Can employment be positively affected by the hosting of the summer Olympics? : a case study of the summer Olympics of London 2012

(1)

Can employment be positively affected by the hosting of the summer Olympics? A case study of the summer Olympics of London 2012

Bachelor thesis University of Amsterdam Economics & Business 19-01-2018

Supervisor: MSc C.W. Haasnoot

Paul Frank Anton Versteegh 5768543

Rode Kruislaan 1408D 1111 XD Diemen

(2)

Abstract

The Olympic Games are the biggest sport event in the world but what does this mean for employment. Can employment be positively affected by the hosting of the summer Olympic Games? This research will be a case study of the Olympics of London 2012. The results of this research can be useful for future host cities and governments but also for its citizens. Recently lots of potential bids and bids for hosting the Olympics are being withdrawn because of referenda. This mostly has to do with the fact that the costs are high and the economic benefits of the Olympics are not clear. This research can contribute to make this clearer. Or at least to what effect is has on employment in London.

In this research quarterly data on all 33 boroughs of London with respect to employment, population, qualifications and wages across both genders within a time period of 2004 till 2016 was used. Two groups were created to identify an Olympic effect on employment. One group of Olympic boroughs consisting of the six boroughs where the Olympics are hosted and one group containing the non-Olympic boroughs. This Olympic effect was tested using a difference-in-difference approach in three different time periods called the three Olympic phases. The results show significant positive effects on employment rates in all three phases in the Olympic boroughs. There are some differences between male and female employment rate effects though, especially in the pre-Olympic phase. A side step in the research shows interesting results for the horeca sector. The results of this research contribute to the literature on effects of hosting the Olympics on employment. The results are not generalizable but the method can be used to identify employment effects in other Olympic host cities.

(4)

Introduction

The Olympic Games come with high costs and the economic benefits are not so clear. This makes it a controversial topic in every country where the Olympics can or will be hosted. It raises the question: why do cities/governments want to host the Olympic Games? One of the possible benefits could be an increase in employment. In this research I will focus on the 2012 summer Olympics of London and I will study if the hosting of the Olympics positively

affected employment.

Hosting the Games is a controversial topic because in Hamburg (2024), St. Moritz (2026) and Innsbruck (2026) citizens voted against a bid in a referendum and potential referenda in Boston (2024) and Budapest (2024) made city councils decide to withdraw the bid. One of the reasons to vote against a bid is the high cost of hosting the event. These costs can be

illustrated by figure 1. The numbers are only the sport-related costs and already show that the costs are huge.

Figure 1: Time series of cost for Olympics 1960-2016

Source: World Economic Forum (2018)

One part of this controversy could be cleared by showing what hosting the event could imply for employment. More specific, increasing the rate of employment in London’s East End, which consists of six boroughs. Figure 2 shows us in which six boroughs the Olympics are hosted. The focus on employment and this part of London was triggered by the mission

(5)

statement of the London Olympic Committee: “The Games will have failed if it does not leave a legacy of growth, employment and expansion and lead to the development of the City’s East End and surrounding neighbourhoods” (Raco and Tunney, 2009, p.2086). This raises a couple of questions: was there an increase of employment? Did employment increase before, during or after the Games? In which part of London was employment generated?

Figure 2. London divided in Olympic and non-Olympic boroughs

Source: International Olympic Committee (2017)

As will be described in chapter 3 these six boroughs are not on the same economic indicators level as the other boroughs of London. The goal of the Olympic committee was to raise them to the same level and they used the Olympics as a tool to achieve this. Using the Games in this way makes it an interesting phenomenon because it is not just the biggest sports event anymore. It could also be a tool for achieving an employment policy objective.

(6)

In this research will be verified if the committee managed to increase employment in the Olympic boroughs compared to the non-Olympic boroughs levels using a Difference-in-Difference approach. This means comparing the change of employment rates in boroughs with an Olympic venue to boroughs without an Olympic venue and to boroughs similar to venue areas also within London. This approach is used in research studies of large sport events like the Olympics and the World Cup soccer. Some relevant examples of this approach can be found in research of the impact of the 1996 summer Olympic Games on

unemployment and wages in Georgia (Hotchkiss et al., 2003) and in research on employment

effects of the summer Olympics Games of 1972 in Munich (Jasmand and Maennig, 2009,

p.999). Results of research with similar approaches showed that there were no significant differences between unemployment rates in German venue cities and in in non-venue cities as a result of the 2006 World Cup soccer for example (Hagn and Maennig, 2009, p.3300). Data on unemployment from 1951 to 2008 used by Tien showed that hosting the summer Olympics has a significant effect on unemployment in the first, third and fourth year before the

Olympics (Tien et al., 2011, p.19).

Most of these papers are focused on more than just employment. This paper however will only focus on effects on employment due to the Olympics in London. The mission statement plays a vital role in the difference between London and other cities that hosted the Olympics. The focus on employment and the clear goal that they want to achieve sets them apart. The Olympics of Barcelona are the most comparable in a way that the event was used to reduce unemployment in Barcelona/Catalonia. The difference however is that London picked six boroughs to host the Olympics because they want to leave a legacy of growth, employment and expansion and lead to the development of the City’s East End and surrounding

neighbourhoods” (Raco and Tunney, 2009, p.2086). This paper will check if they managed to achieve one part of this goal: employment.

For this research accessible data on labour market indicators for all boroughs of London could be found through the Office for National Statistics. Their website provides detailed

information about different kinds of economic variables for all boroughs in London over the period 2004-2016 (ONS, 2017).

(7)

By combining the Olympics and the important issue of employment this research on the effect of hosting the Olympics on employment can be relevant. The result is relevant for National Olympic Committees, the International Olympic Committee (IOC), governments and citizens of (potential) hosting cities. They can all see what employment effects there are in different time periods before, during and after the Olympic Games. This could influence national Olympic Committees and governments their decision whether to be a candidate city or not. It can also influence the decision of citizens to vote for or against a bid to host the Olympics in their city.

The study can also contribute to the literature because of the extensive data set and the implications it can result in. Most studies were not capable to identify effects on a borough level because data was not available and not as extensive. This research will be because of the extensive data set. The organizers of the London Olympics set a clear goal which can be checked for.

(8)

Literature review

Three phases of the Olympic Games

In most study’s authors use three different time periods to structure their study. These three different phases are known as the pre-Olympic phase, the during-Olympic phase and the post-Olympic phase.

