )
1 corr ( At / S t 1 , Ct / S t 1 )
2
(Ct / S t 1 )
(Ct / S t 1 )

(5)

In (5) its clear that a higher variance of supply increases overall market variance
(a positive derivative above) in two situations. First, when there is any negative correlation
between the two. Secondly, when there is the combination of a positive correlation,
together with a supply variance that is sufficiently large relative to the variance in demand.
On the other hand, greater supply-side variance can actually reduce overall market
vacancy when a) the correlation between sides of the market is positive and b) the
variance in completions is sufficiently small relative to that in demand. This latter case is
particularly insightful, for it reveals that supply actually can be a friend of investors in
certain situations by dampening market volatility. Creating new development when it is
needed (as opposed to not-needed) is the key to keeping vacancy smooth and even. With
smooth vacancy, rents, income and investment return should all be smoother as well.
To further illustrate the role of supply, one might assume that the variance in
absorption comes from true demand side random shocks while the variance in supply
depends on the endogenous pattern of market response. This perception of how market
volatility arises is certainly quite common in the literature. In this case, long run
absorption and completion rates must be approximately equal since vacancy rates in
general are stationary and without a long run trend. Thus for purposes of illustration,
equation (4) can be simplified to (6) - if the cyclic variances of each side of the market are
assumed to be equal (here to that in absorption). In this case, market volatility (variance in
vacancy) depends completely and exclusively on the timing of supply - that is the
correlation between completions and absorption).
7

 2 (Vt Vt 1 ) 2 ( At / S t 1 )[2 2corr ( At / S t 1 , Ct / S t 1 )] with equal variances

(6)

In the unlikely situation where supply can respond immediately and so is

completely timed with demand we have a perfect positive correlation (+1.0) between the
two sides of the market and the right hand side of (6) collapses to zero. With high positive
correlation supply can totally eliminate or greatly reduce the impact of demand shocks
to vacancy.
When the timing of the supply response is random with respect to demand shocks,
(a correlation of zero) then the variance in market vacancy is simply the sum of each or
in (6) twice the original variance in demand that is generated by the shocks. It is at this
point where supply variance begins to contribute to overall market variance rather than
reducing it. Moving on, as the pattern of supply response generates a negative correlation
with demand shocks supply becomes a serious additional source of market volatility and
exacerbates overall vacancy variance. In the extreme case where the supply response
results in a perfect negative correlation (-1.0), overall market variance increases to a
maximum of four times the original variance from the demand side.
Recapitulating, if both sides of the market have equal variances with demand
coming from shocks while supply results from the response pattern, then market volatility
in (6) will range from zero (with perfect positive supply correlation) to 2x demand
variance (with zero correlation) on up to 4x demand variance (with perfect negative
correlation).
This discussion makes it clear about what should be examined empirically to
analyze the determinants of market volatility. First, examine the variance in vacancy, the
correlation coefficient between absorption and completions, and then finally use the RHS
of (4) to determine the direction and share that supply contributes to market volatility as
opposed to demand. This decomposition is undertaken for 4 property types using quarterly
time series that span approximately 22 years.

IV. Decomposing Volatility in Metropolitan Office Markets.

The data for the decomposition of office vacancy consists of quarterly time series
spanning the years 1988:1 through 2010:1, from CBRE. The series available are on
vacancy, stock and completions - for 50 US metropolitan areas. To avoid confounding
seasonal variation with cyclic fluctuations, the calculation of completions, vacancy
changes and absorption are all done on a year-over-year basis. This provides 88
overlapping observations for each MSA. Table 1 below presents for each market the
overall variance in vacancy movement, the correlation between the completions and
absorption series and the share of vacancy variance that is due to demand as opposed to
supply sides of the market. If the share is multiplied by the variance in vacancy changes
(the first column) one gets the demand and supply side contributions on the RHS of (4).
In Table 1 the average correlation between the two sides of the market is very
close to zero (.02), and on average, supply accounts for about a third of market volatility.
But the averages hide two distinct patterns. First, there are 10 markets (e.g. New York,
Riverside) in which the positive correlation between completions and absorption is strong
enough (generally greater than .50) so that the share of market volatility that is attributable
to demand is greater than 1.0. In this case the supply contribution is actually negative. In
these markets, the variance in vacancy would be greater if supply was simply always
constant! In other words supply is actually helping to dampen the impact of demand
shocks on overall market volatility. Secondly, there are 10 markets where the contribution
of supply shocks to market volatility is at least 40% or more, and in these markets the
correlation between supply and demand is always less than 0.20 and often negative.
These are markets (e.g. Miami, Boston) where a largely independent supply variance is
contributing quite substantially to overall market volatility.
IV. Decomposing Volatility in Metropolitan Hotel Markets.

