June
2008
Factors Driving Charlotte’s Downtown Office Market
Geoffrey Zawtocki
Senior Associate
gzawtocki@warren-assoc.com
Summary
The record low vacancy rate in Charlotte’s downtown office
market led us to re-evaluate what drives net absorption.
Using statistical analysis, we looked at a number of local
factors and market indicators. We found that over 73% of
downtown office absorption can be attributed to changes in
office-using employment, inflation adjusted rent, and occupied
office space.
The impacts of these factors on absorption can be realized
within a range of three months to 1.5 years. In addition, each
factor affects absorption to a different magnitude. The average
impacts on absorption ranged from approximately 14,000 to 90,000
square feet quarterly.
Finally, we provide a forecast of office absorption
for the first half of 2008. We found that there is currently
significant downward pressure on occupancy, with the regression model
forecasting negative absorption of approximately 35,000 square feet.
However, adjusting this result with the announced move of the
Shaw Group and lease-up at EpiCentre, we forecast
absorption of approximately 117,000 square feet for the first
half of 2008.
Local Factors
and Market Indicators
Office absorption has traditionally been dependent on local
office using employment, office stock, and rental rates.[i]
&
[ii] As companies
grow or move, managers look for office space that fits their
needs. The price of leasing space also impacts managers’
decisions. Therefore, we tested downtown’s absorption for
statistical significance with the following office using
employment sectors:[iii]
• Finance
• Insurance
• Real Estate
• Information
• Professional and Technical Services
• Management of Companies
• Educational Services
• Administration
We combined Finance, Insurance, and Real Estate into one
category (FIRE), and the remaining sectors into a second
category (OFFICE). In addition, we tested downtown absorption
for statistical significance with a number of local market
indicators. In the end, four factors were found to drive
downtown Charlotte’s office absorption.
• Changes in FIRE employment
• Changes OFFICE employment
• CPI Adjusted rents[iv]
• Total occupied office space (OCCUP)
Regression
Model
We tested each local factor individually, and then combined
the factors into a single linear regression model[v],
with the generalized equation shown below:
Net Absorption = β0 + β1 (CPI Adjusted
Rent) + β2 (Change in FIRE) + β3 (OCCUP)
+ β4 (Change in OFFICE) + ε
The “β0”
term in the equation is a constant, and the
βn
terms are the factors’ coefficients, also constants. The “ε”
term represents the error term. As shown in Table 1,
the positive coefficients include FIRE and OFFICE. Increases
in these values affect absorption positively. The negative
terms include CPI Adjusted Rent and OCCUP. Increases in
these values affect absorption negatively.
|
Table
1: Regression Model Coefficients |
|
and
Delay in Effect |
|
|
|
Delay in |
|
|
|
Coefficient/ |
Effect |
|
|
Factor |
Constant |
(months) |
t-Stat |
|
Rent (β1) |
-37,113 |
12 |
2.56 |
|
FIRE (β2) |
55.34 |
3 |
4.71 |
|
OCCUP (β3) |
-0.25 |
3 |
4.91 |
|
OFFICE (β4) |
9.47 |
18 |
2.24 |
|
β0 |
4,172,593 |
N/A
|
4.37 |
|
Source: Warren & Associates |
|
|
|
Notes: Model has an R2 = 0.733, 28
degrees of freedom |
|
Notes:
and a
Durbin-Watson value of 2.01 |
|
We found these factors have their most significant effect on
net absorption in different future periods, making all four
factors leading indicators. This delayed or lag effect is
different for each factor. Changes in FIRE and OCCUP have
more immediate impact on downtown absorption in the
following quarter. However, changes in CPI adjusted rent
have a peak effect on absorption one year later, and changes
in OFFICE employment aren’t felt fully until18 months after
the change.
Regression
Model Accuracy
We tested the accuracy of the model against historical
absorption levels back to 1998[vi].
Graph 1 shows a scatter plot
of downtown absorption compared to our regression model’s
predicted net absorption. The horizontal position of each
point represents the measured downtown absorption, and the
vertical position measures the regression model’s predicted
absorption. A perfect regression model would have an
absolute correlation (r2 = 1). Our regression
model elicited a coefficient of determination, or goodness
of fit (R2) of 0.73. Therefore our model can
explain 73% of the observed downtown net absorption.
Our regression model appears to perform reasonably well,
both in periods of positive and negative absorption.
For example, downtown net absorption was measured at
910,327 square feet in 2002, and our regression model
predicted 752,964 square feet for that period. In
addition, it is able to reasonably predict periods of
negative absorption.
Sensitivity
Analysis
Changes in each factor also affect downtown absorption
by different magnitudes. Taking each factor’s average
change per quarter and multiplying it by the regression
coefficient results in the average effect on downtown’s
absorption. As shown in Table 2,
the average change in FIRE employment has the strongest
impact on absorption. An average change of 1,740
workers per quarter corresponds to approximately 96,300
square feet. In contrast, the average change in
inflation adjusted rent has the weakest effect. The
average quarterly change of $0.38 equates to
approximately 14,000 square feet.
|
Table
2: Sensitivity Analysis of Regression |
|
Factors, 2008 |
|
|
|
Average |
Effect on |
|
|
Coefficient/ |
Change per |
Absorption |
|
Factor |
Value (β) |
Quarter |
(Sq ft) |
|
Rent (β1) |
37,113 |
$0.38 |
14,200 |
|
FIRE (β2) |
55.34 |
1,740 |
96,296 |
|
OCCUP (β3) |
0.25 |
143,248 |
35,377 |
|
OFFICE (β4) |
9.47 |
8,177 |
77,412 |
|
Source: Warren & Associates |
|
|
Downtown
Absorption Forecast
Using local employment forecasts[vii],
we programmed the regression model to forecast downtown
absorption for the first half of 2008. As shown in
Graph 2, the regression
model forecasts negative absorption of almost 35,000
square feet in downtown. The model is reacting to
increasing rents and recent losses in office-using
employment. We agree that rising downtown
rents and losses in office-using employment will create
downward pressure on absorption.
In
forecasting absorption, we use this model and other
tools collaboratively to arrive at a forecast. We
incorporate factors
such as announced moves, pre-leased space,
shadow-space, and leakage, which the regression
model can't take into account. Taking these
factors into account, we adjusted the forecast which
results in a total of 117,142 square feet of absorption
in the first half of 2008. Much of the
absorption is due to the Shaw Group which announced
it is moving into 117,000 square feet. We also
included lease-up of EpiCentre which should have
office space completed shortly. The downtown submarket
is essentially at full occupancy with only 321,000
square feet available through the first quarter of
2008. Until new
space enters the market, this will severely
constrain absorption. In addition, the measurable
negative impacts of rent increases and current job losses are likely to
create greater downward pressure on absorption extending
into 2009.
[i]
Fuerst, Franz: “Predictable or Not? Forecasting Office
Markets with a Simultaneous Equation Approach”, European
Real Estate Society, 2006.
[ii]
Tse, Raymond & Webb, James, “Models of Office Market
Dynamics”, Urban Studies Vol. 40 No. 1, pp 71-89, 2003.
[iii]
North Carolina Employment Security Commission, Labor Market
Information, Employment and Wages by Industry, 1996-2007.
[iv]
Consumer Price Index (CPI)
[v]
Costello, Jim: “New Model For Office Demand”, Torto Wheaton
Research, 2002.
[vi]
Karnes Company, “The Karnes Report, Charlotte Office,
1998-2008.
[vii]
North Carolina Employment Security Commission Labor Forecast
|