The pre-Olympic phase is defined by most studies as the period from the election to one year before the Olympics. The election is when the International Olympic Committee (IOC) chooses which candidate city will be hosting the Olympic Games. This election takes place 7 years before the actual Olympics. In this period there will be an increase in visitor arrivals due to the city’s increased profile (Blake, 2005, p.21). We could expect a larger effect on the construction sector in particular. The pre-Olympic phase is associated with large investments in infrastructure and building of venues also known as the construction phase.

The during-Olympic phase consists of the year of the Olympics and some studies choose a few years surrounding the Olympic year. This period should at least include the Paralympic Games that occur prior or after the Games (Blake, 2005, p.21). The horeca sector (hotel, restaurant and cafe) will be the most affected during this phase. The direct revenues and costs of the Games are part of this phase. Direct revenues are from ticket sales, sponsorships, visitors spending, television contracts and merchandise. Costs in this phase are construction costs, security, hiring of staff on events and public transportation.

The post-Olympic phase consists of the years after the Olympics. Researchers choose their own preferred time frame. Some studies choose periods up to 6 years after the event took place. Most researchers choose shorter periods. Data for London was available up till 2016 so in this research this period will consist of 11 quarters. The effect that the Games have in the years after the Games is called the ‘legacy effect’. This effect includes a higher profile of the city and increased visitor arrivals to the city because of this profile (Blake, 2005, p.21). In addition, the infrastructure developed for the Games will provide value for many years after the Games (Blake, 2005, p.21). So the legacy effect consists of two parts; legacy visitor impacts and legacy infrastructural impacts. The tourism sector will profit the most from the legacy effect.

(9)

Research on previous Olympics

In this part research with regard to employment on Olympic Games and the Soccer World Cup will be summarized. This will only be research on Summer Olympic Games because of a difference in magnitude between summer and Winter Olympics. The number of sports hosted at the Winter Olympics is less. The Sochi Olympics of 2014 had seven sports featuring 15 disciplines resulting in 98 medal events (IOC, 2018). The London Olympics of 2012 had 26 sports featuring 39 disciplines resulting in 302 medal events (IOC, 2018). In Sochi there were contests in 11 venues compared to 34 venues in London. Winter Olympics are usually hosted at places where most of the facilities are already available. In Vancouver 2010 eight out of the 14 venues were pre built/already existed and only some modifications had to be done to meet IOC standards (Olympic World Library, 2009). This means that in the pre-Olympic phase less construction is needed and therefore less employment is generated. Because there is less construction needed and there are less medal events the costs of Winter Olympics are lower as can be seen in figure 1. Because there are less medal events and venues this also means less employment at the venues during the Olympics.

In the research of Brunet (1995) on the 1992 Olympic Games in Barcelona is stated that there was a 72% increase in employment in the construction sector between 1986 and 1992

(Brunet, 1995, p.20). Barcelona used the Olympics to go from a depression to an economic boom. Till 1985 the unemployment rates were rising and after the election of Barcelona as host of the Olympics in 1985 they started falling in Barcelona, surrounding area and in all of Catalonia. The unemployment rate was 18,4% in 1986 but dropped to 9,6% in 1992 during the Olympics. The numbers are even more significant if one considers that the number of active employable people in this period grew by 1,1% (Brunet, 1995, p.21). This indicates 66.889 employed and at least 88,7% of it was due to the impact of the organization of the Olympic Games of 1992 (Brunet, 1995, p.24). The permanent effect of the Olympic Games in terms of employment was calculated to be 20.000 people (Brunet, 1993).

The hosting of the Olympics in Atlanta, Georgia, in 1996 had a significant positive effect on employment in venue areas. Hotchkiss et al. stated 17% more employment during and after the Olympic Games than in non-venue areas (Hotchkiss et al., 2003, p.692). The researchers found three effects on the labor market. First, a direct short-term effect on employment due to spending by the Atlanta Committee for the Olympic Games on goods and services. Second, job training of the 70.000 volunteers and their experience obtained should have impacted their

(10)

employment opportunities. Third, investment in facilities and infrastructure and migration are expected to affect employment and wages positively (Hotchkiss et al., 2003, p.691).

Humphreys and Plummer (1995) used an input-output RIMS II model, which is an updated version of the model that the Economic Research Associates used for their analysis of the Los Angeles Games. They found that the Games generated an additional 77.026 jobs in Georgia and 37.000 of these jobs were in Atlanta (Humphreys, 1995). The Games generated

cumulative job growth in 1994-1996 of 42.448 full- and part-time jobs in Atlanta as an upper bound (Baade, 2002, p.141). This is in line with the findings of Humphreys and Plummer. However their findings also state that at least 40% of the generated jobs were transitory. In the best case scenario the Games generated 24.742 permanent full- or part-time jobs. The other two scenarios indicate a loss of 29.301 and 4540 long term jobs (Baade, 2002, p.142).

Table 1: Research on previous Olympics summarized

Source: Baade and Matheson (2002), Economic Research Associate (1984), Kim et al. (1989), KPMG (1993), NSW Treasury (1997), Madden (2002), Blafousia- Savva et al. (2001) and Papanikos (1999) Los Angeles 1984 Los Angeles 1984 Seoul 1988 Sydney 2000 Sydney 2000 Sydney 2000 Athens 2004 Athens 2004 New jobs 5.043 73.375 336.000 156.198 98.700 90.000 300.400 445.000 Area Los Angeles Southern California South Korea

Australia Australia Australia Greece Greece

Time period 1984 1984 1982-1988 1991-2004 1994-2006 1994-2006 2000- 2010 1998-2011 Author Baade and Matheson (2002) Economic Research Associate (1984) Kim et al. (1989) KPMG (1993) NSW Treasury (1997) Madden (2002) Blafousia- Savva et al. (2001) Papanikos (1999)

(11)

Table 1 shows research on the Olympics and their prediction or stated a number of jobs created in a certain period but were less capable of stating an exact/more precise effect on employment in different time periods by the Olympics.

Soccer World Cups are more comparable to summer Olympics in this regard. At least with regard to costs. Although a World Cup only hosts one sport and most venues are usually already there, the magnitude of the event is comparable. The fact that World Cups are hosted in multiple cities means that employment is generated in multiple cities instead of just one.

The World Cup of 2006 in Germany showed different results in terms of unemployment. In 2001 and 2005 the venue cities had significantly higher unemployment than the non-venue cities (Hagn and Maennig, 2009, p 3297). Unemployment was not lower in venue cities when compared to non-venue cities after the World Cup but after the World Cup but there was significantly higher unemployment in Germany. This higher unemployment was unrelated to the World Cup but more related to the financial crisis.