To examine the Hotel market, data was obtained from Smith Travel Research
covering 50 MSA, over the same period: 1988:1 to 2010:1. The MSA however were
slightly different. Remember that all calculations are done with year-over-year changes so

Table 1: Office Market Decompositions

M a rk e t
ALBU Q U
ATLAN T
A U ST IN
B A L T IM
B O ST O N
C H IC A G
CHRLTE
C IN C IN
CLEVEL
CO LU M B
DALLAS
DEN VER
DETRO I
FO R TLA
FO R TW O
H AR T FO
HO N O LU
H O U STO
IN D IA N
JA C KSO
KAN SAS
LAN GEL
L ISL A N
LVEGAS
M IA M I
M IN N E A
N ASH V I
N EW ARK
N EW YRK
O AKLAN
O RAN GE
O RLAN D
O XN ARD
P H IL A D
PHO EN I
PO RTLA
R IV E R S
SAC R AM
SALTLA
SAN TO N
SD IE G O
SEATTL
SFR AN C
SJO SE
SL O U IS
STAM FO
TAM PA
TU C SO N
W A SH IN
W BEACH
A v e ra g e

v a c . V a r ia n c e c o r r e la t io n D e m a n d % s u p p ly %
0 .0 0 0 7 9 8 0 .0 1 6 7 5 2 6 0 .6 6 3 0 1 4 3
0 .3 3 7 0
0 .0 0 0 6 9 5 0 .4 2 9 3 5 3 1 0 .8 9 5 8 1 1 6
0 .1 0 4 2
0 .0 0 2 3 8 2 0 .1 5 9 3 6 7 0 .4 7 5 3 8 8 8
0 .5 2 4 6
0 .0 0 0 5 3 5 0 .3 8 8 4 8 7 9 0 .7 2 0 9 4 5 9
0 .2 7 9 1
0 .0 0 0 7 7 6 0 .1 8 3 0 2 5 0 .5 7 9 0 7 5 6
0 .4 2 0 9
0 .0 0 0 4 3 0 0 .3 0 4 5 4 5 7 0 .8 4 6 2 8 3 8
0 .1 5 3 7
0 .0 0 0 7 5 6 0 .5 2 0 7 0 2 7
0 .9 1 6 8 1 5
0 .0 8 3 2
0 .0 0 0 4 5 6 0 .5 0 1 0 9 0 4 0 .9 1 4 4 5 4 5
0 .0 8 5 5
0 .0 0 0 4 2 6 0 .3 3 9 2 5 9 1 0 .7 0 7 0 2 6 4
0 .2 9 3 0
0 .0 0 0 6 7 9 0 .2 1 8 5 4 7 4 0 .5 2 6 7 0 3 8
0 .4 7 3 3
0 .0 0 0 6 3 7 0 .0 0 9 6 1 3 0 .5 2 8 4 3 0 1
0 .4 7 1 6
0 .0 0 0 7 2 3 0 .0 3 2 1 3 0 1 0 .5 8 5 6 5 9 9
0 .4 1 4 3
0 .0 0 0 4 9 9 0 .4 0 6 5 8 0 5 0 .7 3 6 6 3 3 3
0 .2 6 3 4
0 .0 0 1 2 1 4
0 .0 8 2 4 4 1
0 .6 7 4 4 0 5
0 .3 2 5 6
0 .0 0 0 4 1 5 0 .4 1 6 8 3 5 8 0 .9 5 5 8 6 9 8
0 .0 4 4 1
0 .0 0 0 9 0 1 0 .1 7 2 9 0 8 5 0 .3 8 8 0 1 7 4
0 .6 1 2 0
0 .0 0 0 2 9 4 0 .8 7 9 6 8 0 6 4 .7 2 3 1 4 4 9
3 .7 2 3 1
0 .0 0 0 4 7 7 0 .2 5 0 9 1 7 0 .6 0 3 3 9 2 9
0 .3 9 6 6
0 .0 0 0 5 4 1 0 .5 2 2 6 4 1 3 0 .9 2 6 5 0 8 5
0 .0 7 3 5
0 .0 0 0 6 1 9 0 .5 1 9 5 0 3 2 1 .0 7 8 0 6 7 3
0 .0 7 8 1
0 .0 0 0 4 9 4 0 .2 7 0 1 5 3 7 1 .0 3 8 4 3 0 1
0 .0 3 8 4
0 .0 0 0 3 7 6 0 .4 4 3 2 5 5 8 0 .9 5 7 0 6 8 5
0 .0 4 2 9
0 .0 0 0 5 4 2 0 .2 4 8 7 5 7 6 0 .5 4 2 1 8 5 9
0 .4 5 7 8
0 .0 0 1 1 8 4 0 .6 0 9 7 4 6 7 1 .2 0 8 4 8 2 3
0 .2 0 8 5
0 .0 0 0 7 6 4 0 .2 2 3 3 5 5 0 .6 9 2 3 0 0 2
0 .3 0 7 7
0 .0 0 0 7 2 1 0 .1 8 5 4 0 3 3
0 .6 4 0 5 0 9
0 .3 5 9 5
0 .0 0 0 5 8 7
0 .4 1 6 4 7 4 0 .7 5 8 4 2 5 8
