THE ECONOMIC IMPACT OF SPORTS
STADIUM CONSTRUCTION:
The Case of the Construction
Industry in St. Louis, MO
PHILLIP A. MILLER*
University of Missouri
ABSTRACT: This article examines the St. Louis construction industry's employment with
special attention given to the periods during which the Kiel Center and the Trans World
Dome were being built. We analyze whether the construction of a major sports stadium
increases construction industry employment. An econometric model is developed to explain
the times series trend of construction industry employment in the St. Louis SMSA. The statistical
evidence suggests that the levels of employment in the construction industry were
neither higher nor lower during the construction of these stadia. It is argued that construction
on these projects merely substituted for other construction projects in this SMSA.
George Bernard Shaw once said "if you laid all economists end-to-end they would never
reach a conclusion." There is much truth to this quip, but there are some issues on which most
economists agree. One of these issues is the economic impact of professional sports teams
and sports facilities. Despite the claims of economic impact statements commissioned and funded
by professional sports teams and other stadium proponents, the independent academic research
on this subject has yielded a consistent conclusion: the existence of a sports franchise in an
SMSA does not generate positive net benefits for the SMSA and could actually generate negative
net benefits.
Baade and Dye (1990), analyzing data for cities that had baseball stadiums renovated or
built between 1965 and 1983, found an overall insignificant effect on the level of aggregate
personal income in the cities and a significant negative effect on the cities' regional share of
income. They also found an insignificant effect on retail sales and a significant negative impact
on the cities' regional share of retail sales. Similarly, Rosentraub, Swindell, Przybylski, and
Mullins (1994) found little if any appreciable difference in employment and payroll growth
between cities when they compared Indianapolis to similar cities to examine Indianapolis' strat*Direct
correspondence to: Phillip A. Miller, Economic and Policy Analysis Research Center, University of Missouri,
Columbia, MO 65211. E-mail: millerpa@missouri.edu
JOURNAL OF URBAN AFFAIRS, Volume 24, Number 2, pages 159­173.
Copyright  2002 Urban Affairs Association
All rights of reproduction in any form reserved.
ISSN: 0735-2166.
egy to invest in sports related downtown redevelopment. Baade (1996) examined whether sports
teams or stadiums have a significant economic impact on the cities in which they reside. Using
data from 48 cities from 1958 to 1987, he found no significant impact on the change in municipality
per-capita income attributable to the existence of a team or a stadium. He also found
that there was no significant impact on municipality share of state employment in the amusement
and recreation industry (SIC 79) in the pooled data but he did find some small significant
impacts in select cities.
Other researchers have found significant negative impacts associated with the existence of
sports teams. Coates and Humphreys (1999) examined the 37 cities in the United States with
NBA, NFL, or MLB teams and found that the existence of these teams decreases the level of
real per capita income in a city. One of the reasons cited for the lack of any measurable economic
impact is that the jobs produced by a sports team appear to be a direct result of purchases
not made in other sectors of the economy. Focusing on the same 37 cities, Coates and
Humphreys (2000) examined the effect of sports teams on specific sectors of the SMSAs' economies.
They found that the presence of a sports team increases employment and earnings in
the amusement and recreation sector but decreases it in all other sectors by an amount that
offsets the increase in the amusement and recreation sector. This provides evidence that consumers
reallocate spending from other sectors of a regional economy to sporting events. This
explains the overall findings in their 1999 article.
These results are not surprising when one considers the resources that produce sporting events.
By in large, most of the spending that goes towards these resources goes to the players and to
the owners of the teams. For example, according to 1996 expenditure data obtained from Financial
World (Badenhausen, Nikolov, Alkin, & Ozanian, 1997), 59% of baseball team operating
expenditures went towards player salaries. Most players do not live in the cities that they represent.
Consequently, it is plausible that a good portion of their income leaves the cities that
they represent. It is also plausible that a large share of owners' incomes derived from the operation
of a sports team will also leave their host cities.
However, the permanent jobs directly created by a sports team within its host city tend to
be seasonal and low-wage jobs (ushers, concessionaires, etc.). Because consumer expenditures
on sporting events displaces spending from other sectors of a regions' economy (e.g.,
restaurants, bars, retail stores, and movie houses), this can lead to lower employment and lower
wages in those sectors. Many of the jobs in these other sectors would pay comparable wages
but would be year-round. It is also plausible that a greater proportion of spending in these
industries will circulate throughout an economy. Hence, the results presented by Coates and
Humphreys are not surprising.
Therefore, the evidence suggests that, at best, the existence of sports teams and sports stadiums
in an SMSA causes consumers to redistribute their purchases between alternative entertainment
purposes or between different geographic areas within the SMSA. At worst, they can
actually decrease earnings and employment in their SMSAs. Hence, the existence of major
league teams and the construction of sports stadiums do not provide a catalyst for economic
development.
However, this body of literature does not claim that there are no benefits to an SMSA having
a sports team that would justify a government subsidy to construct a sporting venue. What
they do question, however, are the claims of the economic impact statements that attempt to
justify public subsidies on the grounds of substantial economic net benefits.
