February, 1995, Revised July 1996.

THE CYCLIC BEHAVIOR OF THE GREATER LONDON OFFICE MARKET

​by

​William C. Wheaton
​Department of Economics and the
​Center for Real Estate
​M.I.T.

​Raymond G. Torto
​Department of Economics
​University of Massachusetts, Boston

​Peter Evans
​DTZ Debenham Thrope, London
 This research was made possible through the support of DTZ Debenham Thorpe, the MIT Center
for Real Estate, and the CB Commercial Real Estate Group, Inc. The authors are solely
responsible for the contents and conclusions of this report.

I. INTRODUCTION
In recent years, several studies have examined the cyclic movements of the U.S. office market, and the
extensive overbuilding of space that occurred in the 1980s [Wheaton and Torto (1987), Voith and Crone (1988),
King and McCue (1987), Rosen (1984)]. By contrast, published research about the office market in the United
Kingdom has been less prevalent and more descriptive [Barras (1994), Gardiner and Henneberry (1991)]. In this
study, we assemble time series data for the greater London office market. The data reveal two large swings in office
rental rates over the last 20 years, but only one significant building boom, occurring during the 1980s. To develop a
better understanding of the market's behavior, we estimate structural equations for office space demand, new supply
and rental movements. The estimated equations suggest that new office construction can be well explained with a
traditional investment model - but only when this previous rental boom is accounted for with a structural effect. On
the other hand, the movements in demand and rents over these two periods seem to be fully explained by the
creation of office jobs in the London economy. The amount of space demanded per worker displays a typical
demand schedule with respect to rental rates, while rents in turn react sharply to vacancy rates and leasing activity,
as would be suggested by modern search theory.

With the estimated equations, we can examine how the London office market might behave in the future -
given an assumed profile for the exogenous variables used in the equations. With a smooth profile of economic
growth, the model generates a dynamically stable office market, but one with only slight real growth in rental rates
from their current historic low values. A profile with a future positive or negative economic shock to the market,
sets off a market cycle, much like that of the past. Such a shock also leads to an "echo" several years later. A
positive shock inevitable generates a bust or market downturn, while a negative shock leads to small market boom.
This "echo" is to be expected given the "backward looking" dynamic structure of the model.

II. EMPLOYMENT AND OFFICE DEMAND

There is considerable evidence that the primary instrument driving office space demand is employment in
selected sectors of an economy [DiPasquale and Wheaton (1995)]. Studies in the U.S. suggest that more than 75%
of the space in larger office buildings is occupied by firms whose SIC is Finance, Insurance, Real Estate, Business or
Professional Services [Wheaton and Torto (1987)]. The remainder of office building space is occupied by the
ancillary activities of firms whose primary business is manufacturing, trade or utilities. It seems reasonable to
assume similar patterns characterize occupancy in the greater London area. Thus in Figure 1, we track the annual
percentage growth of employment in SIC category 8, which in England includes employment in finance, insurance,
business and professional services. The geographic coverage of the data is the South East region of England, which
encompasses a radius of roughly 40 miles around central London.
The actual demand for offices can be measured, ex post, as the amount of office space that is occupied, out
of that existing in the total stock. To distinguish between the two, we first create a series on the total stock of office
space and then combine it with a series on office vacancy. Unlike most U.S. cities, office space in London is widely
scattered throughout many smaller buildings. Keeping tract of the growth in these structures has proven to be a
difficult task for British Surveyors or Brokerage companies, particularly outside of Central London. The British
Government, however, does record the annual value of construction contracts for new office space within all of
Greater London, a geographic coverage which closely matches that of the employment data. It also publishes a
commercial building construction cost index. Deflating the value of contracts by this index yields an estimate of
orders for new construction square footage, which can then be lagged and summed into a stock series.
British surveyors do cooperatively monitor vacant and available office space within the major postal zones
of Central London, and regularly publish office vacancy rates. While this data does not cover the complete stock, it
is felt to be reasonably indicative of the Greater London market. Combining the series on the total stock, with that
on vacancy, produces a series for the occupied stock and its annual change - net absorption.
There are several sources of data series on office rents. British Surveyors closely monitor quoted lease
rates, for a sample of properties in central London. An alternative source of data, and one that better matches the
greater London area is an index of (constant pound) office rents that is produced by the IPD (Independent Property
Databank). Both of these series examine rents for a sample of roughly comparable or "constant quality" buildings.
In this respect, they are similar to recent attempts in the U.S. to statistically produce true rental indices [Wheaton
and Torto (1993)]. The IDP rental index in Figure 2 displays two very striking rental cycles over the last 21 years.
If every office worker in London used an identical and constant amount of office space, then office
employment and the occupied stock would move close to proportionately. Alternatively, the growth in office
employment would match the increase in the occupied stock, or net space absorption. To see if these relationships
hold, Figure 1 compares annual office employment growth in London with our estimated net absorption of office
space.
Figure 1 suggests that while there is a strong relationship between office employment growth and net space
absorption, the relationship is far from perfect. During most of the 1980s, office employment grew quite rapidly,
while net space absorption was unusually low. Conversely, there has been little employment growth over the last 4
years, but significant expansion in the amount of occupied space. There are several reasons why we should not
expect the two series to move in unison, and thus why space per office worker might change over time.

