Investment in stadia and regional economic development--evidence from FIFA World Cup 2006.
Feddersen, Arne ; Grotzinger, Andre Leao ; Maennig, Wolfgang 等
Introduction
A series of studies on Metropolitan Statistical Areas (MSAs) in the
USA revealed that new sport stadiums do not generate significant income
and/or employment effects in their host cities,1 challenging the
"boosters" view of many politicians and sport officials who
claim beneficial effects for the local economy (and hence, a
justification for public financial support).
Using the case of the new stadiums for the 2006 FIFA World Cup in
Germany, this paper is the first multivariate work that examines the
potential income and employment effects of new stadiums outside of the
USA. Such a study is generally interesting because it is set against the
background of the different urban structures in the USA and Europe. In
addition, a non-US study is especially interesting because of
decade-long US tradition of allocating the stadiums in suburban areas,
whereas European stadiums are mostly located near to the city center.
Nelson (2001) argued that (US-)studies concluding insignificant effects
on the home cities of stadiums are misleading because the data are based
on stadia built in the 1960s-1980s. Stadiums in inner cities might be
more efficient for the regional development of the cities, although for
suburban stadiums the effects on regional economic development are
insignificant or even negative (Melaniphy, 1996; Santee, 1996).
This study is the first work on this topic that conducts tests on
the basis of Difference-in-Difference (DD) model with levels and trends.
To address the problem of potential serial correlation in DD models
(Bertrand, Duflo, & Mullainathan, 2004), we use a serial correlation
consistent arbitrary variance-covariance matrix. As robustness check we
use the ignoring time series information (ITSI) model in a form that is
modified for non-synchronous interventions.
The aim of this paper is to isolate the effects of the construction
and use of modern sports stadiums built in connection with the 2006 FIFA
World Cup in Germany on income and employment at a regional scale. Our
analysis distinguishes between effects in the short term (period of
planning and construction, ending with the inauguration of the stadium)
and long term (afterwards). Our study adds evidence on the economic
impact of sport facilities built for mega-sports events, like the FIFA
World Cup, Olympic Games, or professional sports leagues, and to the
political debate on financing sports facilities and mega-sports events.
Background and Literature Review
"And the winner is ... Deutschland!" On June 6, 2000,
these were the words of FIFA President Joseph Blatter, announcing the
host of the 2006 FIFA World Cup. Following the (widely supported) bid by
the German Football Association (DFB), this was the starting signal for
a period of stadium construction (building new stadiums or major
renovations of existing stadiums). All in all, about 20 cities built new
stadiums before the decision of DFB and FIFA on the host cities was
published. FIFA and the local organizing committee fixed the number of
World Cup venues to 12 by means of an open tendering procedure. Thereby,
the DFB forced the competition between applying cities, encouraging them
to build a modern stadium to increase their chances to be elected as a
match venue. Finally, the 2006 FIFA World Cup in Germany was held in
Berlin, Cologne, Dortmund, Frankfurt, Gelsenkirchen, Hamburg, Hannover,
Kaiserslautern, Leipzig, Munich, Nuremberg, and Stuttgart. The
investment costs for new construction or major renovations totaled an
amount of nearly 1.6 billion [euro] for these 12 stadiums (Fifa, 2006).
(2) An additional 1.6 billion [euro] was invested into stadia-related
infrastructure in these cities. In the other cities, another 515 million
[euro] was spent for stadium construction work only.
The size of the cities elected as match venues for the FIFA World
Cup varies strongly. Many of the biggest German cities were among the
list, like Berlin (3.4 million/No. 1), Hamburg (1.7 million/No. 2),
Munich (1.2 million/No. 3), Cologne (1.0 million/No. 4), Frankfurt (0.65
million/No. 5), Dortmund (0.59 million/No. 7), and Stuttgart (0.57
million/No. 8). Also, medium sized cities were named as match venues,
like Gelsenkirchen (0.28 million / No. 21), and one small city with less
than 100,000 habitants (Kaiserslautern) was elected. The cases of
Leipzig and Kaiserslautern are special due to a strong political
influence in the selection process. Leipzig was chosen because at least
one city from eastern Germany should become a match venue, even if no
professional soccer team is located in the city. Kaiserslautern, on the
other hand, has a long tradition in professional soccer in Germany, and
it is the only venue in the federal state of Rhineland-Palatine.
Before the 2006 World Cup in Germany, a series of analyses was
published, according to which, the investments connected with staging
the World Cup as well as the expenditure of the expected 1-2 million
foreign visitors would markedly affect income and employment. As usual
for ex-ante impact studies, the predictions were optimistic. The
estimates fluctuated between a 2 billion [euro] and a 10 billion [euro]
increase in income or up to 10,000 additional jobs (see e.g., Deutsche
Industrie- und Handelskammer, 2006; Deutsche Postbank AG, 2005a, 2005b,
2006).
In some of the above studies, positive long-term effects are also
mentioned resulting from (1) consumer spending due to sports events
staged in the new stadiums including multiplier effects or (2)
intangible benefits like increased image, happiness, or feel-good--and
potentially inducing further investment.
Until now, only few ex-post analysis of the economic impact of the
2006 FIFA World Cup exist. The descriptive analyses of Brenke and Wagner
(2007) and Maennig (2007) found no effect of the event in countrywide
macroeconomic time series but mention feel-good effects and beneficial
effects on the economic structure toward a more service oriented
economy. Time series analyses of Allmers and Maennig (in press) isolate
a significant increase of overnight stays at hotels by foreigners
(approximately 700,000) and an additional 600 to 700 million [euro] (US
$830 to 970 million) in net national tourism income. (3) The only
econometric analysis on the 2006 World Cup using regional data by Hagn
and Maennig (2007)--concentrating on the event itself and not
considering effects during the construction period--finds no significant
job market effects of the 2006 World Cup.