0 .2 4 1 6
0 .0 0 0 4 6 8 0 .4 7 2 0 3 9 8 1 .1 7 0 4 0 9 6
0 .1 7 0 4
0 .0 0 0 3 9 4 0 .5 4 6 3 0 0 8 1 .4 3 4 2 4 7 3
0 .4 3 4 2
0 .0 0 0 7 0 0 0 .2 6 0 9 5 7 3 0 .9 2 2 7 5 0 1
0 .0 7 7 2
0 .0 0 0 9 4 1 0 .4 3 1 2 9 7 3 0 .9 0 8 6 8 5 1
0 .0 9 1 3
0 .0 0 0 8 1 1 0 .6 3 5 4 1 8 7 1 .0 7 9 3 7 8 3
0 .0 7 9 4
0 .0 0 1 6 8 9 0 .1 3 5 0 9 8 7 0 .6 8 7 9 3 6 3
0 .3 1 2 1
0 .0 0 0 3 5 8 0 .4 5 4 5 6 1 4 0 .9 4 3 1 4 0 6
0 .0 5 6 9
0 .0 0 1 3 1 2 0 .0 5 3 2 6 7 2 0 .3 2 8 5 4 6 5
0 .6 7 1 5
0 .0 0 0 6 6 3 0 .0 2 8 3 3 7 0 .6 8 4 3 6 8 6
0 .3 1 5 6
0 .0 0 1 0 7 6 0 .8 2 1 0 3 5 3 2 .7 1 6 7 8 3 3
1 .7 1 6 8
0 .0 0 0 5 6 4 0 .6 6 8 5 7 2 1 1 .6 5 9 0 7 8 2
0 .6 5 9 1
0 .0 0 1 0 9 8 0 .2 4 8 2 4 1 8 0 .7 6 2 8 6 2 1
0 .2 3 7 1
0 .0 0 0 6 2 2 0 .1 5 4 3 1 2 0 .6 5 9 4 5 1 7
0 .3 4 0 5
0 .0 0 0 7 5 9 0 .5 3 3 5 9 5 7 1 .0 3 6 5 2 4 7
0 .0 3 6 5
0 .0 0 0 9 0 7 0 .4 5 3 9 4 9 1 0 .8 4 5 6 8 6 4
0 .1 5 4 3
0 .0 0 1 3 0 1 0 .3 5 9 9 3 3 0 .5 9 0 7 6 2 5
0 .4 0 9 2
0 .0 0 2 1 9 7 0 .1 4 3 7 5 7 0 .5 3 8 1 7 2 6
0 .4 6 1 8
0 .0 0 0 5 1 8 0 .1 8 3 3 6 5 7 0 .5 9 6 2 3 9 2
0 .4 0 3 8
0 .0 0 0 7 0 2 0 .1 0 9 5 5 8 1 0 .7 4 3 2 0 0 4
0 .2 5 6 8
0 .0 0 0 7 8 0 0 .3 1 6 4 7 2 5 0 .6 5 6 1 4 0 8
0 .3 4 3 9
0 .0 0 0 6 4 4 0 .1 6 8 9 7 8 3
0 .7 4 9 5 5 8
0 .2 5 0 4
0 .0 0 0 4 7 0 0 .4 0 5 2 4 3 6 0 .8 0 3 6 4 8 2
0 .1 9 6 4
0 .0 0 1 1 8 7 0 .0 2 9 9 4 2 4 0 .6 3 3 0 9 8 8
0 .3 6 6 9
0 .0 0 0 9 9 2 0 .0 2 3 3 4 7 5 0 .6 4 8 0 5 6 5
0 .3 5 1 9

10

Table 2: Hotel Market Decompositions

Market
vac variance
ALBUQU
0.00217282
ATLANT
0.00183454
AUSTIN
0.00296046
BALTIM
0.00138182
BOSTON
0.00214849
CHICAG
0.00186902
CHRLTE
0.00299764
CINCIN
0.00137673
CLEVEL
0.0016834
COLUMB
0.0014359
COLUSC
0.00268683
DALLAS
0.00191561
DAYTON
0.00267172
DENVER
0.00148274
DETROI
0.00321505
FORTLA
0.00225649
FORTWO
0.00343616
HARTFO
0.00257621
HONOLU
0.00319439
HOUSTO
0.00212315
INDIAN
0.00115846
KANSAS
0.00142719
LANGEL
0.00204192
LISLAN
0.00331032
MEMPHI
0.00177024
MIAMI
0.0027686
MINNEA
0.00182749
NASHVI
0.0019432
NEWARK
0.0027523
NEWORL
0.00298223
NEWYRK
0.00215512
OAKLAN
0.00304129
OMAHA
0.00245661
ORANGE
0.00212325
ORLAND
0.0025132
PHILAD
0.00154526
PHOENI
0.00174925
PITTSB
0.00125968
PORTLA
0.0018355
RALEIG
0.00261297
RICHMO
0.00198655
SANTON
0.00136339
SDIEGO
0.00189932
SEATTL
0.0016184
SFRANC
0.00283149
SJOSE
0.0044589
SLOUIS
0.00122284
TAMPA
0.00081844
TUCSON
0.00148912
WASHIN
0.00114781
WBEACH
0.00189822
Average
0.00203552

