http://peeg.wordpress.com

Papers in Evolutionary Economic Geography

# 20.26

Does Relatedness drive the Diversification of Countries’ Success in Sports? Louis Knuepling & Tom Broekel

Does Relatedness drive the Diversification

of Countries’ Success in Sports?1

Louis Knuepling, Institute of Economic and Cultural Geography, Leibniz University

Hannover, Germany. E-Mail: knuepling@wigeo.uni-hannover.de

Tom Broekel, Business School, Stavanger University, Norway. E-Mail: t.brokel@uis.no

Abstract Research Question: The sources of countries’ success in sporting events have been

investigated for a long time. In addition to socioeconomic factors, sport specialization and

competitive balance have been identified as impacting medal yields. Nevertheless, there

remains little understanding of how countries become successful in sports they did not succeed

in before.

Research Methods: We argue that the concept of related diversification, which is particularly

popular in Evolutionary Economic Geography, can also be applied to sports and can help to

explain why certain sporting countries become successful. We substantiate our argument with

an empirical study of the evolution of national medal portfolios covering 61 sports at the

Olympic Summer Games between 1896 and 2016 using this framework.

Results and Findings: Our empirical findings quantify the relatedness of sports emerging from

physical, financial, cultural, and organizational similarities. We confirm that diversification

processes based on countries’ existing strengths in particular sports are a (small) driving force

behind the Olympic success of countries and help to explain their success in sports in which

they have been previously unsuccessful.

Implications: Our results highlight how examining hidden opportunities for diversification by

incorporating sport-relatedness in a competitive positioning strategy can be extremely

informative for sport policy makers.

Keywords: Relatedness; Olympic Games; diversification; sport policy; elite sport

JEL classification: L83, Z20, O57,

1 This is a pre-print version of the article published by Taylor & Francis Group in European Sport Management Quarterly, available online: http://www.tandfonline.com/10.1080/16184742.2020.1770830.

Introduction

Most of the scientific studies analyzing nations’ success at sport events such as the Olympic

Games concentrate on just four factors: population size, GDP (per capita), political regime

(communist/socialist or not), and host advantage. There are limited studies that consider

additional factors such as comparative advantage/specialization (Du Bois & Heyndels, 2007;

Seiler, 2013; Tcha & Perchin, 2003) or competitive balance (Truyens, De Bosscher, & Heyndels,

2016). Moreover, there are hardly any studies that differentiate between sports (Forrest, McHale,

Sanz, & Tena, 2016). This contrasts with the influence of prioritizing particular sports by policy

on countries’ subsequent success in these sports (De Bosscher, Shibli, & Weber, 2019).

Similarly, non-scientific articles of the ‘Hamburger Abendblatt’ (Leoni, 2012) and ‘The

Guardian’ (Gibson, 2016) attribute the extraordinary success of the British team at the Olympic

Games in London 2012 and Rio de Janeiro 2016 to a change in national sport policy toward

increased funds, national competition, and a focus of support on sports with a competitive

balance and podium potential for the next Games. For example, this strategy was adopted by

the Dutch Olympic Committee, which calls for ‘smart’ investments into a few selected sports

(NOC*NSF, 2016).

This raises questions about how to select the right sports to focus on. In this context,

competitive balance and the number of potential medals a discipline offers are general criteria,

but do not allow for country-specific support policies. Relying on sports that yielded medals in

the past implies the stabilization of the status quo. It does not help to identify the potential to

broaden the medal portfolio and thereby possibly spread and consequently reduce risks.

This paper proposes the concept of identifying sports that are currently un- or

underdeveloped at the national level by combining insights from the literature on regional

economic diversification (e.g. Boschma, Balland, & Kogler, 2015) and sports management.

Therefore, we present the concept of related diversification, which emerged from evolutionary

economic geography, and is based on the theory that economic entities are more likely to

become successful in activities that are related to their current activities in terms of knowledge

and other resources. In essence, this theory reflects the resource-side of the competitive

positioning strategy approach, which has recently been used to investigate the prioritization in

elite sport funding (Weber, De Bosscher, & Kempf, 2019a). Consequently, we argue that a

‘translated version’ of this theory is applicable to the development of countries’ success in elite

sports. We support this argument by means of an empirical study using 61 sports at the Olympic

Games between 1896 and 2016. The results confirm that countries are more likely to become

successful in sports that are related to their current strengths.

The theoretical basis of the related diversification approach and its transfer to the context

of Olympic sports is the subject of the next section. Subsequently, we construct a novel measure

of sport relatedness, followed by the results of the empirical analysis. The findings are then

discussed and exemplified by the application of our framework to the case of Great Britain.

Related Diversification in Sports Success in Sports

Basic socioeconomic factors such as population size and GDP are widely accepted as two of

the best predictors for the number of medals a country wins at the Olympic Games (Berdahl,

Uhlmann, & Bai, 2015; Grimes, Kelly, & Rubin, 1974; Johnson & Ali, 2000, 2004; Lozano,

Villa, Guerrero, & Cortés, 2002). Assuming equally distributed talent worldwide, more people

equals more potential sports talent. Financial resources are crucial for exploiting this talent pool,

although both factors (population and financial investments) tend to be subject to diminishing

returns to scale (Bernard & Busse, 2004; De Bosscher, Heyndels, De Knop, van Bottenburg, &

Shibli, 2008; Hoffmann, Chew, & Ramasamy, 2002). During the Cold War, communist

countries outperformed market economies due to their centralized approach to elite sport

(Andreff, W. Andreff, & Poupaux, 2008; Grimes et al., 1974; Hoffmann et al., 2002). The

existence of a positive effect of hosting the Olympics is also confirmed in a number of studies

(Andreff et al., 2008; Bernard & Busse, 2004; Lui & Suen, 2008; Moosa & Smith, 2004). Being

a neighboring country (with the benefit of low transportation costs and adaption to the climate)

(Johnson & Ali, 2000) and being the upcoming host (Bredtmann, Crede, & Otten, 2016;

Maennig & Wellbrock, 2008; Nevill, Balmer, & Winter, 2009) are also associated with better

Olympic results.

Crucially, some factors influencing countries’ success in sports are specific to disciplines.

For instance, Forrest et al. (2016) find that the effect of GDP per capita differs largely across

sports. This is most likely caused by the differences between sports in terms of their resource-

intensity. Richer countries are more successful in a broader range of sports and in more

resource-intensive ones, whereas poorer countries have to specialize (Du Bois & Heyndels,

2007; Tcha & Pershin, 2003). Rathke & Woitek (2008) show, that relative to its socioeconomic

endowments, Mexico produces an inefficient number of medals in all sports, whereas Mongolia

is efficient by specializing in weightlifting and wrestling. Success in sports is also persistent

over time (Bernard & Busse, 2004; Hunt & Morgan, 1995; Lui & Suen, 2008). On the other

side, the number of medals available at Olympic Games and other world-level championships

remains rather stable, whereas increasing numbers of countries are developing (specialized)

capabilities at the elite sport level (De Bosscher et al., 2019; Truyens et al., 2016). To avoid

declining medal shares induced by increasing competition, countries extend their investments

in elite sport (De Bosscher, De Knop, van Bottenburg, Shibli, & Bingham, 2009; Green, 2004;

Houlihan & Zheng, 2013). In this context, specialization and prioritization of funding is a

prominent and successful strategy (De Bosscher et al., 2019; Green, 2004; Houlihan & Zheng,

2013): Countries tend to devote funds to sports in which they have been successful in the past

to maintain and increase their comparative advantages, whereas other countries refrain from

investing in sports dominated by other countries (De Bosscher et al., 2019).

In a recent study, Weber, De Bosscher, Shibli, and Kempf (2019b) embed the market

potential for sports in the theory of the competitive positioning of firms proposed by Hooley,

Piercy, Nicoulaud, and Rudd (2017). This includes the market side (marked-based view, MBV)

and the resource side (resource-based view, RBV). Regarding the former, competitive balance

indicates the height of the entry barrier to a sport (Truyens et al., 2016). It can be driven by the

number of medals available, but also by the situational character of the sport, such as the method

of result assessment (judgement vs. measurement) (Weber et al., 2019b). Meanwhile, the

resource side refers to the endowments of a country, such as geography (water surface,

mountains), culture (tradition of judo in Japan), and economic resources (capital-intensity of

sports) (De Bosscher et al., 2019). Based on interviews, Weber et al. (2019a) find that National

Sport Associations indeed devote funds according to both the market and the resource side.

