Sport drop-out during adolescence: is it real, or an

artefact of sampling behaviour?

This is the Accepted version of the following publication

Eime, Rochelle, Harvey, John and Charity, Melanie (2019) Sport drop-out
during adolescence: is it real, or an artefact of sampling behaviour?
International Journal of Sport Policy and Politics, 11 (4). pp. 715-726. ISSN
1940-6940

The publisher’s official version can be found at

https://www.tandfonline.com/doi/full/10.1080/19406940.2019.1630468
Note that access to this version may require subscription.

Downloaded from VU Research Repository https://vuir.vu.edu.au/39785/

 International Journal of Sport Policy and Politics

Sport drop-out during adolescence: Is it real, or an artefact

Fo
of sampling behaviour?
rP

Journal: International Journal of Sport Policy and Politics

Manuscript ID Draft
ee

Manuscript Type: Research Article

Keywords: sport, participation, drop-out, sampling, specialising

rR
ev
iew
On
ly

URL: http://mc.manuscriptcentral.com/risp
Page 1 of 47 International Journal of Sport Policy and Politics

1
2
3 Sport drop-out during adolescence: Is it real, or an artefact of
4
5 sampling behaviour?
6
7
8
9
10 Abstract
11
12 Understanding sport participation and drop-out is important for sport management. Many
13
14 children sample or play multiple sports and then specialise. However, quantifying these
15
16 behaviors is challenging. Sport registration databases are potentially useful for this purpose.
Fo
17
18 However, given privacy and data security issues, identification and direct linking of data
19
rP
20 records of individual participants across sports is not possible. This study demonstrates a
21
22 feasible methodology for approximate cross-linking of de-identified data and thereby
23
ee

24 quantifying the extent of sampling behaviour, and hence investigating to what degree the
25
26 decline in community club-based sport participation observed during adolescence is
rR

27
28 attributable to a “sampling to specialisation” effect as opposed to drop-out from sport
29
ev

30 altogether. Participants were registered members of one of eleven state sporting associations
31
32 in 2015. For this analysis, data (907,150 player records) were amalgamated, and players
iew

33
34 were categorised by sex, age group and residential postcode. Numbers of individual players
35
36 were estimated using demographic matching, comparing numbers of registrations and
37
numbers of individual participants across age, sex and region. Results showed that the effect
On

38
39
40 of participants playing multiple sports is highest for ages 5-14, and then participants tend to
41
play fewer sports. This study confirms that the drop-off in community sport participation
ly

42
43
44
during adolescence is real and not simply an artefact of sampling behaviour. It is
45
recommended that national sport policy focuses on overall participation across sports, taking
46
47
into account the sampling and specialising phenomena which naturally occur during
48
49
childhood and adolescence, rather than merely asking individual sports to increase
50
51
participation numbers.
52
53
54
55
56
57
58 1
59
60 URL: http://mc.manuscriptcentral.com/risp
 International Journal of Sport Policy and Politics Page 2 of 47

1
2
3 Key words: sport, participation, drop-out, sampling, specialising
4
5
6
7
8
9
10
11
12
13
14
15
16
Fo
17
18
19
rP
20
21
22
23
ee

24
25
26
rR

27
28
29
ev

30
31
32
iew

33
34
35
36
37
On

38
39
40
41
ly

42
43 Introduction
44
45 It is well known that participation in sport is popular among children, and that sport
46
47 participation rates peak during childhood to early adolescence (Australian Sports
48
49 Commission 2016, Eime, Harvey, Charity and Payne 2016, Wong, Olds et al. 2016). A
50
51
recent survey of children aged 0-14 years reported sport participation peaking at ages 9-11
52
(Australian Sports Commission 2016), and other studies of sport participants have reported
53
54
peak participation at ages 10-14 years (Eime, Harvey, Charity and Payne 2016, Wong, Olds
55
56
et al. 2016). Recent research reports that nearly a third of sports participants are aged 10-14
57
58 2
59
60 URL: http://mc.manuscriptcentral.com/risp
Page 3 of 47 International Journal of Sport Policy and Politics

1
2
3 years (Eime, Harvey, Charity, Casey et al. 2016). However many people drop out of sport
4
5 during childhood and adolescence, especially females (Australian Sports Commission 2016,
6
7 Eime, Harvey, Charity, Casey et al. 2016, Eime, Harvey, Charity and Payne 2016, Wong,
8
9 Olds et al. 2016) . The sport participation decline is further evident throughout the lifespan,
10
11 when there is evidence of a shift away from competitive club-based sport towards non-
12
13 competitive and non-organised forms of leisure-time physical activity (Eime, Sawyer et al.
14
15 2015, Australian Sports Commission 2016, Eime, Harvey, Charity and Payne 2016, Harris,
16
Fo
17 Nichols et al. 2017).
18
19 In a large recent longitudinal study of over 200,00 children aged 4-12 it was found that most
rP
20
21 children in a modified sports (introductory) program in the first year of the study did not
22
23
ee

continue playing that particular sport for throughout a 4-year period (Eime, Casey, Harvey,
24
25 Charity et al. 2015). Furthermore, two-thirds of those children who dropped out of the sport
26
rR