The goals and expectations of London for hosting the Olympics

The organizing committee of the London Olympics stated that the Games would have failed if it does not leave a legacy of growth, employment and expansion and lead to the development of the City’s East End and surrounding neighborhoods (Raco and Tunney, 2009, p.2086). As stated above employment is one of the goals of the organizing committee and especially employment in the city’s East End. Expectations were formed about the amount of jobs that are being generated by the hosting of the Olympics. Adam Blake wrote a paper in 2005 about the economic impact of the London 2012 Olympics and predicted that, for instance, the construction sector would gain an additional 14.354 full time equivalent (FTE) jobs in the pre-Olympic phase, consisting of the years 2005-2011. The city’s East End would benefit the most of this gain due to the larger share of construction impacts and the industrial

composition of employment. Meaning that in East London employment is more weighted towards the construction industry than other London sub-regions (Blake, 2005, p.66).

The hotel, restaurant and bars were also expected to benefit from the hosting of the Olympics and were expected to generate new jobs in the pre-Olympic phase. However restaurants and evening activities are expected to suffer from an expenditure diversion effect. This means that foreign visitors might not visit London because the Olympics led to higher prices which made London a more expensive and hence a less attractive place to stay (Blake, 2005, p.31). This

(12)

effect is expected to be 10% in London and 2% in the United Kingdom during the Olympics. Research on the Olympics of Los Angeles and Atlanta shows that the restaurants had less business during the Olympics than normally. This is due to the fact that Olympic visitors tend to watch Olympic events on television when they are not attending an event instead of

engaging in evening activities (Blake, 2005, p.22). Because of this interesting and surprising result, this research will try to show if this happened as well in London. In the results section a part will be devoted on this interesting topic.

In the pre-games phase East London was expected to have a 33% share of the total new jobs generated mainly because of the composition of employments towards construction. During the games this share of new jobs would only be 10% due to the composition of employment being less heavily weighted towards services (Blake, 2005, p.66). During and post games central and west London were expected to perform better than other parts of London because of higher proportions in service industries (Blake, 200, p.67).

The downside of the expanding sectors is that there are sectors which do not directly benefit from construction activities or from the legacy. Employees move to other jobs due to higher wages in the expanding sectors. For instance, the manufacturing sector was expected to decline equivalent to 18.923 FTE jobs (Blake, 2005, p.44). The higher wages in expanding sectors will lead to a real wage increase leading to a higher labour supply. This additional labour is supplied freely by people who would, at the original wage, prefer not to work

(Blake, 2005, p.60). The short duration of the Olympic Games does not necessarily lead to the hiring of new employees, the generation of permanent full-time jobs and the sustainability of the employment effects. Entrepreneurs will probably exhaust other alternatives such as asking existing employees to work overtime or perform other tasks, before hiring additional work force to satisfy the temporary high demand (Crompton, 1995). Still Blake expected an overall positive effect on the United Kingdom and London economies due to an increase in GDP and 8.164 extra FTE jobs in the United Kingdom.

(13)

Methods

Data on population, economic activity/inactivity, unemployment, willingness to work all divided in a male and female section of all the 33 boroughs from the last quarter of 2004 until the second quarter of 2016 will be used in this research. All the data in this study refer to the population aged 16-64. This is the case for all the data in this study. These variables can be used to check for effects on employment.

Wages are also included in the study but are not quarterly recorded but yearly. This will be represented by the median annual gross pay. Wage will be corrected for by the consumer price index to represent a real wage. According to the literature real wage affects willingness to work and employment.

Six different levels of qualifications are accounted for. They are stated as National Vocational Qualifications (NVQ). NVQ4+ is the highest category in this case. Unfortunately

qualification data are not separated by gender. Data are also not quarterly but yearly. In the regression equation a new variable is created based on the six groups.

NVQHIGH=(NVQ2+/NVQ population 16-64)*100. This new variable refers to the part of the population that is qualified at least NVQ2, in percentages. Regressions using NVQ2+ or NVQ3+ in the equation and everything else kept the same led to more significant results in the case of NVQ2+.

A dummy to identify whether or not a borough is an Olympic borough is included. Six boroughs are identified as Olympic, 27 are not. This dummy shows us whether or not there is a significant difference in the employment rate in the initial state, being the situation before London was elected to host the Olympics.

Three other dummies for different time periods are included: dPRE, dDUR and dPOST. These dummy variables identify the pre-, during- and post-Olympic phase. They will show us whether there is a different effect on employment in each time period. In the during-Olympic phase both the Olympics and Paralympics are hosted which is a requirement according to the literature (Blake, 2005, p.21). Table 2 shows us the distribution of the time periods.

(14)

Table 2: First and last quarter of the three Olympic phases.

Source: Literature review

Interacting the Olympic dummy with the three different time periods: dPREOlympic, dDUROlympic and dPOSTOlympic results in whether or not being an Olympic borough affected the employment rate in each time period. This will identify if the Olympics affected employment and if the Olympic Committee achieved its employment goal.

A crisis dummy is included in the regression. There is a lag in the effect on employment due to the financial crisis. The first effects of the crisis could be seen in the third quarter of 2008, but it took a second quarterly fall for them to be lower on the year (ONS, 2009). The fact that this research includes quarterly data recorded as q3=October-September and q4=January-December resulted in defining the quarters that include the crisis to be: 2009q1 till 2010q4

The industry mix consisting of 9 industries is included as well. Data are quarterly but not separated by gender. The data reflect the percentages of employment distributed across industries of the Standard Occupational Classification (SOC2010). The data can be used to control for borough-level characteristics that might otherwise confound differences measured in borough employment. This data will also be used to check whether or not the horeca sector is affected by the Olympics.

Effects on ethnic minorities will not be specified in the study. This is because the data are incomplete, lacking in gender differences and above all cannot be used as a time series. Estimates on ethnic minorities cannot be used due to changes in the ethnicity questions on the annual Population Survey in 2011 and should therefore not be used as a time series (ONS, 2015). Starts Ends Initial situation 2004q4 2005q2 dPRE 2005q3 2011q4 dDUR 2012q1 2013q2 dPOST 2013q3 2016q2

(15)

This research will not include fixed effects regression. The data is pooled and therefore fixed effects are not expected but random effects. Running the regression with fixed effects leads to the variable dOlympic being omitted because of this. Therefore fixed effects will not be included in this research.

Excluding the borough City of London

The data of the borough City of London was incomplete and statistically unreliable. This mostly had to do with a small number of people (±5.000) in the age of 16-64 leading to unreliable estimates and confidence intervals (Nomis, 2016). In comparison the second smallest borough with respect to people in the age of 16-64 is Kingston upon Thames with just over 100.000.