The econometric ex-post studies published so far that elaborate on
other soccer World Cups do not form a contradiction to such a
pessimistic view. Szymanski (2002) collected data on the 20 largest
economies measured by current GDP over the last 30 years. Many of these
countries have hosted the Olympic Games or the World Cup at least once
in the past 30 years. Using a simple regression, he concluded that the
growth of these countries is significantly lower in World Cup years. (4)
Sterken (2006) finds that World Cups have a positive effect, which is
quite limited. Hagn and Maennig (2008) showed that the 1974 World Cup,
which was held in Germany, did not generate significant short- or
long-term employment effects in that country. Baade and Matheson (2004)
show that, for the 1994 World Cup in the USA, nine of the 13 host cities
suffered falls in growth. Overall, the 13 locations suffered losses on
balance of over US $9 billion.
Such findings are paralleled by studies on other large sport
events. Only few studies have found significant positive effects of
sports facilities and sports events ex-post. Baim (1994) found positive
employment effects for Major League Baseball (MLB) and the National
Football League (NFL) for 15 cities in the USA. Hotchkiss, Moore, and
Zobay (2003) found significant positive effects on employment in regions
of Georgia, affiliated or close to activities of the Atlanta Olympic
Games in 1996, but they did not find significant effects on wages.
Feddersen and Maennig (2009) rejected their findings on employment after
checking for serial correlation and allowing for a simultaneous test of
level and trend effects. Carlino and Coulson (2004) found positive
effects of NFL teams on central on housing rents but no significant
increase in wages, and they interpreted this finding as positive effects
of the NFL. Tu (2005) found significant positive effects of the FedEx
Field (Washington) on real estate prices of its neighborhood, as did
Ahlfeldt and Maennig (2008, 2009) for three arenas in Berlin, Germany.
Kasimati and Dawson (2009) simulated positive short-term income effects
of the 2004 Olympic Games in Greece. Finally, Jasmand and Maennig (2008)
found (limited) long-term positive income effects for German regions
which hosted the 1972 Olympic Games but no employment effects. (5)
The majority of studies suggest that sporting events or sports
stadia have little or no significant effect on regional wages, income,
and/or employment (e.g., Baade, 1987, 1994; Baade & Dye, 1990; Baade
& Sanderson, 1997). A number of works, particularly those of Coates
and Humphreys (1999, 2000, 2001, 2003) or Teigland (1999), have even
arrived at significant negative effects.
Coming back to the case of 2006 FIFA World Cup, Preuss, Schutte,
and Kurscheidt (2007), in a survey based analysis, estimated a primary
economic impact of about 3 billion [euro], resulting mainly from foreign
visitors. Compared with the German GDP of 2,303 billion [euro] in 2006,
this primary economic impact is 0.13%. Thus, it is not surprising that
econometric estimates from countrywide impact studies do not find a
significant effect.
The need for additional analysis on the basis of more disaggregated
and regionalized data is obvious.
Data
The analytical framework for this study comprises data of the 118
most populated large urban districts ("Kreisfreie Stadte") in
Germany in 1995, as reported by the Arbeitskreis Volkswirtschaftliche
Gesamtrechnung der Lander (2007b). (6) As variables for the regional
economic development, the income of private households per capita
(Arbeitskreis Volkswirtschaftliche Gesamtrechnungen Der Lander, 2007b)
as well as the number of people employed (Arbeitskreis
Volkswirtschaftliche Gesamtrechnungen Der Lander, 2007a) in these 118
large urban districts are considered.
[FIGURE 1 OMITTED]
For the income of private households, the period of observation is
1995 to 2005 (a time span of 11 years). A time span of 10 years is
considered for employment data (1996 to 2005). As the data availability
starts in 1995 or 1996, respectively, no structural breaks due to German
reunification have to be considered. (7)
The considered urban districts vary in employment and income size.
Employment numbers varies between about 15,500 and 1.5 million employed
persons. Income disperses between 10,833 [euro] and 23,615 [euro] per
capita with a mean income of 715,868. Due to this variation, we take the
logarithms of income and employment figures; coefficients can be
interpreted as a percentage impact.
Several additional indicators of the regional economic development
could be considered. Hotchkiss, Moore, and Zobay (2003), for instance,
suggested that the DD equation could be estimated for population. Or, a
sport venue or sport franchise (sport club) might increase the
attractiveness of a city from a resident's point of view. As a
consequence, ceteris paribus, migration into the city may occur. Even if
real wages are lower in a city offering such an attractive benefit than
in other cities that have no such sports facilities, some people might
move to this city because the attraction (compensating differentials).
Thus, initially, it might be appropriate to test for a population
effect. However, because it is difficult to assume that unemployed
persons will migrate due to the increased attractiveness of a city, we
can assume that most migrants will be working in their new city. Thus, a
strong correlation between population and employment exists, and an
additional DD analysis on population is unnecessary.
Method and Results
DD Model with Level and Trend
The aim of this paper is to examine if stadium construction
projects in Germany--especially those of the 2006 FIFA World Cup--had a
significant impact on the economic development of the regions in which
they are located. For this purpose, we use a DD estimation. This is a
common approach for identifying the effect of a specific intervention or
treatment. Therefore, one has to compare the differences in outcome
before and after an intervention for groups affected by the intervention
to the difference for unaffected groups (Bertrand, Duflo, &
Mullainathan, 2004).
The 12 cities elected as match venue will be used as the treatment
group in the DD model. During the period of observation, several
additional stadium construction projects were undertaken in Germany. To
avoid biased results, in addition to the FIFA World Cup stadiums, all
relevant stadium construction projects were used as the treatment group
in a second DD regression.