Relatedness in Economics and Economic Geography

In Economics and Economic Geography, researchers have studied firms’, regions’, and nations’

technological and economic diversification patterns over a long time. The relatedness approach

has received considerable attention in this literature (Balland, Boschma, Crespo, & Rigby, 2019;

Boschma, et al. 2015; Kogler, Rigby, & Tucker, 2013). Based on the idea of path-dependent

economic development, researchers argue that economic entities do not diversify in a random

manner but rather do so conditional on their existing set of activities. That is, they are more

likely to try and more likely to be successful in entering2 a new field of activities when this is

2 In this paper, the notion of ‘entry’ always describes the development of success/capabilities in a specific field

related to their existing strengths and competitive advantages. Relatedness thereby implies the

new field being somewhat similar (but not identical) in terms of the competences, resources,

and capabilities required to be successful in it. For instance, a region that has previously had

success in the production of bicycles and coaches is more likely to diversify (and be successful)

in car manufacturing than a region specialized in the manufacturing of textiles (Boschma &

Wenting, 2007). This is because its labor force already possesses crucial skills (mechanical

engineering), and the configuration of its infrastructure is also closer to the needs of the new

industry. Moreover, innovators and entrepreneurs find the new knowledge easier to learn and

to develop further, as it corresponds more with their knowledge basis.

The relevance of relatedness has been confirmed in numerous studies explaining the

economic diversification of firms, regions, and nations (Boschma et al. 2015; He, Yan, & Rigby,

2016; Lo Turco & Maggioni, 2016; Neffke, Henning, & Boschma, 2011; Rigby, 2015) and has

recently been applied to the music industry (Klement & Strambach, 2019).

In the following, we put forward different causes of relatedness between sports and

develop how this concept of related diversification can be applied to the context of sports, and

hence how it may be used to gain a new perspective and better understand in which sports

countries become successful at the elite level.

Sport-Relatedness

Physical relatedness is most obvious at the level of events within the same kind of sport. For

instance, a 100m sprinter is very likely to run the 200m at a similar competitive level. The same

holds for swimmers who are often competitive at multiple distances. In part, this is a result of

the similar physical requirements for a successful execution. Apart from strength, stamina

(running, swimming), tactics (different ball games), and eye-hand-coordination (table tennis,

archery) are examples of aspects that could cause physical relatedness between sports.

This suggests the idea of modularity of sports and their relatedness through similar

components. For instance, a world-class pole vaulter needs speed on the runway and a strong

take-off but also highly developed gymnastic skills. The sport consists of different components

(running, jumping, gymnastics) that can be trained individually. Although the combination is

specific to sports, the individual components are not, which is why many sports comprise at

(industry, sport, etc.) in which the entity (country, region) did not succeed previously. ‘Diversification’ is nearly synonymous in that respect. Diversification into a specific new activity can be called an ‘entry.’

least some components of other sports. Individual athletes might be able to succeed in different

sports that are related by their components, though this rationale makes much more sense on

the regional and national level. There are instructions to many sports on the internet and elite

athletes share training drills via social media, but learning requires a form of ‘trial-and-error’

(Witt, Broekel, & Brenner, 2007). Direct supervision by an expert is important especially when

it comes to perception and sensation during the execution. To reproduce successful athletes,

countries therefore draw upon former elite athletes as top-level coaches (Rynne, 2014).

Furthermore, one reason for the establishment of national training centers is to enable athletes

and coaches of different sports to join in training and learn specific components (Lee & Price,

2016). A pole-vaulter lacking bounce or gymnastic skills might improve through advice from a

professional long jump or gymnastics coach. On the other hand, he might learn only a little

from rowers or swimmers.

Another type of relatedness may arise from similar institutional frameworks and overlap

in infrastructure or capital requirements. Sports can have the same methods of result assessment

(measuring, scoring, judging) or jointly use facilities, such as a gymnasium (handball,

basketball) or an aquatics center (swimming, diving, water polo) (Zheng & Chen, 2016). Other

sports are technology-based and expensive, such as sailing, rowing, golf, cycling, equestrian,

and fencing, which all require high-end material, whereas running, boxing, or weightlifting do

not (see De Bosscher et al., 2019).

Sports may also be related in a cultural and historical sense. The modern pentathlon, for

example, comprises events from equestrian, fencing, shooting, swimming, and running which

all developed from, or were at least frequently performed in a military context (Heck, 2014). A

long history of armed forces, such the former colonial powers have, might therefore induce

success in modern pentathlon and its components. As the athletics events were among the first

belonging to the Olympic program, countries with a long history of participation at the Games

might have established organizational structures and competencies in athletics events in general,

whereas more recent members of the Olympic community (Kenya or Jamaica) occupy niches3

(endurance, sprint) that are, in turn, highly based on physical endowments.

These physical and infrastructural requirements, institutional settings, and historical

developments have been identified as important internal resources of countries necessary for

success in elite sport (Weber et al., 2019a). Countries are endowed with more or less favorable

3 74 out of 77 Jamaican Olympic medals were awarded in sprinting. 89 out of 102 Kenyan Olympic medals were awarded in endurance (800m to marathon) or 4x400m relay (Olympic.org).

conditions depending on how closely their resources match the ones required to become

successful in a specific sport. Furthermore, we assume that the current comparative strengths

of a country (their most successful sports) reflect their endowments with these resources.

Consequently, it will be easier for countries to develop successful athletes in sports that are

related to those in which they are already successful. In this case, relatedness refers to the

overlap in resource requirements.

Imagine country J, which is unsuccessful in diving but performs relatively well in

swimming and trampoline or artistic gymnastics. Hence, it will easily find coaches with

expertise in gymnastics, and swimming pools exist that can be adapted for diving. Its population

is also likely to provide favorable physical characteristics, as the country is successful in

gymnastics. Clearly, country J has a promising foundation to advance in diving. The situation

is opposite for a country S that is only successful in weightlifting. Its coaches, infrastructure,

and general expertise are of almost no use to improve in diving. Accordingly, it will be harder

and hence more unlikely that S will become successful in diving than country J.

The following chapters present how an empirical measure of relatedness between sports

can be calculated and whether relatedness has been a driver of the diversification process

(market entries) of countries at the Olympic Games between 1896 and 2016.

Method Dataset

To measure countries’ success in sports, we rely on a dataset of all Summer Olympic medalists

from 1896 to 2008 retrieved from The Guardian (2016). The medals of the Olympic Games of

London 2012 and Rio de Janeiro 2016 are added manually from the website Olympic.org (2018).

Winter Olympics are not part of the dataset. We leave this to future research.

During Olympic history, some sports completely disappeared from the program, others

were newly introduced, and some disciplines and events experienced minor changes (e.g.

differences in weight categories by 1kg). Moreover, the number of medals distributed per event

is restricted (gold, silver, one or two bronze medals4). To ensure that sports can be traced over

time and the number of medals per sport is sufficiently large, the individual Olympic events are

4 Bronze medals are awarded to both semi-finalists in some sports with knock-out system, like Judo and Taekwondo.

aggregated to 61 sports. It is more straightforward for ball games (e.g. football, handball,

basketball), because they are very heterogeneous and cannot be aggregated. Lacking an

objective criterion, all events in which the same athlete could in principle be able to win an

Olympic medal are aggregated (e.g. 100m breaststroke and 200m breaststroke to breaststroke).

All weight categories in each combat sport are aggregated to one sport, respectively (taekwondo,

judo, etc.)5.

To further increase the comparability of sports over time, all sports that were part of less

than four Olympic Games (e.g. rugby seven’s, golf) or were staged only in the founding period

of the Olympic Games (e.g. tug of war) are excluded. As our manual classification is subjective,

the empirical results are substantiated with another, broader, aggregation of events into 42 sports

(e.g. the swimming styles, fencing styles, and equestrian disciplines are further aggregated, see

Appendix 1). On this basis, the final dataset comprises 13,917 medals won by athletes from 135

countries in 61 sports at the Olympic Summer Games between 1896 and 2016.

Mapping Sport-Relatedness

The basic assumption is that a country’s success in a sport reflects how its physical, knowledge,

infrastructure, and cultural conditions correspond to the conditions necessary for success in this

sport. We approximate success with the presence of an Olympic medal. We assume that sport

training systems are organized in a country-specific fashion, with interactions between sports

being more likely to take place within countries than between countries.

Under these assumptions, a co-occurrence approach in the style of Breschi, Lissoni, and

Malerba (2003) and Hidalgo, Klinger, Barabási, and Hausmann (2007) can be used to construct

a measure of sport-relatedness. In contrast to these studies, which rely on patent and country

export data, we apply the approach to the co-occurrence of Olympic medals at the country level.