correlation
demand%
-0.12752566 0.502457
0.201920988 0.952515
0.009229604 0.627874
0.209145037 0.519514
-0.1882602
0.77902
-0.10515525 0.650259
0.311466889
0.70412
0.161993945 0.691965
0.032326636
0.62457
0.161147142 0.738794
0.435971507 0.689736
-0.11031189 0.795812
0.33963587 0.450488
0.235525314 0.966076
-0.15864869 0.399827
0.337660788 1.074817
0.457971541 0.731631
0.416406681 0.753337
0.176438201 0.992952
-0.03374959 0.744255
0.414067072 0.939579
0.238013107 0.775667
-0.03863364 0.746075
-0.06538912 0.390378
0.208584805 0.797817
0.218203368 0.896878
-0.15579648 0.647605
0.47409533 0.743327
0.082245527
0.49775
0.623476183 1.262182
0.658730594 1.369632
-0.20767912 0.574709
0.225419788 0.478092
0.04590885 0.624971
0.339898403 1.079599
0.182073028 0.629991
0.286180555 0.866802
0.322834856 0.910278
0.043370814 0.580563
0.284882262 0.432802
0.177663755
0.66023
0.495412634 1.146066
0.116603182 0.639901
-0.09081322 0.667847
0.024921384 0.869853
0.127838566 0.832556
0.471957935 0.912359
0.491660607 1.325771
0.254848196 0.965785
0.031417941
0.79407
0.331653284 0.832301
0.102063809 0.667379

11

supply%
0.496325
0.049604
0.372224
0.483311
0.219784
0.348798
0.300817
0.310108
0.375785
0.263223
0.318391
0.203411
0.554501
0.036601
0.598691
-0.070415
0.277294
0.254377
0.008478
0.255443
0.0681
0.227664
0.253583
0.608964
0.204921
0.105782
0.351086
0.266154
0.503221
-0.243953
-0.348401
0.423538
0.524947
0.375542
-0.075178
0.372433
0.137386
0.094699
0.419927
0.571134
0.342114
-0.13555
0.361522
0.331336
0.130341
0.168823
0.097432
-0.318219
0.037301
0.206224
0.173073
0.334699

that the well known hotel seasonality does not impact the analysis. With this in mind
average Hotel vacancy volatility is twice that of offices (.002 versus .001) but the
correlation between the two sides of the market is somewhat higher (.10 versus .02).
Supply in this sector on average has almost the same contribution to overall volatility as in
the case of offices: 33% versus 35%. Examining the same two patterns in markets as with
offices we find that there are only 6 markets (e.g. Fort Lauderdale) in which the demand
share exceeds one and hence where supply helps to reduces volatility. At the other
extreme there are 20 markets (e.g. Chicago) where the correlation is so low or negative
(average .03) that the supply side of the market is contributing at least 33% to overall
market volatility.

V. Decomposing Volatility in Metropolitan Apartment Markets.

In the case of Apartments data was used from MPF Research, which again covered
the same 50 MSA as the office data, over the identical period: 1988:1 to 2010:1. In
apartments the average vacancy volatility is tiny relative to the other commercial property
types (.00029 versus .001 or .002). Partly explaining this, the average correlation between
the two sides of the market is far greater, .47 versus .02 or .07. With this higher
correlation, supply in the apartment sector has an average -.16% contribution to volatility.
Examining the same two patterns of markets as with offices we find that in almost 80% of
the markets (39), there is a negative contribution of supply to volatility. The correlation
between the two sides of the market is strong enough and supply volatility small enough
relative to demand so that supply helps reduce market volatility in the vast majority of
markets. At the other extreme, of the 11 markets in which supply exacerbates volatility,
the average contribution is less than 10%. The apartment market is much better
coordinated than the market for Hotels or Office buildings.