We focus our interest on differences in levels and trends for two
variables: employment and income. Because the stadium construction work
did not start at the same point in time for all cities (see Table 1),
the pre-period and the post-period are not the same for all cities of
the treatment group, and they are not even defined for the control
cities. Thus, in contrast to many DD models8, no dummy variable for the
post-period of all cities will be included. Equation (1) and (2) contain
the modified DD model:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (1)
where [Z.sub.it] is the income of private households in city i in
year t or the employment in city i in year t, respectively. [alpha]
denotes the intercept term. trend is a trend variable for all 118 large
urban districts starting with the value of one in year 1995 (1996) and
ends with a value of 11 (10) in year 2005. No dummy variable for the
treatment group is included because our model is a fixed effects model
(9) with separate dummies for all large urban districts capturing the
treatment group effects. [PT.sub.it] is a dummy for the post
intervention phase of the treatment group. It takes the value of one for
cities with relevant stadia construction projects from the year of the
start of the construction work (10) and zero otherwise. [TT.sub.it] is a
dummy variable indicating an overall trend for the treatment group,
which measures the difference in the growth trend between the treatment
group and the control group for the observation period. It is the
product of the variable trend and a binary variable taking the value of
one if a city belongs to the treatment group. [PTT.sub.it] denotes a
variable that covers a post period trend for the treatment cities. It is
the product of the variables [PT.sub.it] and trend. In the years before
the start of the construction project, it takes the value of zero, and
afterwards, it displays the corresponding value of the trend variable.
[[beta].sub.1], [[beta].sub.2], [[beta].sub.3], and [[beta].sub.4] are
coefficients to be estimated. [m.sub.i] covers the unobserved individual
specific effects (fixed effects), while [n.sub.it] denotes the remainder
disturbance.
The coefficients of interests are [[beta].sub.1] and [[beta].sub.4]
because they are measuring the level and trend effect of the
intervention (stadia construction project) of the treatment cities. If a
stadium construction project has positive regional effects, significant
positive signs should be found in the employment model.
In per capita models, positive short-term effects might result from
the direct spending for the construction work, especially if a sizable
amount of investment is financed from outside the city (e.g., by the
regional or federal government) and if price effect of the construction
demand is absent.
In the long run, the sign of coefficients of income is
theoretically ambiguous in the case positive economic effects of stadium
construction projects. If we assume that the attractiveness of a city
increases in the eyes of residents and non-residents (e.g., because of
an eye-catching new stadium and its associated feel-good effects), then
migration into the city may occur. If the population increases, the
labor supply might increase, potentially leading to decreasing wages
("compensating differentials;" Carlino & Coulson, 2004).
A separation of short-run and long-run effects is thus indicated,
but the model specification chosen in equation (1) does not allow
isolating short-run effects of the construction work.
To isolate the effect of the pure construction phase, a second
variant of the model (1) will be estimated:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (2)
The variables trend, [TT.sub.it], and [PT.sub.it] are identical to
those in model (1). To isolate the effects of the construction phase,
the dummy variable C takes the value of one during the construction work
and the value of zero otherwise. (11) [PTTC.sub.it] is a post
intervention trend for the treatment group that starts after the
construction work has finished because we expect that changes in the
growth trend will occur not due to the construction but, rather, due to
advancements in the attractiveness of the city that are derived only
from the completed stadium. It has to be admitted, though, that, due to
data limitations, for some stadia projects (e.g., Kaiserslautern or
Stuttgart), only a few observations are available for [PTTC.sub.it],
making it statistically demanding to isolate any post-construction
effects for these cities.
As shown by Bertrand, Duflo, and Mullainathan (2004), DD models are
frequently subject to serial correlation, which might lead to an
overestimation of the significance of the "intervention"
dummy. To check for such problems, we performed an LM test for serial
correlation in a fixed effects model as suggested by Baltagi (2001, p.
94-95). (12) This test is performed on the residuals of standard fixed
effects regressions of the above described models (1) and (2) for income
and employment. (13)
The LM statistic indeed rejects the null hypothesis of no serial
correlation in each case.
For such a case, Bertrand, Duflo, and Mullainathan (2004) suggest
using an arbitrary variance-covariance matrix, which is consistent in
the presence of any correlation pattern within cross section over time.
Table 6 and Table 7 show the regression results of the DD coefficients
and the corresponding t-statistics computed using an arbitrary
variance-covariance matrix.
In all eight estimated models, the trend variable trend is
significant at least at the 10% level. This means that there is a
positive trend in income and employment for all regarded 118 German
large urban districts. (14) The treatment trend dummy is insignificant
in almost all models, with the exception of model (1) in the employment
regression, implying that there is no systematic difference between the
treatment and control groups in the growth pattern of urban districts.
The coefficients of the post-period dummy [PT.sub.it] of the treatment
urban districts and the respective coefficient of the post-trend dummy
[PTT.sub.it]--the objects of interest--are insignificant for all
estimations. The results are not affected by accounting for a special
construction effect, as shown in model (2) of the income and employment
regressions. Thus, the hypothesis of no income and employment effect of
the stadia construction projects in the 12 and 20, respectively, urban
districts with completed stadia construction cannot be rejected.
Ignoring Time Series Information DD Model
To check robustness, we will use the ignoring time series
information (ITSI) model in its modification for non-synchronous
interventions (Bertrand, Duflo, & Mullainathan, 2004). In a first
step, [Z.sub.it] (equation 1 and 2) was regressed on city fixed effects,
time fixed effects, and relevant covariates. (15) In the second step,
the residuals of only the treatment group will be taken into account.
These residuals will be divided into two groups: (1) residuals from
years before the start of a stadia construction project and (2)
residuals from years after the start of a stadia construction project.
The stadia effect can then be analyzed by an OLS regression of a
two-period regression of the residuals from the treatment cities only.
Consistent t-statistics can be obtained from this OLS regression. (16)
The results of the ITSI models as shown in Table 8 and Table 9
confirm the findings of the DD model estimated earlier. No coefficient
in the ITSI models is significant on any conventional level. The results
of the robustness check support the results from the DD model using an
arbitrary variance-covariance matrix.
Conclusion
This is the first study to analyze the impact of the 2006 FIFA
World Cup or the construction work of the respective stadiums on both
employment and income. Using the method of difference-in-difference
estimation, we were not able to identify income or employment effects of
the construction of the new stadiums for the 2006 World Cup that are
significantly different from zero in the urban districts with completed
new stadiums in the period leading up to and after the 2006 FIFA World
Cup. In line with recent literature on DD estimates (Bertrand, Duflo,
& Mullainathan, 2004), we estimated serial correlation consistent DD
models if appropriate.