Specifically, the relatedness between each pair of sports 𝒊 and 𝒋 at time 𝒕 is based on their co-

occurrences (𝒄𝒊𝒋 ), i.e., the number of countries (𝒏 ) that won medals both in 𝒊 and 𝒋 at the

respective Olympic Games. 𝒎 represents that a country ‘has won at least one medal’ in that

sport.

5 See Appendix 1 for a documentation of the aggregation.

𝑐 , = 𝑛∈ , , ,

(1)

For example, if Canada, France, and Russia all won medals in both rowing and sailing

at the Olympic Games in Sydney, then 𝒄𝒓𝒐𝒘𝒊𝒏𝒈, 𝒔𝒂𝒊𝒍𝒊𝒏𝒈, 𝟐𝟎𝟎𝟎 = 𝟑 (three co-occurrences). As the 61

sports contains a different numbers of events, the probability index, a version of the association

strength (see for a discussion van Eck & Waltman, 2009), is applied to correct the co-

occurrences for the ‘size’, i.e., number of potential medals in the sports:

𝜑 , =(𝑐 ,𝑇 )

(𝑠 ,𝑇 ∗

𝑠 ,𝑇 − 𝑠 , +

𝑠 ,𝑇 ∗

𝑠 ,𝑇 − 𝑠 , ) ∗ 𝑇

2

(2)

𝑆 = ∑ 𝑐 , ; 𝑆 = ∑ 𝑐 , (3)

𝑇 = ∑ 𝑐 ,, (4)

Put simply, the numerator indicates the share the co-occurrences of 𝑖 and 𝑗 have in the

sample 𝑇 of all co-occurrences of sports in portfolios of countries at time t. It is divided by the

probability of these sports to co-occur by chance, which is calculated with a random draw

(denominator). The resulting values of relatedness 𝜑 range from 0 to infinite, with 𝜑 ≥ 1

indicating deviation from statistical independence/randomness (van Eck & Waltman, 2009).

Each pair of sports receives one value indicating their degree of relatedness.

Figure 1 shows the correlation of the relatedness values between each two Olympic

Games (across all sports present at both of these Games). High values indicate a stronger time

stability of the relatedness measure. However, the correlations are rather weak, even between

two consecutive Olympics. This is a clear drawback of the approach: medals are only a rough

approximation of success because medal decisions are highly situational. They depend on

weather conditions, refereeing decisions, and many other internal (current fitness level), and

external (e.g. spread of a disease in the Olympic village) effects. Therefore, we treat all medals

equal instead of weighing gold, silver and bronze6. To further smoothen the rather unreliable

data of a single Olympics, the relatedness between each pair of sports is adjusted by averaging

the relatedness values from the first Olympics (1896) to the respective year (e.g. the mean of

all Games from 1896 to 1968 for the Games in Mexico City 1968). The consequential increase

6 De Bosscher et al. (2008) and Lui and Suen (2008) show that weighted and unweighted medal counts are highly correlated.

in stability over time is shown in the last column of Figure 1, which indicates the correlation of

the cumulatively calculated relatedness of the respective year to the cumulative relatedness

values of 2016.

Figure 1. Correlation table of the pairwise relatedness values between sports over time. Source: Own calculation and visualization, data: The Guardian (2016), Olympic.org (2018). Notes: The last column shows correlations of the cumulatively calculated relatedness of 2016 with all other cumulatively calculated years.

Figure 2 illustrates the resulting ‘sports space’ of 2016, a network visualization of the

relatedness between sports. Each sport is represented by a node (circle)7. For better visibility,

7 The non-random connections (φ>1) are inputted to the force-directed drawing algorithm (Fruchterman-Reingold), i.e., more similar sports are plotted more closely to each other. For all visualizations we use the statistical software R (package ‘igraph’ for network visualization [Csardi & Nepusz, 2006]).

only the strongest 35%8 of links between sports are displayed. One of the first impressions is

that sports belonging to the same group of sports (colors) tend to be co-located (athletics,

aquatics, combat sports, cycling), but the visual inspection also aligns with a number of our

theoretical arguments. The different swimming styles are closely related, while synchronized

swimming and (synchronized) diving are closer to trampoline. In fact, the direct link between

diving and trampoline is among the strongest (Table 1). This might be explained by the more

technical components in diving compared with swimming. More generally, sports that require

high financial investment, such as equestrian events (eventing, jumping, dressage), sailing,

rowing, and cycling are co-located in the bottom-left of the network. The capital intensity might

also explain the strong links between triathlon and jumping as well as hockey and eventing

(Table 1).

The right part of the network comprises many sports that require technical skills and

accuracy (e.g. table tennis, archery, fencing, gymnastics). In addition to the obvious physical

similarities (e.g. between the fencing styles), there are interesting component-related

connections; for example, the links between archery and badminton as well as between table

tennis and pistol (shooting), which all require excellent eye–hand coordination.

The strong relationships between different types of ball games (e.g. football and beach

volleyball) are likely rooted in the culture (high affinity of the population to ball games in

general), although component relatedness (eye–hand(foot) coordination, knowledge about

tactics) also applies there (e.g. handball and water polo). At second glance, some less obvious

but nonetheless insightful connections emerge. For instance, cross cycling and canoe/kayak

slalom (Table 1) are both forms of ‘outdoor’ sports and require balance, endurance, and

adaptability to different surface conditions.

Moreover, it becomes clear that no single resource requirement exclusively explains a

connection between two sports. Triathlon, for example, is located between its components

endurance, freestyle (swimming), and cycling. This can be interpreted as a physical or as

component relatedness. However, it might also be the capital intensity that explains its location

in the network. Clearly, this deserves further research in the future.

8 This value has been chosen for purely aesthetic reasons.

Figure 2. The 'sports space' (2016). Source: Own calculation and visualization, following Hidalgo et al. (2007): ‘product space’ and Boschma et al. (2015): ‘technology space’, data: The Guardian (2016), Olympic.org (2018). Notes: Each node (circle) represents one sport. Sports are positioned based on their non-random relations with other sports (cumulative relatedness 2016). The force-directed algorithm draws more strongly related sports closer to each other

Table 1. Strongest links in the ‚sports space’.

Source: own calculation, data: The Guardian (2016), Olympic.org (2018). Notes: selected links, to show a variety of different strongly related sports.

Empirical Specification

The co-occurrence structure of sports relates well visually to the components of sport-

relatedness identified in the theory section. With an entry model, we test if and to what degree

it actually explains the diversification of countries into new sports (see as comparison Boschma

et al., 2015, for the entry of technologies to the patent portfolio of US cities). In this model,

observations are country–sport–year combinations (e.g. ‘Australia – rowing – 2004’) and the

dependent variable is binary. A positive value (1) equals an ‘entry’ of a country to a sport, i.e.,

at these Games the country wins a medal in this sport for the first time. The dataset is limited

to those cases for which an entry is actually possible. Hence, all sports that are not part of the

respective Games, all countries that decide to not take part in the respective Games (e.g. mutual

boycotts of the USA and the Soviet Union in 1980 and 1984, respectively), and countries that

do not exist anymore (e.g. Soviet Union after 1992) are excluded. In addition, all sports in which

a country has already won a medal in the past, i.e., in which it has already successfully

‘diversified,’ are excluded for that country. The last point is a very strong restriction, as, for

example, a medal in sprinting would be a great success for Germany which won its last medal

in this sport in 1996.

The main explanatory variable indicates how closely a potentially ‘entered’ sport is

related to the current strengths of the country. We term this ‘proximity’ (𝑃𝑟𝑜𝑥), as it should

reflect the degree to which the resource endowments of a country correspond to those

potentially required for the respective sport. In practice, we calculate Equation 5. Here, the

proximity of country 𝑛 to sport 𝑖 at time 𝑡 is defined as the sum (across all other sports 𝑗) of the

products between the number of medals (NOM) country n has won in the 𝑗 sport at the

respective Games (𝑡) and the relatedness between sport 𝑖 and the 𝑗 sport in 𝑡. Thereby, we

only consider sports 𝑗 that belong to the comparative strengths s of the country, i.e., 𝑗 occupies

a larger share in the country’s medal tally than it does in all medals awarded at these Games.