VI. Decomposing Volatility in Metropolitan Industrial Markets.

The industrial market is surprisingly similar to the Apartments market it is very

well coordinated. The data here is again from CBRE and covers the same period

12

(1988:1 through 2010:1) , but in this case spans only 32 markets. The average variation in
vacancy changes is .00036 very similar to the .00029 of apartments and roughly 1/7th as

Table 3: Apartment Market Decompositions

mkt
ATLANT
AUSTIN
BALTIM
BOSTON
CHICAG
CHRLTE
CINCIN
CLEVEL
COLUMB
DALLAS
DENVER
DETROI
EDISON
FORTLA
FORTWO
HOUSTO
INDIAN
JACKSO
KANSAS
LANGEL
LOUISV
LVEGAS
MEMPHI
MIAMI
MINNEA
NASHVI
NEWARK
NEWYRK
NORFOL
OAKLAN
ORANGE
ORLAND
PHILAD
PHOENI
PITTSB
PORTLA
RALEIG
RICHMO
RIVERS
SACRAM
SALTLA
SANTON
SDIEGO
SEATTL
SFRANC
SJOSE
SLOUIS
TAMPA
WASHIN
WBEACH
average

vvac
0.000291
0.000408
0.000249
0.000116
0.000167
0.000379
0.000339
0.000494
0.000322
0.000236
0.000319
0.000225
0.000127
0.000267
0.000215
0.000519
0.000450
0.000457
0.000326
0.000188
0.000500
0.000347
0.000334
0.000595
0.000201
0.000291
0.000138
0.000021
0.000408
0.000138
0.000177
0.000339
0.000451
0.000278
0.000387
0.000305
0.000246
0.000321
0.000246
0.000223
0.000370
0.000275
0.000121
0.000258
0.000234
0.000379
0.000783
0.000168
0.000105
0.000295
0.000293

corl
0.409277
0.528545
0.187073
0.251737
-0.045356
0.570692
0.262897
0.113784
0.470214
0.434858
0.137856
0.546545
0.309416
0.626796
0.339215
0.085040
0.441793
0.281303
0.445192
0.374128
0.121786
0.943220
0.391334
0.395056
0.478015
0.379155
0.078099
0.453449
0.360253
0.646693
0.729671
0.771828
0.004089
0.449426
0.204888
0.512561
0.642842
0.381613
0.725868
0.555484
0.499295
0.478691
0.758724
0.382370
0.074102
0.085850
0.094295
0.630453
0.260722
0.546588
0.477933

demand
0.910642
1.118101
1.000416
1.007396
0.899333
1.137648
1.068630
1.009310
1.274775
1.147213
0.705990
1.394853
1.074469
1.550786
1.029535
0.881059
1.196130
0.861025
1.243722
1.113637
1.007113
9.010295
1.121010
1.179784
1.290554
1.060405
0.974297
1.174239
1.148864
1.678698
2.135796
2.327951
0.969631
1.186184
1.040127
1.227185
1.488624
1.149831
2.033492
1.417603
1.252094
1.135646
2.355609
1.028877
0.979659
0.926461
0.993977
1.609461
0.964730
1.421547
1.166095

13

supply
0.089358
-0.118101
-0.000416
-0.007396
0.100667
-0.137648
-0.068630
-0.009310
-0.274775
-0.147213
0.294010
-0.394853
-0.074469
-0.550786
-0.029535
0.118941
-0.196130
0.138975
-0.243722
-0.113637
-0.007113
-8.010295
-0.121010
-0.179784
-0.290554
-0.060405
0.025703
-0.174239
-0.148864
-0.678698
-1.135796
-1.327951
0.030369
-0.186184
-0.040127
-0.227185
-0.488624
-0.149831
-1.033492
-0.417603
-0.252094
-0.135646
-1.355609
-0.028877
0.020341
0.073539
0.006023
-0.609461
0.035270
-0.421547
-0.166095

Table 4: Industrial Market Decompositions

msa
ATLANT
BALTIM
BOSTON
CHICAG
CHRLTE
CINCIN
COLUMB
DALLAS
DENVER
HARTFO
HOUSTO
INDIAN
JACKSO
KANSAS
LANGEL
LISLAN
MEMPHI
MIAMI
MINNEA
NASHVI
OAKLAN
ORANGE
PHILAD
PHOENI
PORTLA
RIVERS
SDIEGO
SEATTL
SFRANC
SJOSE
SLOUIS
WASHIN
average

vvac
0.000318
0.000401
0.000599
0.000135
0.000401
0.000206
0.000144
0.000318
0.000275
0.000457
0.000261
0.000229
0.000499
0.000236
0.000138
0.000425
0.000570
0.000328
0.000216
0.000312
0.000360
0.000298
0.000284
0.000578
0.000263
0.000542
0.000218
0.000210
0.000350
0.001329
0.000228
0.000410
0.000364

corl
0.52454567
0.513264212
0.155423928
0.533547782
0.416162219
0.455962965
0.471887201
0.286482631
0.280840007
0.309219498
0.184206308
0.417767125
0.158746838
0.427748021
0.43504988
-0.04854842
0.608598377
0.350897775
0.618704371
0.303410624
0.565416616
0.393208132
-0.09791656
0.350543094
0.567246277
0.886482023
0.781106898
0.569508345
0.065286182
-0.00751774
0.234898777
0.456530936
0.490538