Our empirical results are well backed by theoretical
considerations. The absence of positive short-term effects might be
explained by crowding-out effects on "normal" investment
and/or tourism, including a "time-switching effect" of tourism
and a "carnival effect," describing increased travels abroad
by locals to avoid noise, traffic jams, and other turbulences that are
caused by the FIFA World Cup (Allmers & Maennig, in press). Also,
for the long-term, "booster" arguments suffer from a list of
methodological and theoretical problems (Coates & Humphreys, 2008,
pp. 298-299). Only the spending of true "foreigners" (17)
could be counted, and their share is often small for league matches.
Furthermore, much of the sport event related spending by local residents
comes out of their "entertainment or leisure budget." Thus,
from the viewpoint of a city's economy, this would be identical to
a merely substitution of consumer spending from entertainment to sports,
and no economic impact would be expected from this source.
Unfortunately, the multiplier of sport spending is likely to be smaller
than the multiplier for entertainment. In contrast to owners and
employees in the entertainment sector (e.g., theatres, cinemas, bowling
halls), owners and players from professional sports teams often do not
life within the city. Thus, regarding the high wages of professionals,
the multiplier of entertainment spending should be much higher than the
corresponding multiplier on sports spending. Finalizing this
argumentation, one might say that having a sports stadium and a
professional club might even decrease local income and employment rather
than boost it. Second, the transmission process from happiness to
fundamental economic indicators like income and employment seems to be
weak. So, additional economic growth due to the intangible benefits is
rather unlikely to occur. Third, if the expenditures of the city
government are financed by debts, classical crowding-out effects might
occur, which erase any positive long-term economic impact of a stadium.
We are thus unable to support local, state, or federal subsidies
for sports stadiums and sport events if a "booster effect" on
employment and income is the reasoning. Furthermore, we hesitate to
share the concern expressed both implicitly and explicitly in many of
the comparable sports economic studies that the positive effects of new
stadiums claimed by many sports protagonists are not true for two
reasons. Firstly, other effects such as the feel-good benefit for the
population and/or image effects that are difficult to quantify, may be
sufficiently important to justify major new stadiums and/or subsidies
for them via public funds. With image effects and feel-good effects,
economic empiricism in regards to sports is still in its infancy. (18)
Thus, further research on this topic is needed. Secondly and more
technically, the treatment group variables in the selected form of
municipality areas might still be too large and too highly aggregated to
statistically prove significant effects. Studies on the effects of major
sports venues on property values in surrounding areas indicate a maximum
affect area of some 3,000 meters (Ahlfeldt & Maennig, 2008, 2009;
Tu, 2005).
Appendix
Table A1. Population of the 118 largest urban districts
("kreisfreie Stadte") in Germany in 1995
No. City Population in 1995
1 Berlin 3,471,003
2 Hamburg 1,707,251
3 Munchen 1,240,465
4 Koln 964,597
5 Frankfurt am Main 651,097
6 Essen 616,340
7 Dortmund 599,966
8 Stuttgart 586,954
9 Dusseldorf 572,171
10 Bremen 549,157
11 Duisburg 535,473
12 Leipzig 524,870
13 Hannover 523,574
14 Dresden 496,863
15 Nurnberg 493,940
16 Bochum 400,608
17 Wuppertal 382,600
18 Saarbrucken Stadtverband 358,365
19 Bielefeld 324,115
20 Mannheim Universitatsstadt 313,880
21 Gelsenkirchen 292,061
22 Bonn 291,863
23 Chemnitz 291,331
24 Halle (Saale) 287,052
25 Karlsruhe 276,544
26 Wiesbaden 266,532
27 Monchengladbach 266,095
28 Munster 264,696
29 Magdeburg 262,557
30 Augsburg 260,952
31 Braunschweig 253,513
32 Krefeld 249,821
33 Aachen 247,460
34 Kiel 246,595
35 Rostock 230,768
36 Oberhausen 224,896
37 Lubeck Hansestadt 216,933
38 Hagen 212,909
39 Erfurt 212,532
40 Kassel 201,628
41 Freiburg im Breisgau 198,394
42 Mainz 184,329
43 Hamm 183,734
44 Herne 179,973
45 Mulheim an der Ruhr 176,602
46 Osnabruck 168,106
47 Ludwigshafen am Rhein 167,872
48 Solingen 165,794
49 Leverkusen 162,051
50 Oldenburg (Oldenburg) 150,540
51 Potsdam 144,941
52 Darmstadt 138,973
53 Heidelberg 138,612
54 Bremerhaven 130,720
55 Cottbus 127,791
56 Wurzburg 127,627
57 Wolfsburg 126,782
58 Regensburg 125,809
59 Gera 124,971
60 Remscheid 122,710
61 Heilbronn 121,745
62 Bottrop 120,008
63 Pforzheim 118,460
64 Salzgitter 117,776
65 Schwerin 116,876
66 Offenbach am Main 116,460
67 Ulm Universitatsstadt 115,379
68 Zwickau 112,646
69 Ingolstadt 111,626
70 Koblenz 109,292
71 Furth 108,011
72 Kaiserslautern 101,970
73 Jena 101,724
74 Erlangen 101,372
75 Trier 99,379
76 Dessau 92,030
77 Wilhelmshaven 90,944
78 Brandenburg an der Havel 87,713
79 Flensburg 87,642
80 Neumunster 82,030
81 Neubrandenburg 81,786
82 Frankfurt (Oder) 81,633
83 Worms 79,737
84 Delmenhorst 78,079
85 Plauen 73,318
86 Bayreuth 72,692
87 Bamberg 69,901
88 Gorlitz 68,773
89 Stralsund 66,944
90 Aschaffenburg 66,339
91 Weimar 62,257
92 Greifswald 61,688
93 Kempten (Allgau) 61,494
94 Hoyerswerda 61,441
95 Landshut 59,257
96 Rosenheim 58,704
97 Schweinfurt 55,598
98 Suhl 53,986
99 Neustadt an der Weinstrasse 53,828
100 Baden-Baden 52,677
101 Hof 52,628
102 Emden 51,653
103 Passau 51,035
104 Wismar 50,870
105 Speyer 49,575
106 Pirmasens 48,562
107 Frankenthal (Pfalz) 47,946
108 Eisenach 45,642
109 Amberg 44,177
110 Straubing 44,022
111 Coburg 43,948
112 Weiden i.d.OPf. 43,171
113 Kaufbeuren 42,694
114 Memmingen 40,492
115 Ansbach 39,638
116 Landau in der Pfalz 39,632
117 Schwabach 37,564
118 Zweibrucken 36,039
Source: Arbeitskreis Volkswirtschaftliche Gesamtrechnung
der Lander (2007b).