One explanation for why employment and occupied space might move differently is that space use per
worker varies across occupations, and London's occupational mix has been changing over time. If Clerks use less
space than managers, and if automation is eliminating the former, then the space per (remaining) office worker
might be expected to trend upward over time. To explain the more cyclic movements in space use per worker, we
need only recognize that office space is a factor of production, and as such its use should depend on rental rates. To
test if a demand elasticity exists, Figure 2 compares the London rental index with the ratio of occupied office space
to office workers. There appears to be a clear negative relationship. Both in the mid 1970s and late 1980s, when
rents were high, the amount of space per worker fell.
It has been argued that the use of longer term leases in the office market tends to bind tenants to their space
and thus hinders the adjustment of space consumption to changing demand conditions [DiPasquale and Wheaton
(1995)]. To incorporate such frictions, we model net space absorption with an occupied stock adjustment equation.
*
Let OS represent the amount of space that all firms in the market would in principle occupy if there were no leases,
moving or adjustment costs to obtaining such space. This represents potential office space demand, or demand ex
ante. From the discussion above, market demand should be the product of the number of office workers, Et,
multiplied by the office space per worker that firms would in the long run like to occupy. This desired amount of
space per worker should depend on the recent rent for space, Rt-1. Thus in equation (1), the term within brackets
represents the office space demanded per worker. The coefficient α1 is a base line square feet per worker, while α2
determines how much this space use varies with rent.

*
OS t = α0 + Et[α1 + α2Rt-1] (1)

*
The actual consumption of space, OSt, does not equal OS t, because firms cannot immediately adjust their
consumption in response to changes in demand (resulting either from employment growth or rent movements). If
some fraction of firms in each period, τ1, adjust their actual space consumption to long run demand, then the change
in the occupied stock (net absorption, ABt) will follow equation (2).
*
OSt-OSt-1 = ABt = τ1[OS t - OSt-1] (2)

Combining (1) with (2) yields a linear equation in which absorption and the occupied stock gradually adjust
to a "target" defined by office employment, and rents.
ABt = τ1[α0 + Et[α1 + α2Rt-1]] - τ1OSt-1 (3)

Equation (3) determines how absorption, and through absorption the occupied stock, adjust to reach a long
run desired level of occupancy, that in turn depends on office employment and rents. If employment and rents
*
remain fixed, the occupied space in the market should eventually settle at OS t. If we then take the amount of total
space in the market (St) as given, this occupied stock will yield a stable vacancy rate. Using the London data on
employment, the occupied stock and rents, our statistical estimate of equation (3) is:

ABt = 0.25 [7.3 + Et[294.8 - .92Rt-1]] - 0.25OSt-1

(0.3) (3.9) (-5.5) (-2.6)

2 2
R = .71, N = 21 (1974-1994), DW = 1.81, CHI = .46 (4)

In the equation above, the parameters tell a plausible and interesting story. The amount of office space
demanded per office worker (the term within the inner brackets] is 295 square feet minus 0.92 times rent. At the
current rent index value of 45, for example, the elasticity of space demand per worker is -.20. In reaction to a
demand shock, 25% of the market adjusts in one year, and within 3 years, the adjustment to the occupied stock is
57% completed. The adjustment model fits well, and displays virtually no autocorrelation.