𝑃𝑟𝑜𝑥 , , = ∑ (𝑁𝑂𝑀 , ,∈ ∗ 𝜑 , ) (5)

Accordingly, proximity increases when: firstly, the respective sport is related to many

sports in which the country had success before; secondly, the value of relatedness is higher; and

thirdly, the country has won several medals in the related sport. The largest value of proximity

is observed for the USA to backstroke for the 1908 Games (127.23). It is driven by the large

number of medals the USA has won in related sports at the previous Games. For an easier

interpretation and to reduce the strong effect of high medal counts in some sports, the variable

is transformed into the ‘relatedness density’ as implemented by Boschma et al. (2015). We

divide the sum of the relatedness values of all sports j that are both related to sport i and belong

to the comparative strengths s of the country, by the sum of the relatedness values of all sports

that are related to 𝑖 (Equation 6). Thus, we capture the percentage of relatedness to sport 𝑖

already present in the country’s portfolio. Thereby, the magnitude of success (number of medals)

in each related sport is reduced to whether its share in the regional medal portfolio exceeds the

share across all countries. Nevertheless, proximity and relatedness density are highly correlated

(0.85).

𝑅𝐷 , , = ∑ ,∈ ,

∑ ,∗ 100 (6)

To avoid endogeneity, both relatedness density and proximity are included with a time-

lag of one period. Consequently, we cannot calculate our explanatory variable if the sport was

in the previous Olympic program and if the country chose not to participate at the previous

Games. Neither can we calculate a proximity value before a country has won its first medal at

the Olympic Games. Excluding these cases reduces our dataset from an initial 222,345 cases

(27 Olympics x 61 sports x 135 countries) to 61,768.

Control variables

Proximity and relatedness density partly depend on size. The larger the set of sports a country

is specialized in, the higher the likelihood that some related sports are among these just by

chance. Therefore, we control for the ‘portfolio size’ (𝑃𝑆) (number of sports in which a country

is specialized) at the previous Games. The sport management literature has shown that in

addition to financial, cultural, and geographic factors, ‘competitive balance’ (CB) is a main

reason for the prioritization of funding to specific sports (De Bosscher et al., 2019), and

therefore can be incorporated as a possible factor behind diversification. Our measure of

competitive balance (Equation 7) is the percentage of all medal-winning countries (at the three

most recent Games) that won a medal in this sport (at the three most recent Games) (see Truyens

et al., 2016, PMW-measure9). To isolate the effect of competition from the size effect of sports,

we additionally control for ‘medals available per sport’ (NOM).

9 Contrary to our measure, they divide the number of medal-winning countries by the number of participating countries in each sport. We believe that our measure is more reliable, as the entry barriers for participation are extremely different (sprinting vs. sailing), which might artificially increase the competitive imbalance for sprinting relative to sailing.

𝐶𝐵 , = ∑ ,∑

(7)

Comprehensive socioeconomic data is only available from 1960 (population) or 1990

(GDP per capita) onward. We assume that these factors primarily influence the number of

medals won in total and less in what sports these are won. As we already include the medal

portfolio size, we do not expect these to be of great predictive power. Nevertheless, we add

these variables (their mean value in the three years preceding the respective Games10 ) in a

model that contains only the years following the dissolution of the Soviet Union in a robustness

check. As the literature confirms diminishing returns to scale for these two factors (Bernard &

Busse, 2004), we include both variables in their natural logarithm.

Table 2. Descriptive statistics of the dataset.

Source: own calculation, data: The Guardian (2016), Olympic.org (2018).

As outlined above, there are several reasons for why being the host or the next host of

the Olympic Games increases medal yields considerably. We therefore consider a variable ‘host’

(‘current’, ‘next’, ‘previous’) with the reference category ‘no’ to test this effect in the context

of diversification. The increasing number of events per sport and the participation of more

countries over time sharply increases the share of non-entries (0). Therefore, we consider fixed

effects for the points of observation (each Olympic Games). Descriptive statistics of the dataset

are displayed in Table 2. We do not observe issues of multicollinearity or the presence of outliers

(as indicated by observations with high leverage). The main model (model 2 in Table 4) can be

written in simple form as:

𝐸𝑛𝑡𝑟𝑦 , , = 𝛽0 + 𝛽1𝑃𝑟𝑜𝑥 , , + 𝛽2𝑃𝑆 , + 𝛽 𝐶𝐵 , + 𝛽 𝑁𝑂𝑀 , + 𝐻𝑜𝑠𝑡 , + 𝑌𝑒𝑎𝑟 (8)

10 Population and GDP data are retrieved from the World Bank (2018). Gaps for few countries are filled with United Nations Data (2018) and the Taiwanese national statistics (2018). Population data are extrapolated with the previous and/or posterior growth rate for few missing country–year observations.

Results To get a better impression of the data’s dimensions, the number of ‘entries’ (1,206) (from 1900

to 2016) per country, sport, and year are displayed in Table 3. The number increases sharply

when new sports are introduced in the early Olympics. Subsequently, it remains relatively stable,

as the Olympic sport portfolios become more saturated and are primarily driven by new

countries entering the Olympic sports scene after World War II. It rises again after 1996 with

the dissolution of the Soviet Union. The countries with the highest number of entries tend to be

those that were not part of, or not extremely successful at the earliest Olympics. The USA,

France, or Germany, for example, ‘entered’ many sports when they were part of the Olympic

program for the first time, a period for which we cannot predict their entrance probabilities due

to a lack of information. Italy, the Soviet Union, Spain, Poland, and Australia (35 to 45 entries)

rank highest in this table. As expected, weightlifting, boxing, and judo, with their variety of

weight classes, are the most frequently ‘entered’ sports.

Table 3. Entries to national sport portfolios by country, sport, and year

Source: own calculation, data: The Guardian (2016), Olympic.org (2018).

The first regression model confirms that competitive balance has a significant positive

effect on the probability of winning a medal in a sport for the first time (0.022***11). The effect

11 Raw coefficients in brackets. Coefficients transformed to odds ratios are written in percentages.

goes beyond the mere number of medals available (0.030***), upon which our measure of

balance depends. Countries specialized in more distinct sports also have higher likelihoods of

adding even more sports to their portfolio (0.184***). However, the predictive accuracy of the

regression models is low. This is expected and reasonable given the high uncertainty and

complexity of Olympic competitions. Expressed in odds-ratios, being specialized in one more

sport increases the probability of an entry by about 20%12 , whereas each 1% increase in

competitive balance increases the probability of an entry by approximately 2.2%. The additional

effect of proximity (model 2) is slightly smaller (0.015**). Relatedness density, which is on the

same scale as competitive balance (0–100), has a clearer and more pronounced effect

(0.040***). An increase of 1% makes an entry to this sport 4% more likely. However, the

general probability of entering a sport is low (only 2% of all medal wins are entries).

Accordingly, large changes in relatedness density are necessary to considerably increase or

decrease the likelihood of winning a medal in a new sport (model 3). Noticeably, the effect of

portfolio size decreases when proximity or relatedness density is added (model 2–3). This

signals that the specific selection of sports matters. The more that related sports are part of a

country’s current portfolio, the more likely it will become successful in the focal sport as well.

The magnitude of success in each of the related sports (proximity, model 2) seems to be less

important.

Regardless, by far the most important factor for success in new sports is the host

advantage. Compared to countries that are not hosting the Games (previous/next host excluded),

a host country has a five times higher chance of winning a medal in a sport in which it has not

been successful in the past (model 1–3). This is an expected result given the strong host effect

on overall medal counts (e.g. Bernard & Busse, 2004; Lui & Suen, 2008).

The stability of the models can be confirmed with three alternative specifications.

Excluding all countries that cannot have any relatedness to a potential new sport (those with no

medals in 𝑡 1 at all) changes the results only marginally (model 4). The second one (model 5),

includes only the Games from 2000 onward to eliminate any disturbances such as boycotts or

dissolutions of countries (GDR, Soviet Union). In addition, population and GDP per capita can

be considered. In this specification, the effects of relatedness density and competitive balance

become even more important in relation to the respective control variables medals per sport and

portfolio size.

12 𝑒0.1 = 1.202. Calculation of odds-ratio from the log-odds (coefficients).

Table 4. Entry model – effects of relatedness and competition on the diversification of countries into new sports.