References
Ahlfeldt, G., & Maennig, W. (2008). The impact of sports arenas
on land values: Evidence from Berlin. The Annals of Regional Science.
Retrieved from http://dx.doi.org/10.1007/s0016800008-00249-00164
Ahlfeldt, G., & Maennig, W. (2009). Arenas, arena architecture
and the impact on location desirability: The case of 'Olympic
arenas' in Prenzlauer Berg, Berlin. Urban Studies, 46(7), 13431362.
Allmers, S., & Maennig, W. (in press). Economic impacts of the
FIFA soccer World Cups in France 1998, Germany 2006, and Outlook for
South Africa 2010. Eastern Economic Journal.
Arbeitskreis Volkswirtschaftliche Gesamtrechnungen der Lander.
(2007a). Compensation of employees, gross wages and salaries in Germany
on Nuts3-Level 1996 to 2005 [Electronic Version]. Retrieved from
http://www.vgrdl.de/Arbeitskreis_VGR/R2B2.zip
Arbeitskreis Volkswirtschaftliche Gesamtrechnungen der Lander.
(2007b). Income of private households in Germany on Nuts3-Level 1995 to
2005 [Electronic Version]. Retrieved from
http://www.vgrdl.de/Arbeitskreis_VGR/R2B3.zip
Atkinson, G., Mourato, S., Szymanski, S., & Ozdemiroglu, E.
(2008). Are we willing to pay enough to 'back the bid'?:
Valuing the intangible impacts of London's bid to host the 2012
Summer Olympic Games. Urban Studies, 45(2), 419-444.
Baade, R. A. (1987). Is there an economic rationale for subsidizing
sports stadiums? Heartland Policy Study No.13.
Baade, R. A. (1994). Stadiums, professional sports, and economic
development: Assessing the reality. The Heartland Institute Policy
Study, No. 62.
Baade, R. A. (1996). Professional sports as catalysts for
metropolitan economic development. Journal of Urban Affairs, 18(1),
1-17.
Baade, R. A., & Dye, R. F. (1990). The impact of stadiums and
professional sports on metropolitan area development. Growth and Change,
21(2), 1-14.
Baade, R. A., & Matheson, V. (2004). The quest for the Cup:
Assessing the economic impact of the World Cup. Regional Studies, 38(4),
343-354.
Baade, R. A., & Sanderson, A. R. (1997). The employment effect
of teams and sports facilities. In R. G. Noll & A. Zimbalist (Eds.),
Sports, jobs, and taxes: The economic impact of sports teams and
stadiums (pp. 92-118). Washington, D.C.: Brookings Institution Press.
Baim, D. V. (1994). The sports stadium as a municipal investment.
Westport and London: Greenwood Press.
Baltagi, B. H. (2001). Econometric analysis of panel data (2nd
ed.). New York: Wiley & Sons.
Bertrand, M., Duflo, E., & Mullainathan, S. (2004). How much
should we trust differences-indifferences estimates? Quarterly Journal
of Economics, 119(1), 249-275.
Brenke, K., & Wagner, G. G. (2007). Zum volkswirtschaftlichen
Wert der Fussball-Weltmeisterschaft 2006 in Deutschland. DIW Berlin
Research Notes 19.
Buttner, N., Maennig, W., & Menssen, M. (2007). Relationships
between investment costs for infrastructure and sport stadia: The case
of the World Cup 2006 in Germany. Verkehrswissenschaft, 78(3), 145-175.
Carlino, G., & Coulson, N. E. (2004). Compensating
differentials and the social benefits of the NFL. Journal of Urban
Economics, 56(1), 25-50.
Coates, D., & Humphreys, B. R. (1999). The growth effects of
sport franchises, stadia, and arenas.
Journal of Policy Analysis and Management, 18(4), 601-624.
Coates, D., & Humphreys, B. R. (2001). The economic
consequences of professional sports strikes and lockouts. Southern
Economic Journal, 67(3), 737-747.
Coates, D., & Humphreys, B. R. (2003). Professional sports
facilities, franchises and urban economic development. Public Finance
and Management, 3(3), 335-357.
Coates, D., & Humphreys, B. R. (2008). Do economists reach a
conclusion on subsidies for sports franchises, stadiums, and
mega-events? Econ Journal Watch, 5(3), 294-315.
Deutsche Industrie- und Handelskammer. (2006). Fussball-WM 2006,
Auswirkungen auf die Unternehmen. Retrieved from
http://www.dihk.de/index.html?/inhalt/themen/branchen/
tourismus/fussball/wirtschaft.html
Deutsche Postbank AG. (2005a). FIFA Fussball-Weltmeisterschaft
2006--Deutsche Wirtschaft steht als gewinner bereits fest. Retrieved
from http://www.postbank.de/Datei/R SpezialFebruar05.pdf
Deutsche Postbank AG. (2005b). FIFA Fussball-Weltmeisterschaft
2006--Signifikante Arbeitsplatz- und Beschaftigungszuwachse in einzelnen
Branchen. Retrieved from http://www.post
bank.de/Datei/RSpezialJuli05.pdf
Deutsche Postbank AG. (2006). FIFA Fussball-Weltmeisterschaft
2006--Kleine und "Armere" Austragungsorte profitieren am
meisten. Retrieved from http://postbank.de/Datei/ RSpezialJanuar06.pdf
Feddersen, A., & Maennig, W. (2009). Wage and employment
effects of the Olympic Games in Atlanta 1996 reconsidered. Hamburg
Contemporary Economic Discussions, No. 25.