In figure 3 the red dots indicate employment rates for the City of London. As can be seen this borough is a-typical and responsible for all the outliers in the regression. Excluding this borough leads to more reliable results and a better representation of the labour market indicators between Olympic and non-Olympic boroughs. This supports the decision to exclude this borough from the regressions and comparisons.

Figure 3. Employment rates over quarters. Borough City of London in red, other boroughs in blue

(16)

Labour market indicators before election date

In this part there will be a summary of the state of the labour market across the boroughs before the election of London as the host city, the initial situation. To be more specific: the Olympic boroughs will be compared to the non-Olympic boroughs. The Olympic boroughs are the following six: Barking & Dagenham, Greenwich, Hackney, Newham, Tower Hamlets and Waltham Forest. The remaining 26, after excluding the City of London, boroughs of London will be referred to as the non-Olympic boroughs.

The election of London as the host city was on 6th of July 2005. In the data the first three time periods took place before this election date (2004q4, 2005q1, 2005q2). Data on periods before 2004q4 was not available. A possible effect of this election could not have taken place in these three periods or would be minimal. These periods will therefore be the initial situation and the starting point of the research.

First step was to summarize the data of the Olympic boroughs to create a group of Olympic boroughs and summarize the data of the Olympic boroughs to create a group of non-Olympic boroughs. By summarizing the data averages can be calculated on different labour market indicators. Both groups can be compared to assess the initial state of the labour market across these boroughs.

(17)

Summary of labour market indicators Table 3: Summary of labour market indicators

General population Male population Female Population

Non- Olympic Boroughs Olympic Boroughs Difference in percentage points Non- Olympic Boroughs Olympic Boroughs Difference in percentage points Non- Olympic Boroughs Olympic Boroughs Difference in percentage points Economic activity rate - aged 16-64 74.9 65.5 9.5 82.7 76.0 6.7 67.3 54.9 12.4 Employment rate - aged 16-64 69.9 58.8 11.2 77.2 67.6 9.5 62.8 49.8 13.0 Unemployment rate - aged 16-64 6.7 10.2 -3.5 6.7 10.9 -4.3 6.8 9.3 -2.5 % who are economically inactive - aged 16-64 25.0 34.5 -9.5 17.3 24.0 -6.7 32.7 45.1 -12.5 % of economically inactive who want a job 24.1 19.7 4.4 26.1 22.3 3.8 23.1 18.3 4.8

Data source: Office for National Statistics (2017)

This shows us that there are quite some differences between the two groups of boroughs. The initial situation based on labour market indicators shows that the Olympic boroughs are underperforming. This could indicate why the Olympic Committee stated in their mission statement to leave a legacy of growth, employment and expansion and lead to the

development of surrounding neighbourhoods” (Raco and Tunney, 2010, p.2086).

Females in the Olympic boroughs show a lower economic activity rate and employment rate than females in non-Olympic boroughs. The economic activity rate and employment rate for males in the Olympic boroughs are also lower than in non-Olympic boroughs. The differences for males are smaller between Olympic and non-Olympic than their female counterparts. Indicating that females are even more underperforming based on these labour market indicators than males in the Olympic boroughs compared to the non-Olympic boroughs.

Figures 4 and 5 give a clear visual insight in the differences between the boroughs based on a labour market indicator. The Olympic boroughs are in yellow.

(18)

Figure 4. Employment rate in the initial situation of age 16-64

Data source: Office for National Statistics, 2017

Figure 5. Willingness to work if economically inactive in the initial situation of age 16-64

(19)

Difference by qualifications

Qualifications in England, Wales and Northern Ireland are work based awards achieved through assessment and training. They are expressed in National Vocational Qualifications (NVQ). There are five qualification levels NVQ1 to NVQ5, a group for not having a NVQ and a group with other qualifications. NVQ is not formally the same as conventional academic qualifications but it is used as an equivalent. NVQ2 is an equivalent for a general certificate of secondary education graded A*-C in 4-5 fields (City&Quilds, 2018). Usually enough to enter higher education levels including universities. In this study NVQ will be used as a control variable representing qualification/education. In the data from the Office for National Statistics (ONS) 6 groups appear: no qualifications, other qualifications, NVQ1+, NVQ2+, NVQ3+ and NVQ4+.

Table 4: Differences in National Vocational Qualifications (NVQ) before election date

Source: Office for National Statistics, 2017

The indicators for NVQ1+, NVQ2+, NVQ3+ and NVQ4+ for Olympic boroughs show that each group is at least 10% points lower than in the non-Olympic boroughs. The group with other qualifications than NVQ is a 0,5% point larger and the group with no qualifications is 9,8% point larger than in the non-Olympic boroughs. Because the percentages for NVQ+ are cumulative this means that each group is about the same size in percentages except for

NVQ4+ and no qualifications. In the non-Olympic boroughs NVQ4+ is 10% points larger and this 10% corresponds with no qualifications being 9,8% point lower. This shows us that in the Olympic boroughs the work-based qualifications of persons in the age 16-64 are on average lower than in non-Olympic boroughs.

Non- Olympic Boroughs Olympic Boroughs Difference in percentage points % with NVQ4+ - aged 16-64 33.6 23.6 10.0 % with NVQ3+ - aged 16-64 47.4 35.1 12.3 % with NVQ2+ - aged 16-64 61.1 48.8 12.3 % with NVQ1+ - aged 16-64 71.6 61.3 10.3

% with other qualifications (NVQ) - aged 16-64 15.7 16.2 -0.5 % with no qualifications (NVQ) - aged 16-64 12.7 22.5 -9.8

(20)

Wage

Another important labour market indicator is wage. Unfortunately wage is not quarterly reported as the other indicators mentioned above but annually (ONS, 2016). The earnings information presented refers to gross pay before tax, National Insurance or other deductions, and excludes payments in kind. The results refer to earnings to the survey pay period and payments of arrears from another period made during the survey period; any payments due as a result of a pay settlement but not yet paid at the time of the survey is excluded (Nomis, 2016). Instead of taking an average of the three quarterly periods before the election date this figure will show the initial situation as of 2004. This is because 2005 might be affected by the election of the Games. The statistics of Kensington & Chelsea are unobserved for males and therefore do not produce a reliable estimate. This borough will be excluded from this

comparison.

Figure 6. London by borough median wage 2004 annually in pounds. Excluded the boroughs City of London and Kensington & Chelsea. Olympic boroughs in yellow.

