The impact of the fifa world Cup 2010 on unemployment in South Africa Michiel Antoine Oosterbaan

3. Method and data 
This research examines the employment effects of the FIFA World Cup 2010 on the host 
cities in South Africa. The method is based on Hagn and Maennig’s (2009) multivariate 
study. In their quest to assess to unemployment effects of the FIFA World Cup 2006 in 
Germany, they compared the differences in unemployment between the host cities and non 
host cities in a sample of the 75 largest cities in terms of their population prior to the event.
To capture to which extent the development of unemployment between the two groups differ, 
Hagn and Maennig (2009) use models applied by Baade and Matheson (2000, 2001, 2002, 
2004), Coates and Humphreys (1999, 2000, 2001, 2002, 2003a, b) and Hotchkiss et al. (2003) 
and add their own model to test for method sensitivity. The first two methods test for 
employment effects during the course of the World Cup and monthly data is required. 
Therefore this research focuses on Hotchkiss’s et al. (2003) difference-in-difference method 
plus the application of Hagn and Maennig (2009). 
The difference-in-difference method is used to determine whether the variable of interest, in 
this case unemployment, changes more for one group after an event than the other group 
under observation. The basic idea is to add dummy variables to an OLS regression for 
whether the time period is pre or post event, whether the group under observation is affected 
or not and an interaction of the two indicators. To examine difference-in-difference in the 
intercept, the model takes the following form: 
with: 
lnUnemp
i,t
log unemployment in municipality i at time t, 
lnPop2003
i

log population in municipality i at in the year 2003, 
LF2003
i
 
share of gross value added of the agriculture, hunting, forestry, logging 
and fisheries sector in municipality i in the year 2003, 
HV2003
i 
share of gross value added of the trade, hospitality industry and traffic 
sector in municipality i in the year 2003, 

---

14 
MI2003
i
share of gross value added of the mining industry sector in municipality 
i in the year 2003, 
DL2003
i 
share of gross value added of the community service industry sector in 
municipality i in the year 2003, 
Prod2003
i
 
share of gross value added of the manufacturing industry sector in 
municipality i in the year 2003, 
WC
i 
dummy for match venues of the World Cup 2010 (1 for match venue, 0 
if not a match venue), 
Post
t 
dummy for period after the World Cup 2010 (1 for period after, 0 for 
period before the World Cup) and 
PostWC
i, t 
dummy for match venues and period after the World Cup 2010 (1 if 
match venue and period after the World Cup, otherwise 0). 
In this framework, the variables lnPop2003
i,
LF2003
i
, HV2003
i
, MI2003
i
, DL2003
i
and 
Prod2003
i
control for observable differences between the host municipalities and non host 
municipalities that might entangle the analysis. The location of the FIFA World Cup 2010 
impact is controlled for by the dummy variable WC
i
and the time dummy variable Post
i
controls for whether the event had taken place or not. The variable of interest PostWC
i, t
measures the unemployment in the host municipalities, post-FIFA World Cup 2010 relative to 
pre-FIFA World Cup 2010 compared to the unemployment in the non host municipalities, 
post- relative to pre-FIFA World Cup 2010, the control group.
To test whether the growth rate at which unemployment has changed post- relative to pre-
FIFA World Cup 2010 is different for host municipalities compared to non host 
municipalities, the following specification is modified in accordance with Hotchkiss et al. 
(2003): 
 
with: 

---

15 
Trend 
time trend, 
TrWC
i 
trend variable for match venues of the World Cup 2010 (1 if match 
venue and 1
st
phase of the period under consideration, 2 if match venue 
and 2
nd
phase of the period, etc.., otherwise 0), 
TrPost
t 
trend variable for period after the World Cup 2010 (1 if 1
st
phase after 
the World Cup, 2 if 2
nd
phase, etc.., otherwise 0) and 
TrPostWC
i, t 
trend variable for match venues and period after the World Cup 2010 (1 
if match venue and 1
st
phase after the World Cup, 2 if match venue and 
2
nd
phase after the World Cup, etc.., otherwise 0). 
Again the variables lnPop2003
i,
LF2003
i
, HV2003
i
, MI2003
i
, DL2003
i
and Prod2003
i
are 
used to take into account observable differences between municipalities and time period and 
location are controlled by TrPost
t
and TrWC
i
variables indicating whether the FIFA World 
Cup 2010 had taken place or not and whether the municipality is a host venue or not. The 
difference with model (1) is that model (2) allows the impact of time and location to show up 
through the change in the Trend. The variable TrPostWC
i, t
, the variable of interest, captures 
the differences between the host municipalities and non host municipalities and their change 
in the slope post- versus pre-FIFA World Cup 2010 time period.
Hagn and Maennig (2009) extend the difference in difference approaches of Hotchkiss et al. 
(2003) by combining both approaches into one model. This way they take the changes as well 
as for the trends of the dependent variable into account. Hereby they avoid distorted results, as 
the lower growth after the world cup can be caused by an already existing negative trend. 
 
The variables PostWC
i, t
 and TrPostWC
i, t
examine the potential unemployment effects of the 
FIFA World Cup in the match venues.
The period of observation begins in 2003, which is the year before the announcement that 
South Africa was selected to host the FIFA World Cup 2010 and ends in 2011, the latest 
period with data available. The data is received from IHS Global Insight South Africa. Global 

---

16 
Insight is a consultancy that specializes in economic analyses and forecasting with data from 
Standard & Poor’s Data Resources International and WEFA.
The data being used looks at the 75 largest municipalities in South Africa. The selection is 
made by taking the largest 75 municipalities in terms of their number of inhabitants in 2003.
Host cities for the FIFA World Cup 2010 include Johannesburg, Cape Town, Durban, 
Pretoria, Port Elizabeth, Polokwane, Nelspruit, Bloemfontein and Rustenburg. Their location 
is shown in Figure 1.