FIFA. (2006). World Cup 2006 in Germany, stadia [Electronic
Version]. Retrieved from
http://fifaworldcup.yahoo.com/06/de/d/stadium/index.html
Hagn, F., & Maennig, W. (2007). Labour market effects of the
2006 Soccer World Cup in Germany. Applied Economics, iFirst. Retrieved
from http://dx.doi.org/10.1080/ 00036840701604545
Hagn, F., & Maennig, W. (2008). Employment effects of the
football World Cup 1974 in Germany. Labour Economics, 15(5), 1062-1075.
Heyne, M., Maennig, W., & Sussmuth, B. (2007). Mega-sporting
events as experience goods.
Hamburg Contemporary Economic Discussions, No. 05.
Hotchkiss, J. L., Moore, R. E., & Zobay, S. M. (2003). Impact
of the 1996 Summer Olympic Games on employment and wages in Georgia.
Southern Economic Journal, 69(3), 691-704.
Jasmand, S., & Maennig, W. (2008). Regional income and
employment effects of the 1972 Munich Summer Olympic Games. Regional
Studies, 42(7), 991-1002.
Kasimati, E., & Dawson, P. (2009). Assessing the impact of the
2004 Olympic Games on the Greek economy: A small macroeconometric model.
Economic Modelling, 26(1), 139-146.
Kavetsos, G., & Szymanski, S. (2008). National wellbeing and
international sports events. IASE Working Paper, No. 0804.
Kim, H. J., Gursoy, D., & Lee, S.-B. (2006). The impact of the
2002 World Cup on South Korea: Comparisons of pre- and post-games.
Tourism Management, 27(1), 86-96.
Kim, S. S., & Petrick, J. F. (2005). Residents'
perceptions on impacts of the FIFA 2002 World Cup: The case of Seoul as
a host city. Tourism Management, 26(1), 25-38.
Maennig, W. (2007). One year later: A re-appraisal of the economics
of the 2006 soccer World Cup. Hamburg Contemporary Economic Discussions,
No. 10.
Melaniphy, J. C. (1996). The impact of stadiums and arenas. Real
Estate Issues, 21(3), 36-39.
Nelson, A. C. (2001). Prosperity or bligth? A question of major
league stadia locations. Economic Developmet Quarterly, 15(3), 255-265.
Preuss, H. (2006). The economics of staging the Olympics: A
comparison of the Games 1972-2008. Cheltenham: Edward Elgar.
Preuss, H., Schutte, N., & Kurscheidt, M. (2007). Measuring the
primary economic impact of visitors to mega-sport events--A case study
at the FIFA World Cup 2006. Paper presented at the AK Sportokonomie,
Magglingen / CH, May 11, 2007.
Santee, E. E. (1996). Major league cities. Real Estate Issues,
21(3), 31-35.
Skrentny, W. (2001). Das Grosse Buch der Deutschen Fussball-Stadien
[the Allmanach of the German Soccer Stadia] (2 ed.). Gottingen: Verlag
Die Werkstatt.
Stadionwelt. (2007). Stadium. Neu- Und Umbau [Stadia. New and
Reconstruction] [Electronic Version]. Retrieved from
http://www.stadionwelt.de/neu/sw_stadien/index.php?
folder=sites&site=neubau_d
Sterken, E. (2006). Growth impact of major sporting events.
European Sport Management Quarterly, 6(4), 375-389.
Szymanski, S. (2002). The economic impact of the World Cup. World
Economics, 3(1), 169-177.
Teigland, J. (1999). Mega-events and impacts on tourism; The
predictions and realities of the Lillehammer Olympics. Impact Assessment
and Project Appraisal, 17(4), 305-317.
Tu, C. C. (2005). How does a new sports stadium affect housing
values? The case of Fedex Field. Land Economics, 81(3), 379-395.
Endnotes
(1) See Baade (1987, 1994, 1996); Baade & Dye (1990); Baade
& Sanderson (1997); Coates & Humphreys (1999, 2000, 2001, 2003).
(2) Every World Cup stadium was at least renovated. The average
expenditure per city was 116.7 million [euro], with a minimum investment
of 736.0 million (Dortmund) and a maximum investment of 280.0 million
[euro] (Munich). The volume of transport infrastructure investment was
even larger (Buttner, Maennig, & Menssen, 2007).
(3) Allmers and Maennig (in press) also analyze the effects of 1998
Football World Cup in France on overnight stays, national income from
tourism, and retail sales and find significant effects at all.
(4) No significant effects at all are registered for the Olympic
Games.
(5) In addition to econometric analysis with "realized
data," there are studies for perceived benefits, see Kim and
Petrick (2005) and Kim, Gursoy, and Lee (2006) for the 2002 World Cup.
(6) See Table A1 in the appendix for a complete list of the large
urban districts.
(7) Start and end of the observation periods are determined by data
availability from EUROSTAT and VGRDL.
(8) See e.g., Hotchkiss, Moore, and Zobay (2003) for the use of a
general post period dummy.
(9) The F-tests of the significance of the fixed effects reject the
null hypothesis that all equals zero.
Thus, we consider a fixed effects model as appropriate.
(10) As the employment and income data are on a yearly basis and as
the construction work does not always starts at the beginning of year,
no effect could be found for a year in which a construction project
starts at the year's end. To deal with this problem, stadia
constructions will be considered only for a specific year if the start
of work lies in first three quarters of this year. If the construction
work started in the last quarter of a year, the following year will be
treated as starting point.
(11) The periods of construction can be found in columns 6 and 7 of
Table 1. As construction work is not always started at the beginning of
a year, the dummy takes the value of one if the works starts before
October or does not end before April of the respective year.
(12) The LM test statistic is [MATHEMATICAL EXPRESSION NOT
REPRODUCIBLE IN ASCII], which is asymptomatically distributed N (0,1).