Noll and Zimbalist (1997) argue that sports teams generate external benefits to their host
SMSAs and, as a consequence, subsidies may be justified to ensure the socially optimal quantity
of sports competition in an MSA. However, the economic impact statements ignore such
externalities. Johnson, Groothuis, and Whitehead (2001), analyzing the case of the Pittsburgh
160 6 JOURNAL OF URBAN AFFAIRS 6 Vol. 24/No. 2/2002
Penguins of the NHL, argue that the value of public goods generated by the Penguins is less
than the cost of a new arena. Consequently, while construction of a new stadium or renovation
of an old stadium should not be fully funded by the public, some public funds are warranted.
One issue that has not received much attention in the independent literature is the effect of
building a sports facility on construction employment in an SMSA. Because the construction
activity occurs for a relatively short period of time, the effect on construction employment
and all its derivatives is temporary. However, this effect is still an economic benefit.
The construction generated by the building of any facility will constitute an economic net
benefit on employment levels only if those workers employed on the stadium project would
have been otherwise unemployed. Of course, determining if a person would be unemployed if
they were not working at a particular position or doing a similar job in another location would
be a difficult task. However, one can examine if overall construction employment is significantly
higher in an SMSA during periods when sports facilities are being constructed.
During the 1990s, two major league sports facilities were constructed in St. Louis, Missouri:
the Kiel Center and the Trans World Dome (currently the Edward Jones Dome). The
Kiel Center houses the NHĽs St. Louis Blues, athletic teams from Saint Louis University,
and other athletic and non-athletic events. It was financed with a mixture of public and private
financing: The city of St. Louis provided $34.5 million for the project while 20 corporations
provided $30 million in cash and guaranteed $98 million in construction loans. It was
constructed between March 1992 and October 1994 at a cost of $170 million (Munsey & Suppes,
2001).
The Trans World Dome was constructed from May 1993 to October 1995 at a cost of $280
million. It was constructed primarily to bring an NFL franchise back to St. Louis to replace
the St. Louis Cardinals (a franchise that departed for Phoenix, Arizona in 1988). It was successful
in this endeavor and now houses the St. Louis Rams. The Dome has also hosted various
other sporting events as well as many non-athletic events. It is owned by the city of St.
Louis and was funded with 100% public funds from various state and local government levels
(Munsey & Suppes, 2001).
In comparison, consider road and bridge projects undertaken in the St. Louis metropolitan
area. According to figures released by the Missouri Department of Transportation (personal
communication, Linda Wilson, December 21, 2000), total road construction spending in 1994
on the Missouri side of the St. Louis metropolitan area amounted to $125 million. Considering
construction on the arenas were spread over multiple periods, the projects, especially the
Trans World Dome, compare favorably to the road projects in terms of the dollar size of the
projects.
In this article, the level of construction employment will be theoretically and econometrically
modeled with particular attention given to the St. Louis SMSA and the time periods during
which the Kiel Center and the Trans World Dome were being constructed. We will examine
if overall employment in the construction sector of the St. Louis SMSA was higher in those
periods during which the stadia were under construction. We find no evidence that construction
industry employment in the St. Louis SMSA was higher in the periods during which the
Kiel Center and the Trans World domes were being constructed.
THE THEORETICAL ANALYSIS
The following theoretical model outlines the framework used in the development of the econometric
model of this article. Consider a city defined as the SMSA with an available supply of
construction labor Lc. The construction workers need not live in the SMSA of the city--they
just need to be able to be employed there. These are workers who would be employed in SIC
6 Sports Stadium Construction 6 161
15, 16, and 17 in the United States. Of course, there are other industries that are impacted
directly and indirectly by construction projects. However, according to the Regional InputOutput
Modeling System (RIMSII) multipliers provided by the US Bureau of Economic Analysis
the construction industry receives the lion's share of the employment impacts from a given
expenditure made in the construction industry. For example, according to 1997 RIMSII multipliers
for the state of Missouri, $1 million spent on a construction project will generate roughly
27 jobs, 12 (44.4%) of which are construction jobs (see Table 1). Unfortunately, St. Louis multipliers
were not available for this research project. However, in comparison, if $1 million were
spent on a construction project in Columbia, a city smaller than many St. Louis suburbs, this
spending would generate 11.5 construction jobs in the Columbia metropolitan area (according
to 1997 RIMSII multipliers for the Columbia metropolitan area). This suggests: 1) the effects
of local construction expenditures will be mostly felt locally with only peripheral impacts on
the rest of the state, and 2) because the St. Louis SMSA is vastly larger than the Columbia
metropolitan area, it is likely that the St. Louis RIMSII construction multipliers are the larger
of the two and thus closer to the Missouri multipliers.
Suppose that there are currently n construction projects that utilize the construction workers
with the number of workers employed at building site i denoted as Li where (i 1
n
Li
uc Lc. uc is the number of construction workers out of Lc who are currently unemployed.