(21)

As can be seen in figure 6 above the annual gross pay of the Olympic boroughs is relatively low compared to the other non-Olympic boroughs. Three Olympic boroughs are in the lowest category, two in the second lowest and one in the middle category. The initial situation of the Olympic boroughs citizens based on median annual wage is that they earn less than the non-Olympic boroughs counterparts.

Financial crisis

The financial crisis started on 9th of August 2007. This crisis is expected to have a significant effect on the labour market indicators in all boroughs and is therefore a relevant phenomenon for this research. In this part will be showed how this crisis affected the Olympic and Non-Olympic boroughs. Figure 7 shows that the unemployment rates all went up when the crisis hit the economy. The lines of the unemployment rates show quite a similar movement

although the effect of the crisis seems to kick in a little bit later in the Olympic boroughs. The lines show that it is very likely that the financial crisis affected the labour market in London. The trend was that unemployment rates were declining and all of a sudden they are going up by 3% points in ±2 years. A dummy variable will be included in the regression to identify this effect.

Figure 7. Unemployment rates for Olympic, Non-Olympic boroughs and London as a whole

Source: Office for National Statistics (2017) 5.0 6.0 7.0 8.0 9.0 10.0 11.0 12.0 13.0 Ja n 2 00 4-De c 2 00 4 Ju l 2 00 4-Ju n 2 00 5 Ja n 2 00 5-De c 2 00 5 Ju l 2 00 5-Ju n 2 00 6 Ja n 2 00 6-De c 2 00 6 Ju l 2 00 6-Ju n 2 00 7 Ja n 2 00 7-De c 2 00 7 Ju l 2 00 7-Ju n 2 00 8 Ja n 2 00 8-De c 2 00 8 Ju l 2 00 8-Ju n 2 00 9 Ja n 2 00 9-De c 2 00 9 Ju l 2 00 9-Ju n 2 01 0 Ja n 2 01 0-De c 2 01 0 Ju l 2 01 0-Ju n 2 01 1 Ja n 2 01 1-De c 2 01 1 Ju l 2 01 1-Ju n 2 01 2 Ja n 2 01 2-De c 2 01 2 Ju l 2 01 2-Ju n 2 01 3 Ja n 2 01 3-De c 2 01 3 Ju l 2 01 3-Ju n 2 01 4 Ja n 2 01 4-De c 2 01 4 Ju l 2 01 4-Ju n 2 01 5 Ja n 2 01 5-De c 2 01 5 Ju l 2 01 5-Ju n 2 01 6 Unemployment rate Olympic Boroughs Unemployment rate Non-Olympic Boroughs Unemployment rate London

(22)

Results

In this section the regression results will be presented. This will start with the results for the general population aged 16-64. After this part there will be a results section for males, females and a section for the effects on the horeca sector. The regression results for all sections can be found in table 5. The number of boroughs differs between the regressions. This is because the borough Kensington & Chelsea has missing data for the median wage for males which leads to excluding the borough from regression two and four. The appendix shows a robustness check for the regression.

Results general population aged 16-64

Regression equation 1: EMPrateit=α+β1NVQHIGHi+β2dcrisist+ β3dOlympici + β4dPREt + β5dDURt+ β6dPOSTt+ β7dPREOlympict+ β8dDUROlympict+ β9dPOSTOlympict +εit

Where eit is the employment rate (employment/Populationaged1664) in borough i at quarter t. NVQHIGHit is a control variable indicating a qualification level of borough i at quarter t. dcrisis is a dummy variable for whether or not quarter t is part of the financial crisis. dOlympic is a dummy variable for whether or not borough i is an Olympic borough. dPRE, dDUR and dPOST are dummy variables for if quarter t is in the pre, during or post phase of the Olympics. dPREOlympic, dDUROlympic and dPOSTOlympic are interaction variables to identify the effect of the Olympics on the Olympic boroughs in the three different Olympic phases.

The constant shows us the employment rate for non-Olympic boroughs without any other variable added. The parameter is significantly different from zero and indicates a 62,73% employment rate in the non-Olympic boroughs. NVQHIGH is significantly different from zero at the α=0,05 level. This shows us that a higher fraction of the population with higher qualifications leads to a higher employment rate. This always is a small coefficient because this fraction can only take a value between 0-1.

The dummy variable dcrisis is significantly different from zero at the α=0,05 level. Implying for dcrisis that during the crisis time periods the employment rate was significantly lower. The dummy variable dOlympic is significant at the α=0,001 level. Implying a significant

(23)

difference in employment rate between Olympic and non-Olympic boroughs. The fact that dOlympic is negative means that the Olympic boroughs are on a lower initial employment rate. 9,28% point lower which is in line with the findings stated in the initial state of the labour market indicators.

The dummy variables for the three Olympic phases: dPRE, dDUR and dPOST show insignificant results for both dPRE and dPOST. Implying that these two periods do not

significantly differ from zero across all boroughs. dDUR is significantly different from zero at the α=0,01 level. Implying that the employment rate across all boroughs in the

during-Olympic phase is significantly lower at 2,46% point.

The relevant variables dPREOlympic (α=0,01 level), dDUROlympic (α=0,001 level) and dPOSTOlympic (α=0,05 level) are all significant. Therefore the employment rates in the Olympic boroughs in the Pre-, during- and post- phase of the Olympics are significantly different from zero. In the pre-Olympic phase 3,31% point higher employment rate, during-Olympic phase 4,92% point higher and post-during-Olympics it is 4,41% point higher. All three of them are also positive implying that the employment rate is positively affected by the hosting of the Olympics in all three phases in the Olympic boroughs.

Logistic regression equation 5: EMPrateit=α+β1NVQHIGHi+β2dcrisist+ β3dOlympici + β4dPREt + β5dDURt+ β6dPOSTt+ β7dPREOlympict+ β8dDUROlympict+ β9dPOSTOlympict +εit

This logistic regression using the same variables as regression equation 1 shows some different results. The relevant variables dPREOlympic, dDUROlympic and dPOSTOlympic are not significant anymore at the α=0,05 level. dDUROlympic and dPOSTOlympic are significant at the α=0,10 level though. This shows us that the results for the relevant variables using a logistic regression are not as significant as in regression 1. The dummy variable dcrisis is not significant anymore. Implying that the financial crisis had no significant effect on employment rates for the general population. The control variable NVQHIGH is

significant at the α=0,001 level. The logistic regression shows less significant results but does not differ that much from regression 1 in terms of implications.