Estimation: Binomial (Logit)

Dependent Variable: Entry = 1 (A country’s first medal in a particular sport)

Competition Proximity Rel. Density

Recent success 2000 - 2016 42 Sports

(1) (2) (3) (4) (5) (6) Proximity 0.015** (0.004) Rel. Density 0.040*** 0.037*** 0.067*** 0.028*** [Boschma et al.] (0.004) (0.004) (0.009) (0.004) Competitive balance 0.022*** 0.023*** 0.023*** 0.021*** 0.049*** 0.028*** (0.004) (0.004) (0.004) (0.005) (0.012) (0.004) Medals per sport 0.030*** 0.030*** 0.031*** 0.031*** 0.021** 0.018*** (0.004) (0.004) (0.004) (0.004) (0.007) (0.002) Portfolio size t-1 0.184*** 0.166*** 0.123*** 0.107*** 0.062*** 0.201*** (0.005) (0.008) (0.009) (0.009) (0.017) (0.014) Population t-1 0.128** (0.040) GDP/capita t-1 0.127** (0.044) Host (current) 1.599*** 1.603*** 1.645*** 1.619*** 0.860** 1.634*** (0.124) (0.124) (0.124) (0.124) (0.286) (0.158) Host (next) 0.552** 0.542** 0.583*** 0.433* 0.833** 0.833*** (0.166) (0.166) (0.166) (0.180) (0.306) (0.192) Host (previous) -0.592** -0.739** -0.508* -0.453* -0.950 -0.693* (0.230) (0.242) (0.231) (0.229) (0.541) (0.306) Year F.E. Yes Yes Yes Yes Yes Yes Constant -3.366*** -3.343*** -3.138*** -3.023*** -8.913*** -3.078*** (0.295) (0.294) (0.295) (0.298) (0.877) (0.353) adj. R² 0.06 0.06 0.06 0.07 0.04 0.07 Observations 61,768 61,768 61,768 38,354 28,311 39,097 Note: Own calculation. Data: The Guardian (2016), Olympic.org (2018) *p<0.05; **p<0.01; ***p<0.001

The socioeconomic controls have positive but rather small effects. Countries with one ‘unit’

increase in population or higher GDP/capita on the log-scale have approximately a 14% higher

likelihood to enter a sport. One unit corresponds, for example, to a difference between 10

million and 27 million people and between US$ 10,000 and 27,000 per capita, respectively.

Noticeably, a small share (2.5%) of all entries to new sports is due to the same athlete

being active in multiple sports, whereby the vast majority of these cases is observed in

swimming, fencing, and shooting. When the Olympic events are further aggregated to 42 sport

(model 6), only seven ‘entries’ by the same athlete in multiple sports remain. Most noteworthy,

Carl Albert Andersen (Norway) won a medal in pole vault in 1900 and then one in artistic

gymnastics in 1908. Overall, this is very rare and a phenomenon of the early Olympics, when

the competitive level was considerably lower. The results of model 6 indicate that the

aggregation results in portfolio size gain in relative importance to relatedness density. The

overlap in financial, physical, infrastructural, or cultural requisites appear to be less relevant

when sports are more different on average, as a result of the higher aggregation to 42 sports.

Discussion The empirical results provide robust evidence that relatedness impacts the types of sports

countries become successful in. However, only a small share of this process can be explained.

Nevertheless, it is possible to trace some national differences in diversification patterns. As

Weber et al. (2019a) point out, the internal capabilities of a country and the external market size

and competition of a sport are the two dimensions that impact prioritization of funding and

therefore shape success in elite sport. Relatedness density and competitive balance refer to these

two dimensions in the context of diversification. To showcase their relative influence, we

compare their effects in separate regressions for each country (Figure 3). The bars indicate the

lower and upper 95% confidence intervals, and statistically significant coefficients are indicated

with solid points. For seven countries, competitive balance has a significant effect on their

diversification scheme, and for six countries it is relatedness density, whereas there is no case

in which both variables significantly drive the diversification process13 . Besides this mixed

finding, the figure also reveals that many countries’ diversification pattern cannot be attributed

to either competitive balance or the relatedness density of the sports. Without additional

information about the countries’ actual approaches to success at the Olympic Games, it is

impossible to draw conclusions, yet it is nonetheless an interesting finding that motivates further

investigation.

One of the countries whose diversification seems to be shaped by relatedness density is

Great Britain (GBR, 32 entries). To translate the framework into a more practical tool, we

visualize relatedness density and the competitive balance of all sports that Great Britain could

have possibly entered in two-dimensional plots for each of the five most recent Olympics

(Figure 4). The black lines indicate the median of all considered sports, and the sports that Great

Britain entered are labeled. High scores on both dimensions indicate that the sports fit very well

to the country’s resources and have a high competitive balance. Hence, a competitive

positioning strategy based on this information would suggest prioritizing these sports in funding.

A comparison of the competitive position (Figure 4) with real funding data can show how

closely a country (intentionally or not) targeted sports according to this framework. With data

from the agency ‘UK sport’ it is possible to illustrate distinct funding figures for eight of the

sports in which Great Britain was unsuccessful until 2000 (Figure 5).

13 A few outliers with very large coefficients are excluded from Figure 3.

Figure 3. Country-specific effects of relatedness density and competitive balance on the entry to new sports. Source: own calculation and visualization, data: The Guardian (2016), Olympic.org (2018). Notes: The bars indicate the lower and upper 95% confidence intervals. Statistically significant coefficients are indicated with solid points.

Figure 4. Relatedness and competitive balance in the diversification of Great Britain at the Olympic Games (2000-2016). Source: own calculation and visualization, data: The Guardian (2016), Olympic.org (2018). Notes: All sports Great Britain could have possibly entered are plotted. The sports actually entered by Great Britain are indicated with bright, labeled points.

Of these eight sports, Great Britain entered the Olympic market in badminton (2000),

taekwondo (2008), and triathlon (2012). These three sports received the highest amount of

funding since the foundation of ‘UK sport’ in 1997. Furthermore, taekwondo was the most

balanced and triathlon the second most related sport since their first observation in 2004 until

their entry. However, a substantial share of the total funding could just be a result of the first

Olympic medals won during this time frame and explain why they stand out in comparison to

the other five sports. Moreover, triathlon and taekwondo were newly introduced to the Olympic

program in 2000. Accordingly, professional structures were less likely to be existence in all

countries at that time. Therefore, early and large investments into such ‘new’ markets could

easily pay off. Lastly, badminton is very popular in China, where the Games took place in 2008.

All these are reasons for the variations in funding and it is difficult to connect these to the

framework of relatedness and competition.

Nevertheless, the influence of relatedness on the diversification of Great Britain is also

visible in sports not covered by the funding data. Apart from synchronized diving (2004), the

sports entered were in the upper half of relatedness density. Dressage (2012) was even the most

related sport to the portfolio of Great Britain and kayak sprint (2000) scored high on both

dimensions. However, in some cases, relatedness and competitive balance clearly did not

materialize. Wrestling (Greco-roman), which has a consistently high competitive balance and

cross cycling, with the highest relatedness density in 2016, clearly stand out, However, Great

Britain did not win a medal in these. This underlines the probabilistic nature of the relatedness

framework.

Although the findings are not wholly conclusive, some recommendations for sport

policy makers can be made based on our analysis. Tracking past developments in elite sport,

and in particular the reasons for the emergence of success in specific sports using the framework

of competition, specialization, and relatedness, will reveal important information and help to

design more successful policies. An in-depth analysis is likely to reveal why certain sports have

benefited the most from certain sport policies, or why specific funding strategies have led to

rather disappointing outcomes. For example, according to Figure 4, table tennis was extremely

unlikely to yield success (consistently in the bottom-left quadrant). By contrast, funding for

table tennis was at its highest level for the Olympic Games 2008 in China, where the sport is

very popular. Therefore, it seems fruitful to investigate the link between relatedness,

competition, and targeting in more depth to further detect mechanisms through which countries

become successful in sports. To what degree do countries actively prioritize sports related to

their current strengths? Even more generally: is relatedness between sports a factor that national

sport associations consider in their strategic planning for a national elite sport system (as in De

Bosscher, Shibli, Westerbeek, & Van Bottenburg, 2015)?

Figure 5. Funds dedicated to Olympic sports by the British agency ‘UK sport’. Source: own visualization, data: UK Sport (2019). Notes: Only those sports in which Great Britain was unsuccessful before 2000 and to which funding could be directly assigned.

Conclusion The present study discusses the concept of sport relatedness. Based on the approach of Hidalgo

et al. (2007) for quantifying the degree of relatedness between export products, relatedness

between sports is calculated and visualized as a ‘sport space.’ This network represents related

groups of Olympic sports and indicates which ones could share features such as physical and

infrastructural requirements, or cultural heritage. The subsequent empirical analysis covering

the whole history of modern Olympics confirms that medals in related sports have a positive

effect on countries’ likelihood of winning their first Olympic medal in a specific sport, though

a balanced competition (lower entry barrier) (Truyens et al., 2016) shows similar positive

effects. The host advantage that was shown to affect national success at the Olympic Games

(e.g. Bernard & Busse, 2004; Lui & Suen, 2008), is also a strong driving force behind entry

into new fields of sports. Accordingly, relatedness is but one factor among others that shapes

countries’ diversification into new sports. However, the concept will be informative as a

supplement to the internal factors in a national competitive positioning strategy (Weber et al.,

2019a; Weber et al., 2019b).