III. VACANCY AND RENTAL MOVEMENTS

We have seen how office rents determine absorption and eventually a level of occupancy within the
existing stock. But what determines rents? With the use of long term leases, rents are likely to adjust slowly as
well. It has been argued that occupancy or vacancy rates play a central role in the adjustment of rents. In the U.S.,
for example, researchers have found strong statistical relationships between vacancy and the change in rents [Rosen
(1983), Heckman (1985), Shilling/Sirmans/Corgel (1987)]. Figure 3 takes the office rent index for London and
compares its annual change to the area's office vacancy rate series. Clearly, there appears to be a negative
relationship between the level of vacancy and rents: low (high) vacancy tends to generate higher (lower) rents.
Recent research argues that the relationship between vacancy and rent is more complicated [Arnott (1992)].
Vacancy is a spell through which parcels of space within buildings pass as they wait to be rented, either for the first
time or after a tenant moves. The ratio of the amount of vacant space to the total space needed by moving tenants
represents a measure of the expected time it will take a landlord to rent the space. The inverse of this ratio is the
probability that space is rented over an interval of time.

The search by tenants for space can be a time consuming and expensive process. Parcels of space come in
widely different sizes and configurations, especially within existing buildings. When appropriate space has been
found by a tenant, the stage is set for bargaining between tenant and landlord over rent or other lease terms.
Landlords establish a reservation or minimum rent that would make them indifferent between renting and keeping
space vacant. The longer the expected lease-up time for vacant space, the lower will be the landlord's reservation
rent. Searching tenants also have a reservation rent - the maximum they will offer as opposed to looking elsewhere.
When there are few tenants and many vacant parcels in the market, it becomes easier to find appropriate new space.
Thus when the expected lease-up time is high, the maximum rent a tenant is willing to offer also will be less. The
contract rent that emerges in bargaining must lie between the tenant's maximum offer and the landlord's minimum
reservation. Both of these move inversely with vacancy and positively with the number of tenants looking for space
in the market [Wheaton (1990)].
The modern theory of search and bargaining suggests that for a given level of vacancy and number of
searching tenants, there should emerge a stable level of market rent. Rents do not fall (or rise) continuously in
response to overly high (or low) vacancy [Wheaton and Torto (1994)]. The average number of tenants searching for
space should equal those that move - or gross space absorption. While measures of gross absorption do not exist,
the rate of net space absorption does reflect the outcome of tenant mobility. A rental adjustment model which
incorporates these variables is developed in equation (5).
 *
R = µ0 - µ1Vt-1 + µ2(ABt-1/OSt-1)

*
Rt-Rt-1 = µ3 [ R - Rt-1 ] , and hence,
Rt = µ3 [ µ0 - µ1Vt-1 + µ2(ABt-1/OSt-1) ] + (1-µ3)Rt-1 (5)

*
In (5) above, R represents the equilibrium rent which emerges in the market, determined as a linear
function of both absorption and vacancy rates. Market rents move towards this rent at a rate of µ3 per period. When
equation (5) is estimated with the London data, the results are:

Rt = 0.29 [62.4 - 462.2Vt-1 + 1417.3(ABt-1/OSt-1)] + 0.71Rt-1

(1.7) (-2.6) (3.2) (-9.2)

2 2
R = .89, N = 21 (1974-1994), DW = 1.38, CHI = 3.72 (6)