demand
1.313857835
1.356304116
0.98230287
1.380941217
1.198393262
1.237537842
1.248031134
0.949581758
0.975149735
1.073432273
0.941845121
1.205573263
0.895668444
1.223812939
1.169570014
0.914281072
1.587853674
1.124569659
1.607457711
1.018534755
1.464718377
1.147097551
0.821708858
0.993697662
1.470495426
2.785046696
2.228142535
1.47002787
0.982115044
0.911718673
1.043622266
0.997817737
1.155838

supply
-0.313857835
-0.356304116
0.01769713
-0.380941217
-0.198393262
-0.237537842
-0.248031134
0.050418242
0.024850265
-0.073432273
0.058154879
-0.205573263
0.104331556
-0.223812939
-0.169570014
0.085718928
-0.587853674
-0.124569659
-0.607457711
-0.018534755
-0.464718377
-0.147097551
0.178291142
0.006302338
-0.470495426
-1.785046696
-1.228142535
-0.47002787
0.017884956
0.088281327
-0.043622266
0.002182263
-0.155838

important as hotel or office markets. The average correlation between absorption and
completions is the highest of any property type (.49) and the average share of volatility
due to demand is 1.15%. With this correlation, supply clearly helps to reduce the volatility
in the industrial market by - by almost 15%. In 65% of the markets (21), there is a
negative contribution of supply to volatility. At the other extreme, of the 11 markets in
which supply exacerbates volatility, the average contribution is less than 5%.

14

VI. Cross Section analysis of Decompositions.

A first, and obvious, question that arises is why Hotels are so similar to Offices
and why these two in turn are so different from Industrial and Apartment markets. Since
there are only four observations one can only speculate. If land is plentiful, development
restrictions few and building lags short then in theory supply should be better
coordinated with demand. What also might help coordination is the absence of
speculative development wherein buildings are constructed without pre-leasing
commitments from tenants. Many industrial buildings for example are built to suit in
which the development is done directly for the buyer/occupier. Similarly many apartment
buildings are developed directly by larger apartment landlords who tend to phase their
developments carefully with their perceived market demand.
Within each property type, however, it should be possible to empirically study the
variation in coordination and volatility across markets. This is not trying to make a
statement about causal inferences, but rather simply trying to identify systematic patterns
as to which markets are more volatile, which are better coordinated, and which have a
relatively stronger contribution from the demand side of the market as opposed to supply.
To do this, 4 cross section regressions are undertaken. The first is to explain overall
market vacancy variance. This is followed by the contribution from the demand and
supply sides of the market. Finally, we try and explain the level of coordination
(correlation) between absorption and completions across the markets.
As for covariates, six readily available MSA variables are used. The size and
growth rate of a local economy might be expected to impact overall volatility. In Wheaton
(1991) real estate stock-flow dynamic models were shown theoretically to be more likely
to have internal oscillations when demand growth was rapid. Size is measured with total
employment (in 2010) and growth with its cumulative change over the sample period
(1988:1-2010:1). It might also be instructive to examine market demographics: average
income/worker and the ratio of population/employment (the inverse of Labor Force
Participation). These are measured at the end of the sample period (2010:1). It is not easy
to identify any priors as to the impact of these variables. On the other hand there has been
much discussion about the possible role that a markets supply elasticity could play in
generating market volatility. Mayer (2000) argues that larger (monocentric) cities
15

intrinsically have more inelastic land supply. There are two actual metrics of supply
restrictions readily available: the Wharton-Lurie Index of procedural restrictions (2008),
and the Saiz index (2010) of land constraints.
There are four regressions in Table 5, for office properties and then apartments.
The cross section analysis was restricted to just these two wherein the sample of cities is
almost the same. Hotels and Industrial properties have a very different sample, making
them not strictly comparable. In each equation, the RHS variables described above are
regressed against the variance in the vacancy rate, the variance in the absorption rate
(demand volatility), the correlation-adjusted variance in the completion rate (the supply
contribution) and the absolute correlation between completions and absorption.3
Market size (total employment) is associated with lower vacancy variance in both
property types, but in neither is there a clear explanation. Larger markets generally
experience less volatile demand, but also less supply contribution and lower correlation,
but all of these effects are not statistically significant.
A stronger result emerges for the impact of long term employment growth, which
always is associated with higher demand variance. In Apartments, it is also associated
with a lower supply contribution which then negates the higher demand variance leading
to little overall impact on vacancy volatility. With office properties, however, faster
growing markets also have a slightly higher supply contribution so the impact on overall
vacancy variance is clearly positive.
Moving to an areas demographic/economic makeup, higher income is associated
with greater office vacancy variance with most of the impact originating from the supply
side. On the other hand, higher income areas have lower apartment vacancy volatility,
with no particular assignment of responsibility between the two sides of the market. Areas
with an aging population and lower labor force participation (a higher ratio of
population/employment) are associated with much higher demand variance in the case of
office properties, which is largely offset by such areas having a much a lower variance

Fromequation(4)thecoefficientofanyvariableintheabsorptionequationplusthatvariables
coefficientintheadjustedcompletionrateequationwillequalthatinthevacancyvarianceequation.This
relationshipdoesnotholdforthestandarderrorsofthevariablescoefficientacrossthethreeequations.