However, a number of shortcomings of the empirical study must be pointed out. Firstly,

relatedness based on co-occurrence measures can be rather deterministic as it does not

necessarily reveal actual cross-fertilization and learning. It can be mainly driven by cultural and

historical events or capital intensity. This cannot be distinguished in the present approach.

Studies with more direct measures of relatedness, such as information on co-operations in

training, mutual drills, or the connectedness of athletes in social networks could validate our

findings.

Secondly, the results can be improved by considering alternative measures of sport

success as a basis for the co-occurrence calculations. For instance, participation and results of

world championships as well as world rankings will make the results more reliable and will

allow for a more time-variant measure of relatedness. Moreover, this would likely improve the

dependent variable of ‘entry into a sport.’ In our case, sport capabilities could have been present

before countries win their first Olympic medals, but we do not observe this empirically.

Thirdly, the study assigns a great weight to national borders, i.e., factors determining

countries’ success reside primarily within a country. However, coaches and athletes are mobile

(Jansen & Engbersen, 2017). For instance, nearly the whole silver medal-winning Qatari

handball team at the 2015 World Championships consisted of players born in Eastern Europe

(Hamann, 2015). Similarly, many East African runners emigrate and change nationality due to

high competitive pressure in their home countries (Chemi & Fahey, 2016). Some athletes form

international training groups and practice where the conditions are best. All these processes

imply that the factors impacting athletes’ success are not necessarily related to the countries

they win the medals for. Finally, the consideration of winter sports might reveal interesting

relationships to summer sports, as they are concentrated in a few countries and more expensive

overall.

Despite these shortcomings, the study marks an additional step in understanding the

influence of place-specific factors on sports and sport success. It presents a novel methodology

(at least in this context) for analyzing the relationship between sports. Moreover, it adds to the

recent literature in which the role of differentiation in sport policy is discussed (De Bosscher et

al., 2019; Truyens et al., 2016; Weber et al., 2019a; Weber et al., 2019b) and presents robust

empirical evidence that the relatedness between sports has an effect on the entry of countries to

new sport markets.

References Andreff, M., Andreff, W., & Poupaux, S. (2008). Les détérminants économiques de la

performance olympiques: prévision des médailles qui seront gagnées aux jeux de pékin.

Revue d'économie politique, 118, 135-169. https://doi.org/10.3917/redp.182.0135

Balland, P. A., Boschma, R., Crespo, J., & Rigby, D. (2019). Smart specialization policy in the

European Union: relatedness, knowledge complexity and regional diversification,

Regional Studies, 53, 1252-1268. https://doi.org/10.1080/00343404.2018.1437900

Berdahl, J. L., Uhlmann, E. L., & Bai, F. (2015). Win-win: Female and male athletes from more

gender equal nations perform better in international sports competitions. Journal of

Experimental Social Psychology, 56, 1-3. https://doi.org/10.1016/j.jesp.2014.08.003

Bernard, A. B., & Busse, M. R. (2004). Who wins the Olympic Games: Economic resources

and medal totals. The Review of Economics and Statistics, 86, 413-417.

https://doi.org/10.1162/003465304774201824

Boschma, R. A., Balland, P.-A., & Kogler, D. F. (2015). Relatedness and technological change

in cities: the rise and fall of technological knowledge in US metropolitan areas from

1981 to 2010. Industrial and Corporate Change, 24, 223-250.

https://doi.org/10.1093/icc/dtu012

Boschma, R., & Wenting, R. (2007). The spatial evolution of the British automobile industry:

Does location matter?, Industrial and Corporate Change, 16(2), 213-238.

https://doi.org/10.1093/icc/dtm004

Bredtmann, J., Crede, C. J., & Otten, S. (2016). Olympic medals: Does the past predict the

future? Significance, 13(3), 22-25. https://doi.org/10.1111/j.1740-9713.2016.00915.x

Breschi, S., Lissoni, F., & Malerba, F. (2003). Knowledge-relatedness in firm technological

diversification. Research Policy, 32, 69-87. https://doi.org/10.1016/S0048-

7333(02)00004-5

Chemi, E., & Fahey, M. (2016, August 16). How some Middle East countries are 'buying'

Olympic medals. CNBC. Retrieved from https://www.cnbc.com/2016/08/16/how-

some-middle-east-countries-are-buying-olympic-medals.html

De Bosscher, V., De Knop, P., van Bottenburg, M., Shibli, S., & Bingham, J. (2009). Explaining

international sporting success: An international comparison of elite sport systems and

policies in six countries, Sport Management Review, 12, 113-136.

https://doi.org/10.1016/j.smr.2009.01.001

De Bosscher, V., Heyndels, B., De Knop, P., van Bottenburg, M., & Shibli, S. (2008). The

paradox of measuring success of nations in elite sport. Belgeo, 2, 217-234.

https://doi.org/10.4000/belgeo.10303

De Bosscher, V., Shibli, S., & Weber, A. C. (2019). Is prioritisation of funding in elite sport

effective? An analysis of the investment strategies in 16 countries, European Sport

Management Quarterly, 19(2), 221-243.

https://doi.org/10.1080/16184742.2018.1505926

De Bosscher, V., Shibli, S., Westerbeek, H., & van Bottenburg, M. (2015). Successful Elite Sport

Policies: An International Comparison of the Sports Policy Factors Leading to

International Sporting Success (SPLISS 2.0) in 15 Nations. UK: Meyer & Meyer Sport.

Du Bois, C., & Heyndels, B. (2007). Revealed Comparative Advantage and Specialization in

Athletics (IASE/NAASE Working Paper Series, No. 07-17). Retrieved from

http://web.holycross.edu/RePEc/spe/DuBoisHeyndels_Specialization.pdf

Forrest, D., McHale, I. G., Sanz, I., & Tena, J. D. (2016). An analysis of country medal shares

in individual sports at the Olympics. European Sport Management Quarterly, 17(2),

117-131. https://doi.org/10.1080/16184742.2016.1248463

Gibson, O. (2016, August 15). Team GB's Olympic success: five factors behind their Rio medal

rush. The Guardian. https://www.theguardian.com/sport/blog/2016/aug/15/five-factors-

team-gb-olympic-success-medal-rush

Green, M. (2004). Changing policy priorities for sport in England: the emergence of elite sport

development as a key policy concern. Leisure Studies, 34(4), 365-385.

https://doi.org/10.1080/0261436042000231646

Grimes, A. R., Kelly, W. J., & Rubin, P. H. (1974). A socioeconomic model of national olympic

performance. Social Science Quarterly, 55(3), 777-783.

https://www.jstor.org/stable/42860287

Hamann, B. (2015, January 20). Handball-WM-Gastgeber Katar: Das ist alles nur gekauft.

[Handball world championship host Qatar: That's all just bought]. Spiegel Online.

http://www.spiegel.de/sport/sonst/handball-wm-2015-katar-kauft-sich-spieler-und-

fans-a-1013907.html

He, C., Yan, Y., & Rigby, D. (2016). Regional industrial evolution in China. Papers in Regional

Science, 97, 173-198. https://doi.org/10.1111/pirs.12246

Heck, S. (2014). A Sport for Everyone? Inclusion and Exclusion in the Organisation of the First

Olympic Modern Pentathlon. The International Journal of the History of Sport, 526-

541. https://doi.org/10.1080/09523367.2013.798305

Hidalgo, C. A., Klinger, B., Barabási, A.-L., & Hausmann, R. (2007). The Product Space

Conditions the Development of Nations. Science, 317, 482-487.

https://doi.org/10.1126/science.1144581

Hoffmann, R., Chew, L., & Ramasamy, B. (2002). Public policy and olympic success. Applied

Economics Letters, 9(8), 545-548. https://doi.org/10.1080/13504850110102784

Hooley, G., Piercy, N.F., Nicoulaud, B. and Rudd, J.M. (2017), Marketing Strategy and

Competitive Positioning, 6th ed., Pearson Education, Harlow.