In (5) and (6), both vacancy and absorption are measured as a percent of the stock. Remember that the
*
term in brackets represents the equilibrium or target rent, R . Using historic average values for London (V=.07,
AB/OS=.02), the equation implies that the rental index will head towards 59 per square foot relative to the 1994
value of 44. With vacancy near 2%, and with 4% absorption (such as in 1981) the rental index would move towards
a level of 109. In each of these examples, the annual rate at which rents adjust towards these targets is 29%. The
equation fits extremely well, although the adjustment model still exhibits moderate autocorrelation.
The relationship above, in which vacancy determines rents, completes the demand side of the model. We
have already seen how rents affect property demand and therefore vacancy in the previous section. Here, vacancy
determines a stable level of rents. With rents determining vacancy, and vice-versa, there must exist a short run
market equilibrium which results from a given level of total space (St) and a level of office employment (Et). This
equilibrium has the following characteristics:
1). Given a level of office employment, and total stock of space, equation (3) eventually yields a stable
(zero absorption) occupied stock and hence vacancy rate. Equation (5) takes this vacancy rate and adjusts
rents until they are stable. When rents and vacancy are the same in equation (3) as they are in (5), then the
market is at equilibrium. In this equilibrium, rents lead tenants to demand an amount of space that yields a
vacancy rate, which in turn leads to the same stable value of rents.
2). If office employment increases, absorption turns positive, and with a given stock, vacancy falls. Lower
vacancy causes rents to increase, which in turn partially reduces absorption. Eventually, a new stable
equilibrium is reached with higher real rents, zero absorption, and a lower vacancy rate. The total occupied
space in the market will have increased, but space per worker will be less.

3). If the stock of space increases, vacancy rises, and this causes rents to fall. Lower rents generate positive
absorption, which then helps to bring down the vacancy rate. Eventually, at the new stable equilibrium,
real rents are lower, absorption is zero, and the vacancy rate is higher. Total occupied space has increased
(from the new supply) and lower rents have led to greater space per worker.
Thus, the system of equations (3,5) has all of the intuitive properties of a logically consistent market
clearing mechanism in which rents adjust to bring demand in balance with (fixed) supply. Its distinctive (and
complicating) feature involves the gradual response of office space demand to rents and employment, the continual
presence of vacant space, and the gradual adjustment of rents to changes in vacant space. This mirrors the behavior
of the market, behavior that comes from the market's use of longer term leases and the bargaining that occurs over
lease terms.

IV. CONSTRUCTION AND SUPPLY CYCLES

The construction of office space represents a form of investment in new capital. Following more general
investment theory [Abel and Blanchard (1986), Hayashi (1982)] it is reasonable to assume that the level of new
office construction would depend upon the asset price of office space relative to its replacement cost. The asset
price of office space in turn should be based upon current effective net rental income (considering vacancy), and a
capitalization rate. Thus our equation for new investment or construction orders (Ct) should contain the series on
office rents and vacancy (Rt, Vt), some interest rate or more complicated capitalization rate series (It), and some
measure of replacement costs (RCt).

Ct = β0 + β1Rt + β2Vt + β3It + β4RCt (7)

Figure 4 tracks the estimated construction of new space in greater London market, in comparison to the
office rental index. The building boom of the 1980s appears clearly to be a reaction to strong demand and the
increasing real rental rates during that period. What is puzzling is why the extremely high rental rates of the early
1970s did not result in greater construction levels then. The common explanation given in England is that the Labor
party governments in power during the 1970s enacted a number of restrictions on building activity which effectively
precluded new development. By the late 1970s most of these restrictions were removed. At that time, however, the
economy had not grown for several years and lower rents were sufficient to discourage new development.
​Measures of replacement cost in the real estate industry are particularly complicated. Land is estimated to
account for almost two thirds of the asset value of commercial real estate in London, but there is little reliable data
on land prices. As discussed earlier, however, there is a good series on commercial space construction costs. With
real estate investment, when asset prices are high relative to construction costs, land residual values (shadow prices)
rise. It is these residual values which should enable developers to engage in profitable projects. Thus it is
sometimes argued that in a real estate investment equation, the most appropriate measure of replacement costs is that
for capital alone.
Rental income is converted to an asset price through a capitalization rate, that in principle is based on many
considerations: economy wide interest rates, depreciation or other preferential tax treatments, and expectations of
future rental income. There is good data on the "yields" of publically traded real estate companies, and also yield
data based on the appraised value of privately held real estate. In our model, such series would surely be
endogenous, since property yields are themselves based not only on broader interest rates, but expectations about
market conditions and income growth in the commercial real estate sector. We take the more simple approach of
directly including current 10 year treasury rates in the construction equation.
If we use a shorter 19 year period (as opposed to the 21 year sample in the rent and absorption equations)
we bypass the problem of explaining why extremely high rents in 1973-75 did not generate new construction. With
the 1976-94 sample, our construction equation is estimated as follows.
Ct = 26.3 + .1902Rt - 52.Vt - .2204RCt - 0.421It
(5.0) (6.9) (-3.8) (-5.0) (-1.4)