27 did so after the first year; it was conjectured that many of these had only a single year or
28
29 season of participation in that particular sport (Eime, Casey, Harvey, Charity et al. 2015).
ev

30
31
32 Whilst there are a range of intrapersonal, interpersonal, organisational, and environmental
iew

33
34 and policy factors influencing participation in sport across the lifespan, competency is a key
35
36 determinant (Crane and Temple 2015, Eime, Casey, Harvey, Sawyer et al. 2015). If people
37
On

38 do not perceive themselves as having adequate skills to play, they do not then enjoy playing
39
40 sport and drop out (Crane and Temple 2015). Sport competency is in turn related to the
41
ly

42 development of fundamental motor skills (FMS). There has been much research around the
43
44 importance of FMS for participation in sport (Veldman, Palmer et al. 2017). FMS are
45
46 learned skills, and children do not develop FMS naturally as part of normal growth and
47
48 development. That is, they need to be taught, practiced and reinforced (Robinson and
49
50 Goodway 2009, Veldman, Palmer et al. 2017). Furthermore, FMS development is
51
52 cumulative and relatively permanent, and sport skills may have an effect on sport
53
54 participation that persists across the lifespan (Henrique, Re et al. 2016). There is clear
55
56 evidence that FMS are important for predicting continued sport participation or drop-out in
57
58 3
59
60 URL: http://mc.manuscriptcentral.com/risp
 International Journal of Sport Policy and Politics Page 4 of 47

1
2
3 children (Henrique, Re et al. 2016). This is thought to be associated with those children with
4
5 advanced FMS being more successful in sport and as a result having more enjoyment than
6
7 those children with lower FMS. Having fun or enjoyment in sport is also a major predictor
8
9 of drop-out of sport participation (Crane and Temple 2015, Gardner, Magee et al. 2017).
10
11 Whilst a lack of perceived competency and enjoyment may be contributing factors to the
12
13 drop-out in participation sport, it has been recently acknowledged that the drop-off in late
14
15 childhood to early adolescence may be partly due to children sampling multiple sports, and
16
Fo
17 then specialising in one (Eime, Casey, Harvey, Charity et al. 2015). This may be further
18
19 explained through the Developmental Model of Sports Participation (DMSP) (Coté and
rP
20
21 Vierimaa 2014). Whilst this model is often used to describe the elite sport performance
22
23
ee

pathway we believe also it is relevant here. Within the DMSP the first stage is the sampling
24
25 stage at 6-12 years, followed by the specialisation stage at 12-15 years and the investment
26
rR

27 stage at 15+ years. At age 6-12, sampling is considered beneficial for athletic development
28
29 because of the exposure to a number of different physical, cognitive, affective and
ev

30
31 psychosocial environments, reinforcing physical, personal and mental skills required for
32
iew

33 future sport elite success (Fransen, Pion et al. 2012, Coté and Vierimaa 2014). Conversely,
34
35 early specialisation often leads to dropout (Fraser-Thomas, Coté et al. 2008). Children who
36
37 specialise in a single sport, particularly with heavy training, risk burnout, overuse injuries
On

38
39 and actual decrease in performance due to overtraining (Myer, Jayanthi et al. 2015).
40
41 However, previous research has focused predominantly on elite athletes. We do not have
ly

42
43 knowledge about the extent of sampling and specialisation, or its effect in general
44
45 community sport participants.
46
47
48 Australian sport policy focuses on increasing participation generally, with no specific
49
50 mention of sampling and specialising behaviours (Australian Sports Commission 2015). The
51
52 current policy focuses on achieving annual increases in participation numbers for individual
53
54 sports, and on talent development for children and youth (Australian Sports Commission
55
56 2015). The authors contend that this strategic focus influences sports to prioritise recruitment
57
58 4
59
60 URL: http://mc.manuscriptcentral.com/risp
Page 5 of 47 International Journal of Sport Policy and Politics

1
2
3 over retention, including the development of organised modified sports programs to very
4
5 young participants, in some sports as young as four years of age (Eime, Harvey, Charity and
6
7 Payne 2016). The authors have recommended a greater focus on retention strategies,
8
9 specifically in relation to the sharp decreases in sport participation during adolescence
10
11 (Eime, Harvey, Charity and Payne 2016).
12
13 Sports organisations in Australia tend to work individually, each through their own network
14
15 of state-based organisations and local sport clubs, with an absence of multi-sport clubs
16
Fo
17 which are more evident in Europe (Breuer, Hoekman et al. 2015). It is conjectured that
18
19 multi-sport clubs may make the transition of participants across sports easier than with
rP
20
21 individual sport clubs. It has also been reported that in European countries with larger multi-
22
23
ee

sports clubs, the sport sector is able to be more innovative and to drive public policy
24
25 programs more effectively than individual sport clubs (Harris, Mori et al. 2009).
26
rR

27
28 In Australia, community sports clubs are most often run by local volunteers (Eime, Payne et
29
ev