Hence, (i 1
n
Li represents the level of employment in the industry.
For the purposes of this study, an unemployed worker is one not currently working in the
construction field in the city. This person is not necessarily unemployed or actively seeking
work. Instead the worker must not be working in the construction industry in the city. This
person could come from another industry (i.e., the retail industry) or he could come from the
construction industry in another city. Recall that we are ultimately interested in finding whether
an SMSA realizes an increase in the number of workers employed in the construction industry
when it has a large project under construction. Therefore, whether these workers would have
been employed elsewhere, while an important issue in the overall scheme, is immaterial to
this particular study.
The workers have physical and human capital abilities that are useful in the construction of
a building, and these abilities are assumed to be exogenously determined. Because this is the
case, Lc is also assumed to be exogenously determined.
TABLE 1
Jobs per $1,000,000 Spent in the Construction Industry
in Missouri
Industry Aggregation RIMSII Multiplier
Farm and agricultural services, forestry 0.455
Mining 0.0762
Construction 11.9885
Durable goods 1.6176
Non-durable goods 0.7354
Transportation and public utilities 0.916
Wholesale trade 0.8486
Retail trade 2.6935
FIRE 1.0759
Services 6.2669
Households 0.1573
Total 26.8309
162 6 JOURNAL OF URBAN AFFAIRS 6 Vol. 24/No. 2/2002
Suppose the construction project j will be completed using the production function Qj
f ~Lj , Loj , Kj !. Qj can be thought of as the size of the project. Loj is the number of other labor
inputs (such as plumbers and electricians) and Kj is the number of units of capital used at site
j. Let this production function be twice continuously differentiable in Lj with @]Q~.!#/]Lj 0
and @]2
Q~.!#/]Lj
2
0. Because we are examining construction employment, for simplicity we
assume Loj and Kj are exogenously determined.
When construction begins on a new project in this city, there are two areas from which construction
workers can be pulled: from another job site or from the ranks of the unemployed as
defined above ~uc !. If uc 0 then it must be the case that the number of workers employed at
site j must be drawn from the other n 1 projects. Hence, (k j DLk Lj where DLk represents
the loss of construction workers from site k j.
However, if uc 0, then the new project can pull some workers from the unemployed labor
pool. If this is the case, then the construction industry will realize an increase in the level of
employment. According to the RIMSII multipliers for the state of Missouri given in Table 1,
the Kiel Center's $170 million construction cost should have translated into 2,038 construction
jobs and the Trans World dome's $280 million price tag should have generated 3,357 construction
jobs. As noted above, it is likely that the greatest effects would be in the St. Louis
SMSA and that they occurred in the periods during the periods when the stadia were being
constructed. We shall see if these results pan out in the following section.
THE EMPIRICAL ANALYSIS
The Data and the Models
All data used in the regression analyses presented in this article were obtained from the
databases of the Economic and Policy Analysis Research Center in the Department of Economics
at the University of Missouri. The wage and employment data for the construction industry
used in the analysis are for the St. Louis SMSA. The data consist of 112 quarterly observations
spanning the period from the first quarter of 1971 to the fourth quarter of 1998. Table 2 presents
the list of, and the summary statistics for, the continuous variables used in the following
regressions.
For any observation at quarter q, the basic model analyzed is
SAECCq aXq b1 KIEL b2 TW gTq yq. (1)
TABLE 2
Listing and Means of Continuous Variables
Variable Abbreviation Mean Std Dev Minimum Maximum
St. Louis Employment, Constructiona
SAECC 49553.96 9560.81 26251 68315
St. Louis Quarterly Wages, Constructionb
SAWCC 26126.39 8539.72 12007 42509
US--Invest, Total Nonresidentialc
GINQ 561.24 210.94 277.90 1182.30
St. Louis Building Permits, Totalb
SABP 4059.21 1289.16 1524 7820
Bond Yield, AAA Corporate FYAAAC 9.21 2.08 6.33 15.01
US­CPI, All Itemsd
PZUNEW 102.29 39.57 39.90 164.00
N = 112
Note: Dates 1971Q1 to 1998Q4.
a
Seasonally Adjusted.
b
Seasonally Adjusted at an Annual Rate.
c
Real 1987 Dollars.
d
Not seasonally adjusted.
6 Sports Stadium Construction 6 163
a is a (1 n) matrix of parameters on the continuous variables contained in the (n 1)
matrix Xq. Kiel is a dummy variable equal to one for the quarters during which the Kiel Center
was under construction. TW is a dummy equal to one for the quarters during the Trans
World dome was being constructed. b1 and b2 are the parameters on the Kiel and Trans World
dummy variables, respectively. g is a (1 m) matrix of parameters on the various dummies
contained in the (m 1) matrix Tq which contains dummies that control for time-specific effects
on St. Louis construction industry employment. For example, the variable D74Q2 1 for the
second quarter of 1974 and 0 otherwise. D8889 is equal to one for the years 1988 and 1989
and 0 otherwise. nq is a stochastic disturbance term, which follows particular processes described
below.