(24)

Table 5: Regression results t statistics in parentheses *p < 0.10, ** p < 0.05, *** p < 0.01, **** p < 0.001, (1) (2) (3) (4) (5) Employment rate Employment rate males Employment rate females Elementary Occupations Employment rate logit NVQHIGH 0.00119** -0.000798**** 0.00818**** (2.34) (-4.05) (5.52) dcrisis -0.00847** -0.00910* -0.00389 0.00150 -0.0449 (-2.09) (-1.87) (-0.61) (0.56) (-1.54) dOlympic -0.0928**** -0.0881**** -0.120**** 0.0221*** -0.370**** (-4.57) (-4.59) (-3.82) (3.16) (-4.19) dPRE -0.00885 -0.00801 0.00169 0.00462 -0.0534 (-1.36) (-0.99) (0.20) (1.26) (-1.15) dDUR -0.0246*** -0.0105 -0.00180 0.0133** -0.156*** (-2.67) (-0.97) (-0.17) (2.18) (-2.79) dPOST 0.00558 0.0230** 0.0278** 0.0118** -0.0118 (0.60) (2.37) (2.41) (2.32) (-0.23) dPREOlympic 0.0331*** 0.0397*** 0.0254 0.0105 0.135 (2.86) (2.95) (1.09) (0.84) (1.49) dDUROlympic 0.0492**** 0.0678**** 0.0382* 0.00111 0.198* (3.76) (3.64) (1.72) (0.06) (1.87) dPOSTOlympic 0.0441** 0.0430* 0.0578** 0.00697 0.160* (2.35) (1.94) (2.02) (0.39) (1.68) RWImale 0.000625* (1.92) RWIfemale -0.0000942 (-0.20) RWI -0.00116**** (-4.79) _cons 0.627**** 0.712**** 0.635**** 0.236**** 0.346**** (18.86) (21.65) (15.99) (11.43) (3.36) N Boroughs R2 Pseudo R2 384 32 0.3825 350 31 0.2948 360 32 0.3118 338 31 0.5960 384 32 0.0041

(25)

Results male population aged 16-64

Regression equation 2: EMPmaleit=α+β1RWImaleit+β2dcrisist+ β3dOlympici + β4dPREt + β5dDURt+ β6dPOSTt+ β7dPREOlympict+ β8dDUROlympict+ β9dPOSTOlympict +εit

Where EMPmaleit is the male employment rate (male employment/male

Populationaged16-64) in borough i at quarter t. RWImaleit is a control variable indicating a real wage index for males in borough i at quarter t. dcrisis is a dummy variable for whether or not quarter t is part of the financial crisis. dOlympic is a dummy variable for whether or not borough i is an Olympic borough. dPRE, dDUR and dPOST are dummy variables for if quarter t is in the pre, during or post phase of the Olympics. dPREOlympic, dDUROlympic and dPOSTOlympic are interaction variables to identify the effect of the Olympics on the Olympic boroughs in the three different Olympic phases.

RWImale is significantly different from zero at the α=0,10 level. This shows us that an increase in the real wages for males positively affects the male employment rate.

The dummy variable dcrisis is significantly different from zero at the α=0,10 level. Implying that during the financial crisis the employment rate for males was significantly lower with 0,9% point.

dOlympic is significantly different from zero at the α=0,001 level. Implying that the Olympic boroughs are on a lower initial employment rate for males. 8,81% point lower which is in line with the findings stated in the initial state of the labour market indicators.

The dummy variables for the three Olympic phases: dPRE, dDUR and dPOST show insignificant results for both dPRE and dDUR. Implying that these two time periods do not significantly differ from zero in all boroughs. dPOST is significantly different from zero at the α=0,05 level. Implying that the male employment rate across all boroughs in the post-Olympic phase is significantly higher at 2,3% point.

The relevant variables dPREOlympic, dDUROlympic and dPOSTOlympic are all significant different from zero. dPREOlympic is significant at the α=0,01 level. dDUROlympic is significant at the α=0,001 level. dPOSTOlympic is significant at the α=0,10 level Therefore the levels of employment in the Olympic boroughs in the Pre-, during- and post- phase of the

(26)

Olympics are significantly different from zero. All three of them are also positive implying that the male employment rate in the Olympic boroughs is positively affected by the hosting of the Olympics in all three Olympic phases.

Results female population aged 16-64

Regression equation 3: EMPfemaleit=α+β1RWIfemaleit+β2dcrisist+ β3dOlympici + β4dPREt + β5dDURt+ β6dPOSTt+ β7dPREOlympict+ β8dDUROlympict+ β9dPOSTOlympict +εit

Where EMPfemaleit is the female employment rate (female employment/female

Populationaged16-64) in borough i at quarter t. RWIfemaleit is a control variable indicating a real wage index for females in borough i at quarter t. dcrisis is a dummy variable for whether or not quarter t is part of the financial crisis. dOlympic is a dummy variable for whether or not borough i is an Olympic borough. dPRE, dDUR and dPOST are dummy variables for if quarter t is in the pre, during or post phase of the Olympics. dPREOlympic, dDUROlympic and dPOSTOlympic are interaction variables to identify the effect of the Olympics on the Olympic boroughs in the three different Olympic phases. In this regression the borough City of London is excluded.

RWIfemale is not significantly different from zero. This shows us that an increase in the real wages for females does not significantly affect the female employment rate. The dummy variable dcrisis is not significantly different from zero. Implying that during the financial crisis the employment rate for females was not significantly different.

dOlympic is significantly different from zero at the α=0,001 level. Implying that the Olympic boroughs are on a lower initial employment rate for females. 12,0% point lower which is in line with the findings stated in the initial state of the labour market indicators.

The dummy variables for the three Olympic phases: dPRE, dDUR and dPOST show insignificant results for both dPRE and dDUR. Implying that these two time periods do not significantly differ from zero across all boroughs. dPOST is significantly different from zero at the α=0,05 level. Implying that the female employment rate across all boroughs in the post-Olympic phase is significantly higher at 2,78% point.

(27)

There are different results for the relevant variables dPREOlympic, dDUROlympic and dPOSTOlympic. dPREOlympic is not significantly different from zero. This implies that the female employment rate for females in the Olympic boroughs is not significantly different from the females in the non-Olympic boroughs. dDUROlympic is significantly different from zero at the α=0,10 level. Implying that the employment rate for females in the

during-Olympic phase in the during-Olympic boroughs is significantly more positively affected than in the non-Olympic boroughs at 3,82% point. dPOSTOlympic is significantly different from zero at the α=0,05 level. Implying that the employment rate for females in the post-Olympic phase in the Olympic boroughs is significantly more positively affected than in the non-Olympic boroughs at 5,78% point. The difference between males and females can be seen in the pre-Olympic phase. An pre-Olympic phase in which employment is mainly influenced by

construction. One could assume that the construction sector is mainly employed by males. So an insignificant result for females in this phase could possibly be explained by this

assumption.