16

contribution from the supply side. In the apartment sector, areas with an aging population
have essentially no impact on vacancy volatility from either side of the market.
The results for the market supply variables are often counter-intuitive. For office
properties, areas with a high regulatory score (more restrictive) have both higher demand
side variance and a lower variance contribution from supply. On net such areas
experience little difference in overall vacancy variance. In the case of apartments, more
restrictive areas behave in reverse with a higher supply side contribution to variance,
that is more than offset by such areas being associated with lower demand variance.
Turning to the land variable, areas with constrained land supply are associated with
greater demand side variance for both office properties and apartments. For office
properties constrained areas also have a larger supply contribution so such areas clearly
experience greater vacancy volatility. In the apartment sector, constrained areas have a
smaller contribution from the supply side, so the overall impact of land constraints on
apartment vacancy is more ambiguous.

Table 5: Cross Section Determinants of Decomposition (t statistics in

parenthesis)
Equation

Vacancy
Variance

Absorption
Variance

Supply
contribution

correlation

R2

0.36

0.51

0.25

0.13

Constant

1.7e-004 (.38)

-8.1e-004(1.99)

9.9e-004 (1.93)

.26 (2.33)

Employment

-7.7e-008 (-1.4) -4.7e-008(-.95)

-3.1e-008(-.49)

1.7e-005(.42)

Emp. Growth

2.1e-004 (3.27)

1.6e-004 (2.79)

.51e-004 (.69)

-.03 (.61)

Inc/Employment

5.5e-009 (1.62)

-6.6e-01 (-.21)

6.1e-009(1.61)

-4.9e-006(-1.90)

Offices

Pop/Employment -1.3e-004 (-.78) 5.5e-004 (3.34)

-6.7e-004(-3.34) .23 (1.73)

WLURI

-3.1e-005 (-.3)

1.6e-004 (1.86)

-1.9e-004(-1.74) .08 (1.02)

Land constraint

5.4e-004 (1.96)

3.3e-004 (1.34)

2.1e-004 (.67)

17

-.09 (.45)

Apartments

Vacancy
Variance

Absorption
Variance

Supply
Contribution

Correlation

R2

0.33

0.36

0.32

0.45

Constant

6.4e-004(3.66)

2.5e-004 (.48)

3.9e-004 (.77)

-.18 (-.72)

Employment

-4.1e-008(-1.9)

-5.1e-009 (-.08) -3.5e-008 (-.61)

Emp. Growth

-1.8e005(-.85)

2.1e-004 (3.77)

-2.3e-004(-3.71) .11 (3.53)

Inc/Employment

-2.8e-009(-1.9)

-1.1e-009(-.25)

-1.6e-009(-.38)

-1.3e-006(-.62)

Pop/Employment -3.3e-006(-.05)

4.7e-005 (.23)

4.4e-005 (.24)

.22 (1.93)

WLURI

-5.9e-005(-1.5)

-1.8e-004 (-1.5) 1.2e-004 (1.08)

Land constraint

9.3e-005(1.06)

5.1e-004 (1.92)

-7.6e-006(-.26)

-.027 (-.45)

-4.2e-004(-1.61) .24 (1.91)

VI. Conclusion

This paper has been somewhat exploratory in nature, developing a simply way of
decomposing the variance in a markets vacancy rate into that due to absorption (demand)
and that due to construction (supply). From the well-known variance formula, a key
ingredient in determining the contribution from construction (supply) is its intertemporal
correlation with demand. When buildings are constructed at a time close to when
absorption is high then the variance in market vacancy can be little or none. If
construction occurs at times orthogonal from demand, then the supply effect is additive. If
development occurs when absorption (demand) is not growing then the supply side of the
market greatly exacerbates market volatility.
Decomposing the variance in 50 markets for 4 types of property reveals some
interesting patterns. There are stark average differences between the property types
(offices, apartments, hotels, industrial buildings). As for the differences across markets
there is noticeable variation, but it is not well-explained. Economic growth is clearly
associated with greater volatility that appropriately originates from absorption. The well
known WLURI regulatory index, however, yields very mixed results that are almost the
18

opposite for offices as opposed to Apartments. A newly developed index of geographic

land constraints is associated with higher volatility mostly from the demand side
which would seem to make sense if such constraints are thought of as increasing the
inelasticity of land supply for new development. Hopefully the methodology developed
here will see continued application.