ARCH Models and Regression Results
We analyze several different models because we feel that similar results will strengthen the
conclusions. The first four models were analyzed using various ARCH models using maximumlikelihood
estimation. Time series variables may appear statistically correlated merely because
they both move together over time--not because of any causation between the variables. ARCH
models control for these autoregressive processes that frequently occur in time series data. ARCH
models also control for differences in variances that can occur in some time series variables.
This is an important control because construction employment increases over time and, as a
result, its variance will also increase.
In each case, a backstep selection process was performed to determine the appropriate lag
to place on the disturbance terms. In this approach, the software package used (SAS) deleted
insignificant estimated lagged disturbances from the model, leaving only significant estimated
lagged disturbances. This method was employed because it allows the data to determine
what the appropriate lag is rather than having the researcher use trial-and-error to discover
it. In each case, the disturbance was found to follow an AR(1) process. In addition, Lagrangian
Multiplier tests and Q tests were performed on the disturbances to test for potential ARCH
processes. The tests suggested that the disturbances follow an ARCH(1) process in each case.
Hence, the disturbance terms are assumed to be generated by the following AR(1)-ARCH(1)
process:
yq q wyq 1
q !hqmq
hq  dq 1
2
mq ; iid~0,1!.
Model 1 is given by the following equation:
SAECCq a0 a1 GINQq b1 KIEL b2 TW gTq yq
GINQq is the level of national real non-residential fixed investment and is used to control for
the amount of construction spending in St. Louis. We use the national level of investment because
we did not have state or local investment expenditures available. However, this is a satisfactory
control because non-residential investment spending in St. Louis and non-residential investment
spending in the nation should be positively correlated. The regression results are given
164 6 JOURNAL OF URBAN AFFAIRS 6 Vol. 24/No. 2/2002
in the Table 3. First note that the model fit the data very well (R2
0.9841) and most parameters
are significant at least at the 10% level of significance. A normality test on the estimate
of the error term q /!hq suggests that they are normally distributed (the p-value on the null
hypothesis that they are not normally distributed is 0.1561). The residuals were uncorrelated
over time and were homescedastic.
The parameter estimate for non-residential investment is significant and positive. This suggests
that a $1 billion expenditure in national non-residential investment increases construction
employment in St. Louis by over 35 workers. However, the parameter estimates for both
the Kiel and Trans World dummies are positive but highly insignificant. Thus, this model suggests
that after controlling for investment expenditures, the level of construction employment
in St. Louis was not larger on average than in other quarters covered in the study.
Model 2 is the same as model 1 except it includes building permits for the St. Louis SMSA
lagged one quarter. We include this variable as an additional control on the amount of St. Louisspecific
non-residential investment spending. While using national non-residential investment
helps control for construction expenditures, it does not fully control for construction expenditures
in St. Louis. Adding lagged building permits to the model provides an additional control
for this. We use the lagged variable because building permits are issued before construction
begins on a project. Model 2 is given by the following equation:
SAECCq a0 a1 GINQq a2 SABPq 1 b1 KIEL b2 TW gTq yq.
The regression results for this model are also given in Table 3. Once again, the model fit
the data very well (R2
0.9823) and most variables are significant at least at the 10% level
of significance. A normality test on the estimated errors found them to be normally distributed
(p-value on the null is 0.6405). The residuals were uncorrelated and homoscedastic.
The parameter estimate on non-residential investment is significant and positive suggesting
that a $1 billion expenditure on non-residential investment in the US gives work to approximately
36 construction workers in St. Louis. The parameter estimate on lagged building permits
is also positive and significant, but it is very small. It suggests that if the number of building
permits issued in any particular quarter rises by 2500, only one additional construction worker
will be employed in the current quarter, all else equal. However, as in model 1, the parameter
estimates on the Kiel and Trans World dummies are both positive but highly insignificant suggesting
that, on average, after controlling for non-residential investment expenditures and the
number of lagged building permits, there is no evidence of higher or lower levels of construction
employment during the periods when the two major sports facilities were being built.
Models 3 and 4 are the same as model 2 with the addition of an average interest rate measure.
While businesses will consider their expectations of interest rates when deciding whether
to invest in a project, the amount of expenditures may not appropriately control for changes
in the expectations. Consequently, we added average interest rates to the model. Model 3 contains
the interest rate measure FYAAACqrq 3, the average AAA corporate bond rate during
the past four quarters including the current quarter. Model 3 is thus given by
SAECCq a0 a1 GINQq a2 SABPq 1 a3 FYAAACqrq 3
b1 KIEL b2 TW gTq yq.