Results horeca sector aged 16-64

In this section the results of the regression 4 will be stated. Regression equation 4 is based on interesting findings in the research on the Olympics of Los Angeles and Atlanta indicating an expenditure diversion effect. The regression will check for this effect by regression on the employment in the horeca sector. A downside to this regression is that the data for this sector shows not just the horeca sector but elementary occupations. This group best represents the workforce in the horeca sector although there are many more occupations that are part of this group. Jobs that are part of the horeca sector and related to the elementary occupations based on the SOC (2010) are bar supervisor, barmaid, barperson, bartender, glass collector, waiter, waitress, hotel assistant and porter for instance (UK & Visas, 2015).

Regression equation 4: ElemOccuEMPit=α+β1RWIit+ β2NVQHIGH+β3dcrisist+ β4dOlympici + β5dPREt + β6dDURt+ β7dPOSTt+ β8dPREOlympict+ β9dDUROlympict+β10dPOSTOlympict +εit

Where ElemOccuEMPit is the rate of the total employment that is employed in the SOC (2010) group Elementary Occupations in borough i at quarter t. RWIit is a control variable indicating a real wage index in borough i at quarter t. dcrisis is a dummy variable for whether or not quarter t is part of the financial crisis. dOlympic is a dummy variable for whether or not

(28)

borough i is an Olympic borough. dPRE, dDUR and dPOST are dummy variables for if quarter t is in the pre, during or post phase of the Olympics. dPREOlympic, dDUROlympic and dPOSTOlympic are interaction variables to identify the effect of the Olympics on the Olympic boroughs in the three different Olympic phases. In this regression the borough City of London is excluded.

RWI is significantly different from zero at the α=0,001 level. This shows us that an increase in the real wages significantly negatively affects the rate that is employed in the elementary occupations. This is in line with the expectation because elementary occupations are low paid occupations. The same expectations can be formed for NVQHIGH. A higher rate of higher qualifications leads to a lower rate that is employed in an elementary occupation. The regression confirms this expectation and is significant at the α=0,001 level.

The dummy variable dcrisis is not significantly different from zero. Implying that during the financial crisis the rate employed in this sector was not significantly different. dOlympic is significantly different from zero at the α=0,01 level. This implies that the Olympic boroughs are on a higher initial rate that is employed in this sector. 2,2% points higher.

The dummy variables for the three Olympic phases: dPRE, dDUR and dPOST show an insignificant result for dPRE. Implying that the employment rates in this period do not significantly differ from across all boroughs. dDUR and dPOST are significantly different from zero at the α=0,05 level. Implying that the rate of employed people in this sector in all the boroughs in the during- and post-Olympic phase is significantly higher. This meets the expectations that the horeca sector will profit during and after the Olympics and not before the Olympics. The increased profile of the city will lead to more tourism and therefore to more consumption in the horeca sector.

The results for the relevant variables dPREOlympic, dDUROlympic and dPOSTOlympic are not significantly different from zero. This means that the rate that is employed in this sector in the Olympic boroughs is not significantly different from the non-Olympic boroughs in all three time periods. This could mean that visitors do not spend more of their evening activities in the Olympic boroughs than in the non-Olympic boroughs.

(29)

Conclusion and discussion

This research demonstrated that the 2012 Olympics of London has positively affected employment rates in the Olympic boroughs in all three Olympic phases. Employment rates went significantly up compared to the non-Olympic boroughs. This result means that the Olympic Committee of London achieved one part of its goal. “The Games will have failed if it does not leave a legacy of growth, employment and expansion and lead to the development of the City’s East End and surrounding neighbourhoods” (Raco and Tunney, 2009, p.2086). The Committee achieved the employment goal, implying that hosting the Olympics can be used as a tool to achieve economic policy.

A similar regression identifying the different effects per gender shows some different results. For males there is a significant positive effect on the employment rate in all three Olympic phases in the Olympic boroughs compared to the non-Olympic boroughs. For females this is not significantly the case in the Olympic phase. This could have to do with the pre-Olympic phase being weighted towards construction. In the during- and post-pre-Olympic phase the effects are positive just like for their male counterparts. Interesting result is also that the financial crisis did not seem to significantly affect female employment rate nor did the indicator for real female wage.

A side step in the research to identify an effect on a particular sector in the labour market namely the horeca sector led to some interesting results as well. Assumed in most literature is that the horeca sector will benefit the most in the during- and post-Olympic phase. Blake (2005) already stated that an expenditure diversion effect could take place instead of the expected benefits. The regression implied that there was no significant difference in the percentage of employed people in this sector in the Olympic boroughs during all three phases compared to the non-Olympic boroughs. This could mean that Olympic visitors tend to spend their evening activities throughout London and not just in the Olympic boroughs. The

percentage of employed people across all boroughs was significantly higher in both the during- and post-Olympic phase. This could imply that there is a legacy effect on tourism because more employment is needed in this sector compared to before the Olympics. This effect is not just for the Olympic boroughs but for London in general. An important note on this regression is that it includes all elementary occupations and does not only refer to the horeca related occupations.

(30)

Discussion

Due to constraints things could have been done better. More variables could be added to get a more precise representation of the situation. Variables like growth in GDP per capita,

qualifications per gender or ethnic minorities could all add to a more precise representation. This could potentially lead to a better fit of the model and increase the R2.

Data is only available from 2004-present and updated every quarter. Repeating this research in a few years could give a better representation of the post-Olympic phase. This could maybe even identify when the Olympic effect ends. Data on periods before 2004 would be very useful as well to identify the period before pre-Olympic better. A longer time period in general results in more observations and a better representation of the effect over time on employment rates.

An interesting regression to run would be to check for effects on employment in the

construction sector. Especially in the pre-Olympic phase an effect can be expected due to the construction of the venues and infrastructural projects. This regression could very well be the explanation for the differences between males and females in the pre-Olympic phase.

The other goals that the Olympic Committee set are not researched in this study. These are also very interesting goals and questions to review. Due to time constraints these goals were not reviewed. This could be an interesting topic for future research and would shine a brighter light on the effects of hosting the Olympics on more than just employment.

This method could also be applied to other Olympics to check if they have similar results. One of the problems could be the availability of the necessary data in other Olympic cities. For London this data is very extensive, meaning that it is on borough level and quarterly. For other cities this could be a potential problem.