2
R = .88, N = 19 (1976-1994) , DW = 2.46 (8)

The implications of this model are reasonable and also easy to interpret. At current (1994) construction
costs (L87), vacancy (.10), and interest rates (10%), the greater London rent index would have to be 13 in order to
get any construction started. At the current (1994) rent index value of 45, construction should average about 6
million square feet. The historic high rent index levels that prevailed in the late 1980s (120) would eventually
generate yearly construction of 20 million square feet - assuming that construction costs and vacancy remained at
current levels. Several measures of both real and nominal interest rates have the appropriately (negative) signed
coefficient in the equation, but none proved to be very important statistically.

V. ECONOMIC OUTLOOKS AND MARKET BEHAVIOR

Equations (4), (6) and (8) give three behavioral relationships among the 6 endogenous variables:
Absorption (ABt), rent (Rt), new construction orders (Ct), vacancy (Vt), the total and occupied stocks (St, OSt). To
complete the model, we need only recall the three identities in (9) below. The exogenous instruments for the model
are interest rates (It), office employment (Et) and real construction costs (RCt).

​S t
= St-1 + Ct(1-δ) ​
​V t
= (St - OSt)/St ​
OSt = OSt-1 + ABt (9)
Using this system of equation, the future behavior of the London market can be studied - based on an
assumed macroeconomic outlook that provides estimates of the exogenous variables: employment, interest rates, and
inflation or construction costs. It is useful to consider several alternative outlooks, in order to examine the dynamic
properties of the model.
Our first outlook is a "base forecast" using smooth and somewhat slow growth in office employment over
the next decade, about 1.5% annually. Long term interest rates are assumed to hold at current levels (10%) and there
is only a
slight (1%) yearly increase in real construction costs - consistent with its long term trend. Figure 5 gives the outlook
for office space absorption and construction, while Figure 6 gives the resulting movements in vacancy and rent.
The "base forecast" illustrates the market "echo" from past economic shocks. The lower rent levels in
recent years, resulting from the overbuilding and cooling economy of the early 1990s has now generated new
absorption and falling vacancy. These lead rents to recover somewhat, from 45 in 1994 to 62 by 1997, and this sets
off a slight increase in construction. While the model is dynamically stable, shocks do have later repercussions. By
the year 2000, the market settles down - assuming smooth economic growth - along a steady state growth path.
With a rent index in the low 50 range, and vacancy around 8%, construction and absorption are closely matched at
around 6 million square feet annually - maintaining these market values. While a rent index of 50 is low
historically, this corresponds to central London rents of around L30 (net). Such rents could easily generate new
construction, given that replacement (capital) costs are only L88. These figures generate significant land residuals.

The second "forecast" [Figures 7,8] examines the impact of a negative economic shock to the London
market, such as occurred with the recessions of 1975, 1981 or 1991. Office employment is assumed to contract by
7% over 2 years (1996-1997) and then resume the slow growth rate of the "base forecast". Interest rates and
construction costs are assumed stable as in the "base forecast". The model reacts with a quick and sharp drop in
absorption, but then recovers strongly around the year 2000, largely induced from falling rents. Rents fall because it
takes construction at least 2 years to even begin reacting to these changes. Thus vacancy rises from 7% to 13% and
this together with negative absorption drives rents from 45 to 30. The rise in vacancy is less than the employment
contraction because of the rental elasticity of space demand. At the end of the forecast period, vacancy falls sharply
since modest employment growth has resumed with no new space entering the market. This allows rents to start
recovering.
The final forecast [Figures 9,10] examines the impact of a positive employment shock to the London
economy, such as occurred between 1984 and 1988. Office employment is assumed to grow by 5% annually for 5
years (1996-1999) and then resume the slower growth rate. Again, interest rates and construction costs are assumed
stable at current (real) levels. The model's response to this shock is much like that which occurred during the mid
1980s. Strong net absorption, in the 11-12 million square feet range drives down vacancy to 4%, because again it
takes construction at least 2 years to begin reacting to the shock. Rents rise sharply, from 45 to 83, and these in turn
then cause absorption to contract. By the end of the forecast period, the construction boom arrives, employment and
absorption are much lower, so vacancy rises and rents start falling.