The regression results from model 3 are given in Table 3. As in models 1 and 2, the model
fits the data very well ~R2
0.9844) and most parameters are highly significant. The residu6
Sports Stadium Construction 6 165
TABLE 3
ARCH Model Regression Results
Variable
Parameter
Estimate
Standard
Error Variable
Parameter
Estimate
Standard
Error
Model 1: Model 3:
Intercept** 29369 5877.5 Intercept** 22955 5078
US--Invest, Total Nonresidential** 35.693784 6.9491 US--Invest, Total Nonresidential** 37.135506 5.8532
KIEL 464.045247 1352.3 SL--Building Permits, Total: Lag 1** 0.615098 0.2485
TW 669.407129 7965.5 Bond Yield, AAA Corporate: Curr to Lag 3 339.007383 267.9
D72Q3** -1870.31584 933.8 KIEL 293.048079 1359.9
D74Q2** -13480 1078.7 TW 706.871 5221.9
D80Q2** -4024.501309 837.5 D74Q2** -13848 1305.9
D81Q2* -1129.948883 607.1 D79Q2* 1116.614146 640.5
D88Q1* 881.418519 494.7 D80Q2** -4144.594403 1174.4
D91Q1* -2928.234377 1629.1 D81Q2* -1037.668709 588.8
AR(1)** -0.956396 0.0328 D87Q2* -1822.689824 812.2
ARCH0** 1664195 41.3859 D91Q1 -2649.985721 2246.9
ARCH1 1.08E-19 4.22E-11 AR(1)** -0.950729 0.0358
ARCH0** 1708046 11.0812
R-Sq 0.9841 ARCH1 8.67E-19 3.42E-11
R-Sq 0.9844
Model 2: Model 4:
Intercept** 27172 5327.8 Intercept** 25750 5940.5
US--Invest, Total Nonresidential** 36.396445 6.9857 US--Invest, Total Nonresidential** 36.651932 6.5606
SL--Building Permits, Total: Lag 1* 0.407488 0.2108 SL--Building Permits, Total: Lag 1** 0.519568 0.236
KIEL 205.76392 1334.4 Bond Yield, AAA Corporate: Lag1 to Lag 4 93.824587 274.7
TW 773.59061 7806.4 KIEL 154.993845 1162.1
D74Q2** -13733 1303.6 TW 826.830093 6778.7
D79Q2 957.707762 706.8 D74Q2** -13847 1187.6
D80Q2** -3992.895504 1257.5 D79Q2* 1048.255951 635.3
D88Q1 598.159731 572.9 D80Q2** -3936.562106 1196.3
AR(1)** -0.947544 0.0361 D87Q1** 2555.622824 1135
ARCH0** 1831391 32.1129 D91Q1 -2689.020407 2065.7
ARCH1 2.12E-22 5.38E-11 AR(1)** -0.953003 0.0372
ARCH0** 1664566 23.6761
R-Sq 0.9823 ARCH1 1.08E-19 3.84E-11
R-Sq 0.9847
*Significant at 5% to 10%
**Significant at less than 5%
1666JOURNALOFURBANAFFAIRS6Vol.24/No.2/2002
als were uncorrelated and homoscedastic and normality tests on the estimated errors suggest
they are normally distributed (p-value 0.3635).
The parameter estimate on non-residential investment expenditures is once again positive
and significant. According to this estimate, a $1 billion increase in this expenditure will cause
St. Louis construction industry employment to increase by 37 workers. The parameter estimate
on lagged building permits is also positive and significant, suggesting that a 3000 unit
increase in building permits in one period causes just less than two construction workers to be
employed on average during the following period in St. Louis. The parameter estimate on the
average AAA bond rate is insignificant. This suggests that non-residential investment expenditures
adequately capture interest rate expectations. In addition, there is no evidence that the
Kiel and Trans World dummies are significant.
Model 4 uses the average AAA corporate bond rate during the previous four quarters,
FYAAACq 1rq 4. The following model was thus estimated:
SAECCq a0 a1 GINQq a2 SABPq 1 a3 FYAAACq 1rq 4
b1 KIEL b2 TW gTq yq.
The results from the regression run on model 4 are also given in Table 3. The model fits the
data well (R2
0.9847) and most parameters are significant at least at the 10% level of significance.
A normality test on the estimated errors suggests they are normally distributed
(p-value 0.6544). The residuals were uncorrelated and homoscedastic.
As in the previous models, the parameter estimates on non-residential investment and on
lagged building permits are both positive and highly significant. They suggest that a $1 billion
increase in national non-residential investment expenditures and a 2500 unit increase in
the previous quarter will increase construction employment in St. Louis by 36.7 and 1 worker,
respectively. As in model 3, the parameter estimate on the average of the four previous quarters'AAA
corporate bond rate is insignificant suggesting, once again, that non-residential investment
expenditures adequately controls for interest rate expectations. As in all other models,
there is no evidence that the Kiel and Trans World dummies are significant. Hence, regardless
of how Equation 1 was modeled, there is no evidence that the level of employment in the construction
industry was any higher or lower than in other quarters, all else equal.
Two-Stage Regression Model and Regression Results
One of the potential shortcomings of models 1 through 4 is that they each include nonresidential
investment expenditures as a regressor. It is conceivable that any employment effects
caused by construction on the Kiel Center and the Trans World Dome are captured by this
regressor, which causes the parameter estimates for their respective time dummies to be insignificant.