19

References

Alberts, W. W.(1962), "Business Cycles, Residential Construction Cycles, and the

Mortgage Market," The Journal of Political Economy, LXX,1, 263-281.
Campbell, S., M. Davis, J. Gallin, R. Martin ( 2009), What moves Housing Markets: A
variance decomposition of the Price-Rent Ratio, Journal of Urban Economics 66(2), 90102.
Cannon, S., N. Miller, G. Pandher (2006),Risk and Return in the US Housing Market: A
Cross-Section Asset-Pricing Approach, Real Estate Economics, 34, 4, 519-552.
Case, Karl and Robert Shiller (1989), The Efficiency of Market for Single-Family
Homes, American Economic Review, Vol79, no.1, 125-137.
Capozza, D., P. Hendershott, C. Mack (2004), An Anatomy of Price Dynamics in Illiquid
Markets: Analysis and Evidence from Local Housing Markets, Real Estate Economics,
32, 1, 1-21.
Childs, P, Ott, S, and Riddiough,T, (1996), "Incomplete Information, Excise Policy and
Valuation of Claims on Noisy Real Estate Assets", MIT Working Paper, Center for Real
Estate.
DiPasquale, D. , Wheaton, W.C., (1996), Urban Economics and Real Estate Markets,
Prentice Hall, New Jersey.
Glaeser,E., Gyrouko,J. (2005), Why Have Housing Prices Gone Up?, NBER Working
Paper 11129.
Gabriel, S., F. Nothaft,(2001) "Rental Housing markets: Incidence and Duration of
Vacancy, and the Natural Vacancy Rate", Journal of Urban Economics, 49,1, 121-150.
Grebler, L., and W. Burns,(1982), "Construction cycles in the U.S.",AREUEA Journal,
10,2, 201-222.
Grenadier, S. (1995), "Local and National Determinants of Office Vacancies", Journal of
Urban Economics, 37,1, 57-72.
Grenadier, S. (1995), "The Persistence of Real Estate Cycles", Journal of Real Estate
Finance and Economics, 10,2, 95-121.
Grenadier, S. (1996), "The Strategic Exercise of Options: Development Cascades and
Overbuilding in Real Estate Markets", Journal of Finance, 51,5, 1653-1679 .
Gyourko,J., A. Saiz, A. Summers, (2008), A New Measure of the Local Regulatory
Environment for Housing Markets.., Urban Studies, 45,3, 693-729.
20

Harter-Dreiman,M. (2004) Drawing Inferences about Housing Supply Elasticity from

house price responses to income shocks, Journal of Urban Economics, 55, 2, 316-337.
Hendershott, P.H, and E. Kane,(1995), "Office Market Values during the Last Decade,
How Distorted Have Appraisals Been?", Real Estate Economics, 23,2, 101-117.
King, J. and T. McCue (1991), Office Building Investment and the Macroeconomy,
Real Estate Economics, 15,3, 234-255.
Kiyotaki, N. and John Moore, (1997), "Credit Cycles", Journal of Political Economy, 105,
2, 211-248.
Mankiw, G. and D. Weil, (1989),The Baby Boom, the Baby Bust, and the Housing
Market, Regional Science and Urban Economics, 19(2), 235-258.
Mayer, C. and T. Somerville,( 2000), Residential Construction: Using the Urban Growth
Model to estimate Housing Supply, Journal of Urban Economics, 48, 5, 85-109.
Ortalo-Magne, F. and S. Rady , 2006, Housing Market Dynamics: On the contribution of
Income Shocks and Credit Constraints, Review of Economic Studies, 73, 459-485.
Poterba, J. M., (1984), "Tax Subsidies to Owner-Occupied Housing: An Asset-Market
Approach," The Quarterly Journal of Economics, 99,4,729-752.
Rosen, Ken, and L. Smith, (1983), The Price Adjustment process for Rental Housing and
the Natural Vacancy Rate, American Economic Review ,73, 1, 779-786.
Saiz, A., (2010). The Geographic Determinants of Housing Supply, Quarterly Journal
of Economics, 100, 3 500-530.
Stein, J. (1995), Prices and Trading Volume in the Housing Market: A Model with
Down payment Effects, Quarterly Journal of Economics 110, May, pp. 379-406.
Topel, R. and S. Rosen,(1988), "Housing Investment in the United States," Journal of
Political Economy, 96, 3, 718-740.
Voith, Richard and Theodore Crone,(1988), "National Vacancy Rates and the Persistence
of Shocks in U.S. Office Markets", AREUEA Journal,16, 4, 437-458.
Wheaton, W.,(1987), "The Cyclic Behavior of the National Office Market", Real Estate
Economics, 15:4 281-289.
Wheaton, W. (1999), "Real Estate Cycles: Some Fundamentals", Real Estate Economics,
27,2, 209-231.

21