Houlihan, B. & Zheng, J. (2013). The Olympics and Elite Sport Policy: Where Will it All End?,

The International Journal of the History of Sport, 30(4), 338-355.

https://doi.org/10.1080/09523367.2013.765726

Hunt, S. D. & Morgan, R. M. (1995). The Comparative Advantage Theory of Competition.

Journal of Marketing, 59(2), 1-15. https://doi.org/10.2307/1252069

Jansen, J., & Engbersen, G. (2017). Have the Olympic Games become more migratory? A

comparative historical perspective. Comparative Migration Studies, 5(11), 1-15.

https://doi.org/10.1186/s40878-017-0054-2

Johnson, D. K. N., & Ali, A. (2000). Coming to Play or Coming to Win: Participation and

Success at the Olympic Games. Wellesley College Working Paper, No. 2000-10.

Johnson, D. K. N. & Ali, A. (2004). A Tale of Two Seasons: Participation and Medal Counts at

the Summer and Winter Olympic Games. Social Science Quarterly, 85(4), 974-993.

https://doi.org/10.1111/j.0038-4941.2004.00254.x

Kogler, D. F., Rigby, D., & Tucker, I. (2013). Mapping Knowledge Space and Technological

Relatedness in US Cities, European Planning Studies, 21(9), 1374-1391.

https://doi.org/10.1080/09654313.2012.755832

Klement, B. & Strambach, S. (2019). Innovation in Creative Industries: Does (Related) Variety

Matter for the Creativity of Urban Music Scenes? Economic Geography, 95(4), 385-

417. https://doi.org/10.1080/00130095.2018.1549944

Lee, J. & Price, N. (2016). A national sports institute as a learning culture, Physical Education

and Sport Pedagogy, 21(1), 10-23. https://doi.org/10.1080/17408989.2015.1072507

Leoni, A. (2012, August 13). Arme deutsche Athleten: 44 Medaillen, 626 Euro monatlich. [Poor

German athletes: 44 medals, 626 Euro per month]. Hamburger Abendblatt. Retrieved

from https://www.abendblatt.de/sport/article108592586/Arme-deutsche-Athleten-44-

Medaillen-626-Euro-monatlich.html

Lo Turco, A., & Maggioni, D. (2016). On firms' product space evolution: the role of firm and

local product relatedness. Journal of Economic Geography, 16, 975-1006.

https://doi.org/10.1093/jeg/lbv024

Lozano, S., Villa, G., Guerrero, F., & Cortés, P. (2002). Measuring the performance of nations

at the summer Olympics using data envelop analysis. Journal of the Operational

Research Society, 53, 501-511. https://doi.org/10.1057/palgrave.jors.2601327

Lui, H.-K., & Suen, W. (2008). Men, Money, and Medals: An Economic Analysis of the

Olympic Games. Pacific Economic Review, 13(1), 1-16. https://doi.org/10.1111/j.1468-

0106.2007.00386.x

Maennig, W., & Wellbrock, C.-M. (2008). Sozio-ökonomische Schätzungen olympischer

Medaillengewinne: Analyse-, Prognose- und Benchmarkmöglichkeiten. [socio-

economic estimations of olypmic medal yields: possibilities to analyse, predict, and set

benchmarks] (Hamburg contemporary economic discussions, No. 20). Retrieved from

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1520581

Moosa, I. A., & Smith, L. (2004). Economic development indicators as determinants of medal

winning at the Sydney Olympics: An extreme bounds analysis. Australian Economic

Papers, 43(3), 288-301. https://doi.org/10.1111/j.1467-8454.2004.00231.x

Neffke, F., Henning, M., & Boschma, R. A. (2011). How Do Regions Diversify over Time?

Industry Relatedness and the Development of New Growth Paths in Regions. Economic

Geography, 87(3), 237-265. https://doi.org/10.1111/j.1944-8287.2011.01121.x

Nevill, A. M., Balmer, N. J., & Winter, E. M. (2009). Why Great Britain's success in Beijing

could have been anticipated and why it should continue beyond 2012. British Journal

of Sports Medicine, 43(14), 1108-1110. https://doi.org/10.1136/bjsm.2008.057174

NOC*NSF (2016). Sport inspires! Sport Agenda 2016. Retrieved from

https://www.nocnsf.nl/stream/12046wtmposterspagenda2016planouk.pdf

Olympic.org (2018, March 16). Olympic.org. International Olypmic Committee. Retrieved

March 16, 2018 from https://www.olympic.org/olympic-results

Rathke, A., & Woitek, U. (2008). Economics and the Summer Olympics - An Efficiency

Analysis. Journal of Sports Economics, 9(5), 520-537.

https://doi.org/10.1177/1527002507313743

Rigby, D. L. (2015). Technological Relatedness and Knowledge Space: Entry and Exit of US

Cities from Patent Classes. Regional Studies, 49(11), 1922-1937.

https://doi.org/10.1080/00343404.2013.854878

Rynne, S. (2014). 'Fast track' and 'traditional path' coaches: affordances, agency and social

capital. Sport, Education and Society, 19(3), 299-313.

https://doi.org/10.1080/13573322.2012.670113

Seiler, S. (2013). Evaluating the (Your Country Here) Olympic Medal Count. International

Journal of Sports Physiology and Performance, 8, 203-210.

https://doi.org/10.1123/ijspp.8.2.203

Tcha, M., & Perchin, V. (2003). Reconsidering Performance at the Summer Olympics and

Revealed Comparative Advantage. Journal of Sports Economics, 4(3), 2016-239.

https://doi.org/10.1177/1527002503251636

The Guardian (2016, March 16). Olympic medal winners: every one since 1896 as open data.

The Guardian. Retrieved March 16, 2018 from https://docs.google.com/spreadsheets/d/

1zeeZQzFoHE2j_ZrqDkVJK9eF7OH1yvg75c8SaBcxaU/edit#gid=322436777

Taiwanese national statistics (2018, March 16). National Statistics, Republic of China (Taiwan).

Retrieved March 16, 2018, from https://eng.stat.gov.tw/mp.asp?mp=5

Truyens, J., De Bosscher, V., & Heyndels, B. (2016). Competitive balance in athletics.

Managing Sport and Leisure, 21(1), 23-43.

https://doi.org/10.1080/23750472.2016.1169213

United Nations Data (2018, March 16). UN data, A world of information. United Nations

Statistics Division. Retrieved March 16, 2018, from http://data.un.org/

UK Sport (2019, November 15). UK sport: inspire the nation. UK Sport. Retrieved November,

15, 2019, from https://www.uksport.gov.uk/our-work/investing-in-sport/historical-

funding-figures

van Eck, N. J., & Waltman, L. (2009). How to Normalize Co-Occurrence Data? An Analysis of

Some Well-Known Similarity Measures. Journal of the American Society for

Information Science and Technology, 60(8), 1635-1651.

https://doi.org/10.1002/asi.21075

Weber, A. C., De Bosscher, V., & Kempf, H. (2019a). Positioning at the Olympic Winter Games.

Examining the targeting of Olympic Winter Sports by medal-winning nations. Sport,

Business and Management: An International Journal, 9(5), 417-442.

https://doi.org/10.1108/SBM-01-2018-0002

Weber, A. C., De Bosscher, V., Shibli, S., & Kempf, H. (2019b). Strategic analysis of medal

markets at the Winter Olympics. Introducing an index to analyse the market potential of

sports disciplines. Team Performance Management: An International Journal, 25(3/4),

229-252. doi.org/10.1108/TPM-10-2018-0068

Witt, U., Broekel, T., & Brenner, T. (2007). Knowledge and its Economic Characteristics - A

Conceptual Clarification (Jena Economic Research Papers, 2007-013). Retrieved from

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1019222

World Bank (2018, March 16). World Bank Open Data. The World Bank Group. Retrieved

March 16, 2018, from https://data.worldbank.org/

Zheng, J., & Chen, S. (2016). Exploring China's success at the Olympic Games: a competitive

advantage approach. European Sport Management Quarterly, 16(2), 148-171.

https://doi.org/10.1080/16184742.2016.1140797

1

Appendix 1. Different aggregations of sports.