Hence, it is valuable to analyze an alternative model that does not contain nonresidential
investment as an explanatory variable.
Consequently, a regression was run on the following model
SAECCq a0 a1 SAWCCq a2 SABPq 1 b1 KIEL b2 TW gTq yq. (2)
To control for possible correlation between SAWCCq, quarterly construction wages in the St.
Louis SMSA, and yq, SAWCCq was estimated using the following model:
SAWCCq f0 f1 CPIq 2 b2
'
T '
yq
'
. (3)
6 Sports Stadium Construction 6 167
CPIq 2, the US Consumer Price Index for all items lagged 2 quarters, was used as the instrument.
We use this as a regressor because changes in wages are often due to cost-of-living adjustments,
which can be captured by including a measure of prices. Backstep selection was used
to determine the autoregressive nature of the disturbance yq
'
and performing a Lagrange Multiplier
test and a Q test to determine the existence of ARCH processes suggests the disturbance
follows the following AR(1,3)-ARCH(1) process:
yq
'
q
'
w'
yq 1
'
w2
'
yq 3
'
q
'
!hq
'
mq
'
hq
'
'
ď
q 1
2
mq
'
; iid~0,1!
Equation 3 was estimated using maximum likelihood estimation and predictions for nominal
SAWCCq were calculated. (Regression results are available from the author.) These predicted
values were then used in the estimation of Equation 2. Two second-stage models were
analyzed: one using predicted nominal wages and one using predicted real wages. Predicted
real wages were obtained by dividing predicted nominal wages by the St. Louis CPI for all
urban consumers. Backstep selection to determine the order of autocorrelation and Lagrangian
Multiplier and Q tests to determine the ARCH process on the disturbance yq suggests the
following AR(1, 8)- ARCH(1) process generates the disturbances for the equation using nominal
wages:
yq q w1 yq 1 w2 yq 8
q !hqmq
hq  dq 1
2
mq ; iid~0,1!
The same selection and test criteria were used to determine the ARCH process on the disturbance
yq suggests the following AR(1, 2)- ARCH(1) process generates the disturbances for
the equation using real wages:
yq q w1 yq 1 w2 yq 2
q !hqmq
hq  dq 1
2
mq ; iid~0,1!
The regression results were obtained using maximum likelihood estimation and are given
in Table 4. Consider the estimated equation using the predicted nominal wages. The model
fits the data well (R2
0.9840) and most variables are significant at least at the 10% level of
168 6 JOURNAL OF URBAN AFFAIRS 6 Vol. 24/No. 2/2002
TABLE 4
Two-Stage Least Squares Model: Second Stage Estimated Wages Only
Nominal Wages Real Wages
Variable
Parameter
Estimate
Standard
Error Variable
Parameter
Estimate
Standard
Error
Intercept** 17699 1743.1 Intercept** 100449 1103.6
SL--Predicted Quarterly Wages,
Construction** 1.016408 0.0508
SL--Predicted Quarterly Wages,
Construction** -1.986248 0.1904
SL--Building Permits, Total: Lag 1** 1.31001 0.2907 SL--Building Permits, Total: Lag 1 0.669251 0.909
KIEL -1441.83951 1383 KIEL -717.391222 5922.2
TW 18.056543 2096.2 TW 2511.01028 5335.7
D8889** 3812.29599 1478.5 D8889 4559.31856 3809.1
D1990** 3724.73980 1786 D1990 2943.33894 3844.4
D74Q2** -18585 1355 D74Q2 5043.16289 3961.8
D80Q2* -4397.34006 2563.7 D80Q2 -5143.89738 3978
D96Q3 3377.13357 2495 D96Q3 663.211524 6044.8
D73Q1 781.286829 629 D73Q1 -1932.78358 2962.2
D79Q2** 2172.63814 795.5 D79Q2 396.591709 4266.9
D83Q2** -2965.51869 947.3 D83Q2 -2940.47247 3408
D87Q2** -3929.48361 1094.9 D87Q2 -894.849068 4637.9
D88Q2* 1476.94402 783.9 D88Q2** -270.453755 29.2974
D95Q1** -2988.66835 1433 D95Q1** 3719.68702 361.8
D96Q4** 4000.08810 1249.8 D96Q4 2734.59618 3716.1
AR(1)** -0.820418 0.0661 AR(1)** -0.710236 0.2765
AR(8)** 0.208833 0.0667 AR(2) -0.275571 0.2895
ARCH0** 1874847 3.7774 ARCH0** 8757154 2.0928
ARCH1 2.60E-18 2.18E-11 ARCH1 2.12E-22 4.54E-10
R-Sq 0.984 R-Sq 0.961
Note. *p < .10. **p < .05.
6SportsStadiumConstruction6169
significance. A normality test suggests the estimated error is normally distributed (p-value
0.3917). Analysis also suggests that neither autocorrelation nor heteroscedasticity is a problem
with our estimated model.