(13) The "intervention" coefficients of these regressions
are often significant. But, in line with Bertrand, Duflo, and
Mullainathan (2004), the estimates might be inefficient.
(14) Substituting the time trend by time invariant fixed effects
does yield comparable results. The variables of interest ([PT.sub.it],
[PTT.sub.it]) change only marginal and are still insignificant.
(15) As done in the previous section, two different variants have
been analyzed: (1) no covariates are considered; (2) only a construction
dummy is considered.
(16) As the numbers of cities is not small, the t-statistics
don't have to be adjusted.
(17) "Foreigners" are defined as consumers coming from
outside of the city for a visit in the city.
In the framework of economic impact studies, the requirements to be
named a foreigner are even stronger. If they switched only the time of
their visit to a date when a match will be played, no additional money
and thus no economic impact will be generated ("time
switchers"). For this and further methodological problems
concerning "foreigners" see e.g., Preuss (2006).
(18) For the measurement of the benefit of the 2012 Olympic Games
in London see Atkinson et al. (2008). Heyne, Maennig, and Sussmuth
(2007) show a study of intangible effects of the 2006 FIFA World Cup.
For a first attempt to measure the impact of sports events as well as
success of a national team on happiness, see Kavetsos and Szymanski
(2008).
Arne Feddersen [1], Andre Leao Grotzinger [1], and Wolfgang Maennig
[1]
[1] University of Hamburg
Arne Feddersen is a research fellow in the Department of Economics.
His research interests include sports economics, applied regional
economics, and media economics.
Andre Leao Grotzinger is an associate director in private banking.
His research interests include sports economics and regional economics.
Wolfgang Maennig is a professor and the chair for Economic Policy
in the Department of Economics. His research interests include sports
economics, transportation economics, and real estate economics.
Table 1: Relevant Stadia Construction Projects in Germany,
1996 to 2005
City Stadium Capacity Team(s)
Berlin# Olympiastadion# 74,000# Hertha BSC
Berlin#
Bremen Weserstadion 42,100 Werder Bremen
Cologne# RheinEnergy- 50,374# 1. FC Koln#
Stadion#
Cottbus Stadion der 22,746 FC Energie
Freundschaft Cottbus
Dortmund# Signal Iduna Park# 83,000# Borussia
Dortmund#
Dusseldorf LTU arena 52,000 Fortuna
Dusseldorf
Duisburg MSV-Arena 31,514 MSV Duisburg
Frankfurt# Commerzbank- 51,500# Eintracht
Arena# Frankfurt#
Gelsen- Veltins-Arena# 61,524# FC Schalke 04#
kirchen#
Hamburg# HSH-Nordbank- 57,000# Hamburger SV#
Arena#
Hannover# AWD-Arena# 49,000# Hannover 96#
Kaisers- Fritz-Walter- 48,500 1, FC#
lautern# Stadion# Kaiserslautern#
Leipzig# Zentralstadion# 44,193# Sachsen Leipzig#
Magdeburg Stadion Magdeburg 27,000 1, FC Magdeburg
Monchen- Borussia-Park 54,057 Borussia
gladbach M'gladbach
Munich# Allianz Arena# 69,901# FC Bayern
Munchen#
Nuremberg# easyCredit-Stadion# 46,780# 1, FC Nurnberg#
Rostock DKB-Arena 30,000 FC Hansa Rostock
Stuttgart# Gottlieb-Daimler- 55,896# VfB Stuttgart#
Stadion#
Wolfsburg Volkswagen Arena 29,161 VfL Wolfsburg
City Costs Construction
Start End
Berlin# 242.0# Aug 2000# Aug 2004#
Bremen 18.0 May 2003 Jul 2004
Cologne# 117.5# Jan 2002# Jul 2004#
Cottbus 12.0 Apr 2002 Jul 2003
Dortmund# 36.0# May 2002# Jul 2003#
Dusseldorf 218.0 Sep 2002 Jan 2005
Duisburg 43.0 Oct 2003 Jan 2005
Frankfurt# 126.0# Jul 2002# May 2005#
Gelsen- 192.0# Nov 1998# Jul 2001#
kirchen#
Hamburg# 97.0# Jun 1998# Aug 2000#
Hannover# 63.0# Feb 2003# Jan 2005#
Kaisers- 48.3# Aug 2004# Apr 2006#
lautern#
Leipzig# 90.6# Dec 2000# March 2004#
Magdeburg 30.9 March 2005 Dec 2006
Monchen- 87.0 Jan 2002 Jul 2004
gladbach
Munich# 280.0# Feb 2002# May 2005#
Nuremberg# 56.0# Nov 2003# Jul 2005#
Rostock 55.0 May 2000 Aug 2001
Stuttgart# 51.6# Jan 2004# Jan 2006#
Wolfsburg 51.0 May 2001 Nov 2002
Source: Skrentny (2001); FIFA (2006); Stadionwelt (2007);
FIFA World Cup 2006 stadia are marked in bold letters.
Note: FIFA World Cup 2006 stadia are marked indicated
with #.
Table 2: Descriptive Statistics (nominal values)
Income
All Control Treatment
Group Group
N 118 106 12
Mean 15,868 15,786 16,266
Median 15,773 15,769 15,881
Std. Dev. 2,192 2,086 2,619
Min 10,833 10,833 11,287
Max 23,615 23,615 22,908
Employment
All Control Treatment
Group Group
N 118 106 12
Mean 128,122 76,558 380,784
Median 67,543 56,938 281,361
Std. Dev. 186,240 55,618 336,195
Min 15,477 15,477 57,189
Max 1,436,573 290,340 1,436,573
Table 3: Descriptive Statistics (logarithms)
Income
All Control Treatment
Group Group
N 118 106 12
Mean 9.6626 9.6582 9.6840
Median 9.6661 9.6658 9.6728
Std. Dev. 0.1378 0.1324 0.1600
Min 9.2904 9.2904 9.3314
Max 10.0696 10.0696 10.0392
Employment
All Control Treatment
Group Group
N 118 106 12
Mean 11.2501 11.0018 12.4668
Median 11.1205 10.9497 12.5458
Std. Dev. 0.9210 0.7003 0.9063
Min 9.6471 9.6471 10.9541
Max 14.1778 12.5788 14.1778
Table 4: Test for Significance of the Fixed Effects
Endogenous Model (1)
variable Treatment WC Treatment ALL
Income 319.16 *** 321.34 ***
Employment 4,690.77 *** 4,567.26 ***
Endogenous Model (2)
variable Treatment WC Treatment ALL
Income 318.57 *** 322.20 ***
Employment 4,584.37 *** 4,499.86 ***
Notes: *** p <0.01, ** p <0.05, * p <0.10.