These exercises suggest that much of the volatility in the London market can be explained by employment
movements in the city's office sectors. In each experiment, the demand shocks closely replicate historic market
changes. The market also displays considerable volatility. Shocks of historic magnitude cause vacancy to vary
between only 4% and 13% - somewhat small by the fluctuations typical of U.S. markets. These movements in
vacancy, however, generate major swings in rents and construction. In each case, the shock also generates an
"echo". This pattern of behavior comes from the lags in the market. Construction may respond quickly to rents, but
rents react more gradually to vacancy and absorption. With longer term leases, it also takes absorption and vacancy
some time to fully respond to economic change. Thus a negative shock leads to a market rebound, and a positive
shock eventually generates a market downturn.
Our London model demonstrates that commercial property in European cities is forcastable - to the extent
that any economic variable is - dependent upon a pattern of economic growth. The dynamic structure of property
markets, however, means that the response of these markets to economic change will be quite different than that
which occurs in the markets for other goods and services.

​REFERENCES

Abel, Andrew and Olivier Blanchard, "The present value of profits and cyclic movements in
investment", Econometrica, 54,1986,249-73.

Arnott, Richard, "Housing Vacancies, Thin Markets and Idiosyncratic Tastes", Journal of Real
Estate Finance and Economics, 2,1989, 5-30.
Barras, Richard, "Property and the economic cycle:building cycles revisited", Journal of Property
Research, 11, 1994, 183-197.

DiPasquale, Denise and William Wheaton, The Economics of Real Estate Markets, Prentice Hall,
New Jersey (1995).

Gardiner, C. and Henneberry, J. "The Development of a Simple Regional Model of Office Rent
Prediction", Journal of Property Valuation, 7, 1988, 36-52.

Hayashi, Fumio, "Tobin's marginal and average q: a neoclassical interpretation", Econometrica,

50, 1982, 213-24.

Hamilton, James D., Time Series Analysis, Princeton University Press, Princeton, N.J. (1994).

​Heckman, J.,"Rental Price Adjustment and Investment in the Office Market", AREUEA Journal,
13:1, 1985.

King, John L., and Thomas E. McCue. "Office Building Investment and the Macroeconomy:
Empirical Evidence 1973-85." AREUEA Journal 15, 3 (1987): 234-255.

​Rosen,K. and Smith,L.,"The Price Adjustment Process for Rental Housing, and the Natural Vacancy
Rate", American Economic Review, 73:779-86, 1983.

Rosen, Kenneth T. "Toward a Model of the Office Building Sector." AREUEA Journal 12, 3
(1984): 261-269.

​Shilling, J.D., C.F. Sirmans, J.B. Corgel, "Price Adjustment Process for Rental Office Space",
Journal of Urban Economics, 22:90-100, 1987.
Voith, Richard and Theodore Crone. "National Vacancy Rates and the Persistence of Shocks in
U.S. Office Markets." AREUEA Journal 16, 4 (1988): 437-458.

​Wheaton, William,"Vacancy, Search and Prices in a Housing Market Matching Model", Journal of
Political Economy, 98: 1270-1293, 1990.

Wheaton, William C. "The Cyclic Behavior of the National Office Market." AREUEA Journal
15, 4 (1987): 281-299.

Wheaton, William C. and Raymond G. Torto. "Office Rent Indices and their Behavior Over
Time." Journal of Urban Economics 35 (1984): 112-139.