The parameter estimate on the predicted quarterly nominal wage is significant and positive
suggesting that a $1000 increase in annual rate quarterly nominal wages brings an additional
worker into the construction industry. The parameter estimate on lagged building permits is
also positive and significant suggesting that an additional 1000 building permits issued in a
particular quarter will bring one more construction worker employment during the following
quarter. As in models 1 through 4, the parameter estimates on the Kiel and Trans World dummies
are insignificant.
Now consider the estimated equation using the predicted real wages. The model fits the data
well (R2
0.9623) and a normality test suggests the estimated error is normally distributed
(p-value 0.1703).
The parameter estimate on the predicted quarterly wage is significant and negative suggesting
that a $1000 increase in annual rate quarterly wages forces two workers out of the construction
industry. The parameter estimate on lagged building permits is also positive and
significant suggesting that an additional 2000 building permits issued in a particular quarter
will bring one more construction worker employment during the following quarter. As in all
previous models analyzed, the parameter estimates on the Kiel and Trans World dummies are
insignificant. Hence, despite the way we empirically model construction employment, the data
suggest that there was no more nor no less employment in St. Louis' construction industry
during the construction of the Kiel Center or the Trans World Dome.
DISCUSSION AND CONCLUSION
Independent economists agree that the economic impact statements used to justify public
subsidies for sport stadia overstate the economic benefits of the stadia and the teams they house.
They do so by providing gross benefits as the total benefits of the teams and stadiums rather
than subtracting the benefits that would have been derived had sports fans spent their incomes
in another fashion rather than on sports. They also do so by ignoring what would have been
done with public subsidies had they not been given to sports teams.
Previous studies have examined the impact of sports teams and facilities and have found no
impact on the economies of the host cities. The present study examines a potential but ignored
effect of major stadium and arena construction: its impact on the regional construction industry.
The case examined was that of the impact of stadium construction on the construction sector
of the St. Louis, Missouri, SMSA during the early and mid-1990s. By econometrically
modeling construction employment during the 1970s, 1980s and 1990s, we found that there
was neither more nor less construction employment within the St. Louis MSA during the time
the Kiel Center and the Trans World Dome were being constructed.
According to the RIMSII multipliers, 44% of the jobs created by a $1,000,000 construction
project in Missouri would be in the construction industry. The vast majority of these jobs would
be local. However, the present study suggests that instead of creating new construction jobs,
jobs were shifted from projects that would otherwise have been undertaken, resulting in no
new job creation in the construction industry.
In addition, there were no out-of-the ordinary wage changes that occurred during the three
and one-half year period when the stadiums in St. Louis were being constructed that can be
directly linked to the stadia. These results, coupled with the more extensive analysis on construction
employment, suggest that the net impact of stadium construction on construction employment
and worker incomes is zero.
170 6 JOURNAL OF URBAN AFFAIRS 6 Vol. 24/No. 2/2002
Given that the analysis in the present article focuses on the case study of St. Louis, one
may wonder about the generalizability of these findings: Is St. Louis a typical city? Table 5
presents some average macroeconomic indicators for 24 selected major league cities in the
US during the period from 1974 to 1999. Note that St. Louis is below average in three of the
four categories. In the fourth category, SMSA unemployment rate, it has a higher-thanaverage
unemployment rate. In only one category (real personal per-capita personal income
growth) is the difference between St. Louis and the average not statistically significant. These
data show that St. Louis is not a typical city: it is smaller, has a lower personal per-capita
personal income, and a higher unemployment rate. There is also no reason to believe that construction
makes up a different proportion of the St. Louis workforce than the other cities included
in Table 5. Consequently, if sports stadium construction were to have any positive effect on
employment and wages in a construction industry, we should see it in St. Louis.
Another question to ask is the following: were the economic conditions in the St. Louis SMSA
such that they would mask any potential effects that construction of the arenas might have
had on the construction industry? Table 6 presents the St. Louis unemployment rate and the
year-ago growth rate of per-capita personal income from the period from 1976 to 1998. Recall
that the Kiel Center was built from 1992 to 1994 and the Trans World Dome was built during
the period from 1993 to 1995. The unemployment rate was relatively high in 1992 and 1993,
TABLE 5
Selected Major Sports City Economic Indicators
City
Average Real
Per-Captia
Personal
Income
Average Real
Per-Capita
Personal Income
Year-Ago
Growth
Average
SMSA
Population
Average
Unemployment
Rate
Los Angeles 15794 1.14 10,461,491 7.47
Atlanta 14729 2.27 2,724,819 5.27
Baltimore 15194 1.72 2,313,527 6.28
Boston 16027 2.48 5,544,153 5.43
Chicago 16446 1.75 7,425,188 6.62
Cincinnati 14083 1.94 1,511,199 5.98
Cleveland 15174 1.56 2,252,113 6.53
Dallas/Fort Worth 15210 1.95 3,668,908 4.93
Denver 16219 2.05 1,584,649 4.91
Detroit 15359 1.69 4,344,968 8.48
Houston 15440 1.66 3,139,876 5.84
Kansas City 14628 1.60 1,540,576 5.32
Miami 13416 0.87 1,811,352 7.66
Milwaukee 15257 1.65 1,417,404 5.12
Minneapolis/Saint Paul 16268 2.10 2,412,096 4.17
New York 17467 2.04 8,523,633 8.13
San Francisco 19217 1.89 3,530,071 5.47
Philadelphia 15428 1.96 4,869,470 6.30
Phoenix 13611 1.84 2,021,957 5.19
Pittsburgh 14168 1.68 2,477,542 6.87
San Diego 14822 1.47 2,203,744 6.45
Seattle 17025 2.36 1,861,506 6.20
St Louis 14708 1.81 2,467,621 6.43
Tampa Bay 13324 2.10 1,868,262 5.52
6 Sports Stadium Construction 6 171
and fell quite precipitously in 1994. Per-capita personal income growth ranged from 1.1% to
2% during this period. During the mid 1990s, the US economy was recovering from the early1990s
recession, and this explains why the St. Louis unemployment rate fell in the mid1990s.