(31)

Bibliography

Baade RA, Matheson V. (2002) Bidding for the Olympics: fool’s gold? In Transatlantic Sport: 442 E. Kasimati the Comparative Economics of North America and European Sports, Barros CP, Ibrahimo M, Szymanski S (eds). Edward Elgar: London; 127–151.

Balfousia-Savva S, Athanassiou L, Zaragas L, Milonas A. (2001) The Economic Effects of the

Athens Olympic Games. Centre of Planning and Economic Research: Athens. (In Greek.)

Blake A. (2005) The Economic Impact of the London 2012 Olympics, Nottingham: Nottingham University Business School.

Brunet F. (1993) Economy of the 1992 Barcelona Olympic Games, Lausanne: International Olympic Committee.

Brunet F. (1995) An economic analysis of the Barcelona '92 Olympic Games: Resources, financing, and impact, The Keys to success: the social, sporting, economic and

communications impact of Barcelona '92(1995)

City&Quilds

https://www.cityandguilds.com/qualifications-and-apprenticeships/qualifications-explained/qualification-comparisons (25-01-2018)

Coates, D, Humphreys, B,R. (2003) The effect of professional sports on earnings and employment in the services and retail sectors in US cities, Regional Science and urban

economics, 33(2003) 175-198

Crompton JL. (1995) Economic impact analysis of sports facilities and events: eleven sources of misapplication. Journal of Sport Management 9(1): 14–35

Hagn, F, Maennig, W. (2009) Large sport events and unemployment: the case of the 2006 soccer World cup in Germany, Applied Economics, 41: 3295-3302

Hotchkiss J, Moore R.E, Zobay S.M. (2003) Impact of the 1996 Summer Olympic Games on

Employment and Wages in Georgia, Southern Economic Journal, 69: 691-704

Humphreys JM, Plummer MK. (1995) The Economic Impact on the State of Georgia of

Hosting the 1996 Olympic Games. Selig Center for Economic Growth: Georgia.

International Olympic Committee (IOC) https://www.olympic.org/sochi-2014 (04-01-2018)

International Olympic Committee (IOC) https://www.olympic.org/london-2012 (04-01-2018)

Jasmand, S, Maennig, W. (2009) Regional Income and Employment Effects of the 1972 Munich Summer Olympic Games, Regional studies, 42: 7:991-1002

Jeux olympiques d’hiver. Comité d’organisation (2010) Official Games report VANOC

(32)

Kasamati, E (2003) Economic Aspects and the Summer Olympics: a Review of Related Research, Int. J. Tourism Res. 5, 433–444

Kim JG, Rhee SW, Yu JC, et al. (1989) Impact of The Seoul Olympic Games on National

Development. Korea Development Institute: Seoul.

KPMG Peat Marwick. (1993) Sydney Olympics 2000:Economic Impact Study. Sydney Olympics 2000 Bid Ltd: Sydney.

Madden, JR (2002) The Economic Consequences of the Sydney

Olympics: The CREA/Arthur Andersen Study, Current Issues in Tourism, 5:1, 7-21

Nomis https://www.nomisweb.co.uk/ (03-09-2016)

NSW Treasury. (1997) Research and Information Paper: the Economic Impact of The Sydney

Olympic Games. New South Wales Treasury and CREA: Centre for Regional Economic

Analysis (University of Tasmania).

Office for National Statistics (2015) http://www.ons.gov.uk (02-04-2015)

ONS (2009) The impact of the regression on the labour market. Newport: Office for National Statistics

Papanikos GT. (1999). Tourism Impact of the 2004Olympic Games. Tourism Research Institute: Athens. (In Greek.)

Rose, A.K, Spiegel. (2011) The Olympic effect, The Economic Journal, 121 (June): 652–677

The World Bank (2015) http://data.worldbank.org/country/china (02-04-2015)

Tien C, Lo H.C, Lin H.W. (2011) The economic benefits of mega events: A myth or a reality? A longitudinal study on Olympic games, Journal of sports management, 25: 11-23

UK Visas & Immigration (2015) Codes of practice for skilled workers. Glasgow: UK visas & Immigration

World Economic Forum

(33)

Appendix

Table 6: Robustness check

(1) (2) (3)

Employment rate Equation 1

Employment rate PRE reduced by the

first 4 quarters

Employment rate DUR extended by two

quarters NVQHIGH 0.00119* 0.00117* 0.00110* (2.34) (2.37) (2.16) dcrisis -0.00847* -0.00805* -0.00829* (-2.09) (-2.00) (-2.04) dOlympic -0.0928*** -0.0853*** -0.0939*** (-4.57) (-4.22) (-4.57) dPRE -0.00885 -0.00776 -0.00845 (-1.36) (-1.82) (-1.29) dDUR -0.0246** -0.0223** -0.0159 (-2.67) (-3.06) (-1.73) dPOST 0.00558 0.00784 0.0149 (0.60) (1.05) (1.49) dPREOlympic 0.0331** 0.0284* 0.0333** (2.86) (2.10) (2.86) dDUROlympic 0.0492*** 0.0416*** 0.0442** (3.76) (3.50) (2.88) dPOSTOlympic 0.0441* 0.0365* 0.0474* (2.35) (2.14) (2.37) _cons 0.627*** 0.627*** 0.633*** (18.86) (19.60) (19.06) N 384 384 384 t statistics in parentheses *_{p < 0.05,}**_{p < 0.01,}***_{p < 0.001}

Table 6 shows us a robustness check. Three regressions have been done using different time periods to identify the pre-, during- and post-Olympic phases. The first column of results shows us the results of equation 1, which is used in this research. In the second regression the dummy for the pre-Olympic phase has been reduced by one year, four quarters, so instead of starting in 2005q3 this phase starts in 2006q3 and still ends in 2011q4. This compared to the

(34)

first equation. In the third regression the during-Olympic phase is extended by two quarters compared to the first equation. This means that this phase ends in 2013q4 instead of 2013q2. It still starts in 2012q1. It also means that the post-Olympic phase is reduced by two periods and starts in 2014q1 instead of 2013q3.

The results of the regressions show very similar results. Almost all variables have roughly the same coefficients and significance levels. The only variable that is not significant compared to equation 1 is the during-Olympic across all boroughs variable in regression 3. This makes sense because of the literature showing us that this phase usually is one year. In this regression this period is two years. The effect on the post-Olympic phase shows similar results for all boroughs and for the Olympic boroughs. This shows that the regression is robust.