Sport 42 Sports 61 Sports Original Event

Aquatics

Diving Diving

Synchronized Diving

Swimming

Backstroke all distances

Breaststroke “

Butterfly “

Freestyle “

Medley “

Synchronized Swimming

Water polo

Archery

Athletics (track & field)

Combined Combined Heptathlon and Decathlon

Endurance Endurance 3000m to Marathon incl. Steeplechase

Middle Distance 800m and 1500m

Jumps Jumps Long Jump, Triple Jump, High Jump

Pole vault

Race Walk all distances

Sprint Hurdles all distances

Sprint 60 to 400m

Throws

Discus

Hammer

Javelin

Shot put

Badminton

Baseball

Basketball

Boxing

Canoe / Kayak

Canoe Sprint

Kayak Sprint

Canoe/Kayak Slalom

Cycling

Road Cycling

Track Cycling all events on indoor race track

Cross Cycling bmx, mountain bike

Equestrian

Dressage

Eventing

Jumping

Fencing

Epéé

Foil

Sabre

Football

Gymnastics

Artistic Gymnastics

Rhythmic Gymnastics

Trampoline

2

Handball

Hockey

Judo

Modern Pentathlon

Rowing

Sailing

Shooting

Pistol pistol

Rifle rifle, running target

Shotgun trap, skeet, clay pigeons

Softball

Table tennis

Taekwondo

Tennis

Triathlon

Volleyball Beach Volleyball

Volleyball

Weightlifting

Wrestling Wrestling Freestyle

Wrestling Greco-Roman

Notes: The aggregation of sports shows which original events of the Olympic Summer Games

(last column) have been aggregated to 61 sports, 42 sports (for the robustness check), belonging

to the broader categories of sports (first column). Sports that were part of less than five Olympic

Games or only part of the earliest Olympics (e.g. Tug of War) have been excluded.

3

Appendix 2. Robustness check: Count data regression. Estimation: Quasi maximum-likelihood

Dependent Variable: Number of medals (of a country in a particular sport at one Olympic Games)

Success in the same

sport

Speciali-zation Proximity Without

1980 and 1984 2000 - 2016 42 Sports

(7) (8) (9) (10) (11) (12)

Proximity t-1 0.007*** 0.007*** 0.014*** 0.002*

(0.001) (0.001) (0.001) (0.001)

Number of medals t-1 1.303*** 1.183*** 1.122*** 1.131*** 1.186*** 1.099***

(0.014) (0.015) (0.016) (0.017) (0.029) (0.014)

Specialization t-1 0.028*** 0.028*** 0.029*** 0.036*** 0.028***

(0.001) (0.001) (0.001) (0.002) (0.001)

Host (current) 0.691*** 0.706*** 0.692*** 0.737*** 0.413*** 0.712***

(0.032) (0.033) (0.034) (0.037) (0.083) (0.036)

Host (next) 0.076 0.091* 0.083 0.083 0.178 0.102*

(0.041) (0.043) (0.044) (0.044) (0.092) (0.045)

Host (previous) -0.320*** -0.244*** -0.336*** -0.337*** -0.127 -0.266***

(0.042) (0.045) (0.047) (0.050) (0.076) (0.046)

Country, Sport, Year, Fixed Effects Yes Yes Yes Yes Yes Yes

Constant -4.434** -4.515*** -4.368*** -4.416*** -30.068*** -4.184***

(1.006) (1.022) (1.029) (1.071) (6.173) (1.068)

adj. R² 0.42 0.44 0.44 0.43 0.52 0.58

Observations 73,512 72,551 72,551 65,669 31,075 51,381 Note: Own calculation. Data: Olympic.org (2018) and The Guardian (2016). *p<0.05; **p<0.01; ***p<0.001

The count data regression (Poisson quasi-maximum-likelihood) with the number of medals of

a country in a sport at the Olympic Games substantiates the findings of the article. The number

of medals the country won in that sport at the previous Games (Number of medals t-1), meaning

ongoing success in the same sport, is by far the best predictor for the number of medals. The

degree of specialization in that sport (Specialization t-1) and the host advantage also positively

affect medal counts. Nevertheless, the proximity of a country to a sport (as calculated in the

main document) also has a significant positive effect on the number of medals a country wins

in a sport. Though, the additional variation explained by the variable is, again, very small. Two

subsets of the data (model 10 and 11) confirm the robustness of these findings. However, just

like in the main findings (model 6), reduced to 42 sports, the connection between relatedness

and sporting success declines (model 12). Again, the stronger aggregation and, therefore, lower

average relatedness between sport might explain the sharp decrease in effect size.

4 A

ppendix 3. Possible entries at the Olym

pic Gam

es in Tokyo 2021. C

ountry Sport

Probability

Country

Sport R

el. Density

RU

S Football

0.32

RU

S Football

67.75 JPN

Tram

poline 0.31

U

SA Tram

poline 61.65

USA

Badminton

0.24

GB

R

Cross C

ycling 57.64

GER

B

adminton

0.19

GER

B

adminton

56.28 G

BR

C

ross Cycling

0.19

FRA

R

ace Walk

45.25 FR

A

Race W

alk 0.15

JPN

Tram

poline 42.12

CH

N

Kayak Sprint

0.09

CH

N

Kayak Sprint

41.53 C

AN

C

anoe/Kayak Slalom

0.06

C

AN

C

anoe/Kayak Slalom

39.10

ITA

Table Tennis 0.04

ITA

Table Tennis

34.57 A

US

Dressage (Equestrian)

0.04

BR

A

Weightlifting

33.55 B

RA

W

eightlifting 0.03

A

US

Dressage (Equestrian)

32.88 ESP

Sprint (Athletics) 0.03

ESP

Trampoline

30.58 R

SA

Canoe/K

ayak Slalom

0.03

RSA

C

anoe/Kayak Slalom

30.58

SWE

Judo 0.03

D

EN

Race W

alk 30.12

DEN

R

ace Walk

0.03

NZL

Road C

ycling 28.66

NZL

Road C

ycling 0.03

C

ZE D

ressage (Equestrian) 27.86

AZE

Artistic G

ymnastics

0.03

UK

R

Javelin Throw

27.13 C

ZE Track C

ycling 0.03

K

OR

M

odern Pentathlon 25.91

UK

R

Javelin Throw

0.02

CR

O

Race W

alk 24.75

Source: Ow

n calculation. Data: O

lympic.org (2018) and The G

uardian (2016). Left: Predicted Probabilities (highest per country) based on Rel. D

ensity, Com

p. Balance, H

ost status and available m

edals (coefficients of model 3). R

ight: Highest value of relatedness density per country.

The table shows the highest probability of entry (left) and the highest values of R

elatedness Density (right) (from

high to low and only the highest per

country – otherwise Japan (as upcom

ing host) and the USA

would appear several tim

es). The rows highlighted in italics are the only ones in w

hich

the left and the right table differ. For example, Track C

ycling delivers more m

edal winning opportunities than D

ressage and has a higher competitive

balance. Therefore, the model predicts a higher probability of entry for the C

zech Republic than expected solely by the value of relatedness density.

The probabilities are relatively low for m

ost cases as the overall explanatory power of the m

odel is rather weak.

5

Appendix 4. The network of Olympic medal-winning countries (2016). Notes: Nodes represent countries. Nodes are positioned based on their non-random links (𝜑 > 1) (adjusted relatedness 2016). For better visibility, only the strongest 15% of links between countries are displayed. Node size is given by GDP per capita (2013-2016 average). Source: own calculation and visualization. Data: Olympic.org (2018) and The Guardian (2016).

The network of sports can be transposed to the network of countries (Appendix 4), based on the

similarity of their medal portfolios at the Olympic Games. It shows historical similarities

between countries through the sports in which they are successful (co-occurrences of countries’

athletes on the ‘same’ sports’ podium). We find a clustering of countries located on the same

continent. In addition, countries with a higher GDP per capita are grouped together in the central

part of the network. This reveals similarities in (sport) culture, as well as the influence of socio-

economic conditions on success in sports. There are some sub-continental clusters as well.

6

Though not showing strong direct links, Azerbaijan (AZE), Armenia (ARM), and Georgia

(GEO) are close to each other (upper part of the network). Moreover, former and current

communist countries China (CHN), North Korea (PRK), Cuba (CUB), GDR, the Soviet Union

(URS), and many of its successor states are relatively close to each other at the left of the center.

Some counterintuitive strong connections (e.g. Luxembourg and Jamaica) arise from occasional

co-occurrences. In this case, both countries won a medal in mid-distance running at the 1952

Olympics in Helsinki.

References

Olympic.org (2018, March 16). Olympic.org. International Olypmic Committee. Retrieved

March 16, 2018 from https://www.olympic.org/olympic-results

The Guardian (2016, March 16). Olympic medal winners: every one since 1896 as open data.

The Guardian. Retrieved March 16, 2018 from https://docs.google.com/spreadsheets/d/

1zeeZQzFoHE2j_ZrqDkVJK9eF7OH1yvg75c8SaBcxaU/edit#gid=322436777