However, recall from the analysis that we found no effect that construction of the two
sports stadiums analyzed had any noticeable impact on employment in the construction industry
in St. Louis. This suggests that the growth in employment during 1994 must have come
from other sectors within the St. Louis economy.
Last, consider the use of public funds for the building of sports stadiums. One could plausibly
argue that if these funds were spent elsewhere in the economy, either by the government
(on a different project) or by consumers (in terms of tax breaks and the resulting consumption
spending), the resulting change in overall demand would stimulate new construction projects.
Thus, one would not expect public spending on sports stadiums to cause employment in the
construction industry to increase. Yet the method of financing does not explain why we do not
observe any more construction employment during the building of sports stadiums in a locale.
The results presented in this article suggest that construction workers at a given site would
have been working at another site. Whether the financing for either project is public or private
is immaterial.
The evidence within the independent literature on the economic impact of sports teams and
stadiums suggests that the only plausible positive effects are external benefits. However, Johnson,
Groothuis, and Whitehead (2001) argue that the externalities are less than the cost of a
TABLE 6
St. Louis Economic Indicators
Year
St. Louis
Unemployment
Rate
Real St. Louis
Per-Capita Personal
Income Year-Ago
Growth
1976 7.1 4.4
1977 6.6 4.4
1978 5.7 3.4
1979 5.4 -0.3
1980 8.2 -3.1
1981 8.5 0.6
1982 9.9 0.4
1983 10.6 3.7
1984 7.3 5.3
1985 7.4 2.6
1986 7.0 2.9
1987 6.6 1.8
1988 6.0 2.1
1989 5.5 1.2
1990 6.0 -0.2
1991 7.0 -1.3
1992 6.2 2.0
1993 6.3 1.1
1994 4.8 2.0
1995 4.7 1.8
1996 4.5 0.9
1997 4.2 3.5
1998 4.3 2.5
172 6 JOURNAL OF URBAN AFFAIRS 6 Vol. 24/No. 2/2002
new stadium. Hence, a government may plausibly provide some subsidies to sports teams, but
it should never fully fund the projects. If the public sector wants to generate economic development,
its money might be better spent elsewhere.
ACKNOWLEDGMENT: This manuscript has greatly benefited from the suggestions of four anonymous
reviewers. Any remaining errors or admissions are my own.
REFERENCES
Baade, R. A. (1996). Professional sports as catalysts for metropolitan economic development. Journal of
Urban Affairs, 18, 1­17
Baade, R. A., & Dye, R. F. (1990). The impact of stadiums and professional sports on metropolitan area
development. Growth and Change, 21, 1­14.
Badenhausen, K., Nikolov, C., Alkin, M., & Ozanian, M. K. (1997, June). More than a game: An inside
look at the raging bull market in sports franchises. Financial World, 40­50.
Coates, D., & Humphreys, B. R. (1999). The growth effects of sports franchises, stadia, and arenas. Journal
of Policy Analysis and Management, 18, 601­624
Coates, D., & Humphreys, B. R. (2000). The effect of sports on earnings and employment in US cities.
Unpublished working paper, University of Maryland-Baltimore County.
Johnson, B. K., Groothuis, P. A., & Whitehead, J. C. (2001). The value of public goods generated by a
major league sports team: The CVM approach. Journal of Sports Economics, 2, 6­21
Munsey, P., & Suppes, C. (2001). Ballparks. Available: http://www.ballparks.com [December 2001].
Noll, R., & Zimbalist, A. (Eds.). (1997). Sports, jobs, and taxes: The economic impact of sports teams
and stadiums. Brookings Institute.
Rosentraub, M., Swindell, D., Przybylski, M., & Mullins, D. R. (1994). Sports and downtown development
strategy: If you build it, will jobs come? Journal of Urban Affairs, 16, 221­239.
6 Sports Stadium Construction 6 173