Table 5: Test for Serial Correlation
Endogenous Model (1)
variable Treatment WC Treatment ALL
Income 23.411 *** 23.413 ***
Employment 33.964 *** 23.186 ***
Endogenous Model (2)
variable Treatment WC Treatment ALL
Income 23.249 *** 23.251 ***
Employment 23.367 *** 23.371 ***
Notes: *** p <0.01, ** p <0.05, * p <0.10.
Table 6: DD Model with Fixed Effects for Income of Private Households
Model (1)
Treatment Treatment
WC_12 ALL_20
Constant 9.551 *** 9.550 ***
(3,539.662) (3,663.220)
PT [-6.109e.sup.-4] -0.008
(-0.432) (-0.820)
C -- --
trend 0.019 *** 0.019 ***
(38.905) (37.428)
TT -0.005 [6.297e.sup.-4]
(-0.465) (0.733)
PTT [8.260e.sup.-4] [-7.980e.sup.-5]
(0.201) (-0.018)
PTTC -- --
[R.sup.2] 0.881 0.881
adj. [R.sup.2] 0.881 0.881
F-Stat 441.250 *** 550.130 ***
N 118 118
T 11 11
N*T 1298 1298
Model (2)
Treatment Treatment
WC_12 ALL_20
Constant 9.551 *** 9.551 ***
(3,502.198) (3,594.649)
PT -0.017 -0.020 *
(-1.257) (-1.787)
C -0.005 -0.008 *
(-0.694) (-1.742)
trend 0.019 *** 0.019 ***
(38.897) (37.413)
TT [8.063e.sup.-4] [2.243e.sup.-4]
(-0.528) (0.213)
PTT -- --
PTTC 0.010 0.009
(1.365) (1.228)
[R.sup.2] 0.882 0.882
adj. [R.sup.2] 0.881 0.881
F-Stat 352.928 374.450 ***
N 118 118
T 11 11
N*T 1298 1298
Notes: *** p<0.01, ** p<0.05, * p<0.10. t-statistics are in
parentheses. Standard errors are computed using an arbitrary
variance-covariance matrix as suggested by Bertrand, Duflo,
& Mullainathan (2004, pp. 270-272).
Table 7: DD Model with Fixed Effects for Employment
Model (1)
Treatment Treatment
WC_12 ALL_20
Constant 11.244 *** 11.245 ***
(1,804.627) (1,954.357)
PT 0.006 0.015
(0.723) (1.106)
C -- --
trend [2.189e.sup.-4] [3.647e.sup.-4]
(0.222) (0.375)
TT 0.006 0.005
(1.494) (1.224)
PTT -0.008 -0.006
(-1.523) (-1.146)
PTTC -- --
[R.sup.2] 0.979 0.979
Adj.[R.sup.2] 0.979 0.979
F-Stat 10,256.240 *** 10,330.930 ***
N 118 118
T 10 10
N*T 1,180 1,180
Model (2)
Treatment Treatment
WC_12 ALL_20
Constant 11.245 *** 11.244 ***
(1,950.524) (1,868.912)
PT 0.015 0.002
(0.623) (0.100)
C -0.009 -0.010
(-1.159) (-1.371)
trend [9.702e.sup.-4] [2.189e.sup.-4]
(0.003) (0.222)
TT 0.004 0.005
(1.089) (1.413)
PTT -- --
PTTC -0.001 -0.007
(-0.071) (-0.525)
[R.sup.2] 0.979 0.979
Adj.[R.sup.2] 0.978 0.978
F-Stat 9,620.552 *** 9,596.174 ***
N 118 118
T 10 10
N*T 1,180 1,180
Notes: *** p<0.01, ** p<0.05, * p<0.10. t-statistics are in
parentheses. Standard errors are computed using an arbitrary
variance-covariance matrix as suggested by Bertrand, Duflo,
& Mullainathan (2004, pp. 270-272).
Table 8: ITSI DD Model for Income of Private Households
Model (1)
Treatment WC Treatment ALL
Constant 0.046 0.020
(1.081) (0.596)
POST -0.006 -0.001
(-0.102) (-0.030)
[R.sup.2] 0.045 0.019
adj. [R.sup.2] 0.001 0.007
Model (2)
Treatment WC Treatment ALL
Constant 0.047 0.022
(1.111) (0.635)
POST -0.005 -0.001
(-0.084) (-0.020)
[R.sup.2] 0.030 0.024
adj. [R.sup.2] 0.014 0.002
Notes: *** p<0.01, ** p<0.05, * p<0.10. t-statistics
are in parentheses. Coefficients are from a two-step
process using OLS.
Table 9: ITSI DD Model for Employment
Model (1)
Treatment WC Treatment ALL
Constant 1.573 *** 1.192 ***
(6.016) (5.734)
POST -0.008 -0.004
(-0.022) (-0.013)
[R.sup.2] 0.095 0.057
adj. [R.sup.2] 0.054 0.032
Model (2)
Treatment WC Treatment ALL
Constant 1.577 *** 1.196 ***
(6.050) (5.760)
POST -0.014 -0.003
(-0.039) (-0.010)
[R.sup.2] 0.035 0.016
adj. [R.sup.2] 0.009 0.010
Notes: *** p<0.01, ** p<0.05, *p<0.10. t-statistics
are in parentheses. Coefficients are from a two-step
process using OLS.