30 al. 2009). In the Australian state of Victoria, some sports are funded by the Victorian Health
31
32 Promotion Foundation to focus not just on traditional competitive sport participation, but
iew

33
34 also on the development of more social and recreational forms of sport participation, in an
35
36 effort to get more people active through sport (VicHealth 2018).
37
On

38
It is well-known that many children sample or play multiple sports and then specialise.
39
40 Sampling may involve playing a sequence of different sports or playing multiple sports in
41
ly

42 the same year. Some of the decline we see in sport during childhood and adolescence may
43
44 be due to a decline in the number of sports played in a particular year and not necessarily a
45
46 true drop-out from sport participation (Eime, Casey, Harvey, Charity et al. 2015, Eime,
47
48 Harvey, Charity, Casey et al. 2016). However, quantifying these behaviors is a challenging
49
50 task. Population-wide sample surveys can provide direct evidence of the participation
51
52 patterns of individuals, but sample sizes are determined from a national perspective, and so
53
54 the capacity to produce estimates with acceptable levels of sampling error for smaller
55
56 geographical areas or population segments is limited. Also, historically the scope of such
57
58 5
59
60 URL: http://mc.manuscriptcentral.com/risp
 International Journal of Sport Policy and Politics Page 6 of 47

1
2
3 surveys in Australia has generally been limited to persons 15+ years of age (Eime, Sawyer et
4
5 al. 2015). Furthermore, non-response and refusal rates are high, and recent research has
6
7 suggested that, in addition to sampling error, estimates of participation counts and rates from
8
9 such surveys are subject to substantial non-response bias (Harvey, Charity et al. 2018).
10
11 Sport registration data are prima facie much less susceptible to bias, and as they purport to
12
13 be complete enumerations of participants in each sport, are not subject to sampling error
14
15 (Harvey, Charity et al. 2018). When registration data from multiple sports are
16
Fo
17 amalgamated, the aggregated counts are inflated by individuals playing multiple sports in the
18
19 same year.
rP
20
21
22
23
ee

24 Ideally, to quantify this phenomenon would require the capacity to identify individual
25
26 participants across multiple sports. However, particularly given current awareness of
rR

27
28 privacy and data security issues, identification and direct linking of data records of
29
ev

30 individual participants across sports is not possible. The aim of this study was to
31
32 demonstrate a feasible methodology, based on demographic characteristics, for approximate
iew

33
34 cross-linking of de-identified data and thereby to use sport registration data to quantify, for
35
36 each age level, the extent of participation in multiple sports in a given year, and hence to
37
On

38 investigate to what degree the decline observed in the per capita rate of aggregated
39
40 registrations in community club-based sport during later childhood and adolescence is
41
ly

42 attributable to a reduction in the extent of participation in multiple sports, i.e. a “sampling to

43
44 specialisation” effect.
45
46
47 Methods
48
49 Data for this study were collected as part of the Sport and Recreation Spatial project, a
50
51 research project funded by government and public health agencies in the Australian state of
52
53 Victoria to monitor participation in sport and active recreation, for the purpose of informing
54
55 policy development and program planning in the sport and recreation sector. The data have
56
57
58 6
59
60 URL: http://mc.manuscriptcentral.com/risp
Page 7 of 47 International Journal of Sport Policy and Politics

1
2
3 been previously described in detail (Eime, Harvey, Charity, Casey et al. 2016, Eime,
4
5 Harvey, Charity and Payne 2016). Briefly, participants were registered participants in one of
6
7 eleven state sporting associations (Australian football, basketball, bowls, cricket, football
8
9 (soccer), golf, gymnastics, hockey, netball, sailing and tennis) in Victoria in 2015.
10
11 For this analysis, data for the eleven sports were amalgamated, and registered Victorian
12
13 players aged 4-100 in 2015 were categorised by sex, age group and residential postcode.
14
15 Individuals may play more than one sport, and therefore it is possible a player could be
16
Fo
17 registered with more than one sport in any given year. Actual numbers of registrations and
18
19 estimated numbers of individual players were calculated, to allow comparison of these two
rP
20
21 indicators across age, sex and region.
22
23
ee

24
25 The number of registrations in each year were direct counts of the total of all registrations
26
rR

27 for the 11 sports. As the registrations for each sport were de-identified, it was not possible
28
29 to directly link all data records for a particularly individual. An approximate count of
ev

30
31 individuals was generated by assuming that two registrations for different sports that were
32
iew

33 matched on year, sex, date of birth and residential postcode were the same person. Hence,
34
35 by excluding all but the first occurrence of a particular combination of these four
36
37 characteristics, estimated counts of individual players were obtained. This was done for each
On

38
39 sport separately to provide an indication of ‘multiples’ per sport, and then for the sample as a
40
41 whole.
ly

42
43 A second player count estimate was also generated by assuming that multiple matching
44
45 registrations within a sport genuinely represented different individuals (i.e. same-sex twins
46
47 or children fortuitously matched on demographic characteristics). The ‘Multiple’
48
49 combinations of DOB, postcode and sex within each sport were retained, up to a maximum
50
51 of five occurrences per sport (arbitrarily chosen in order to exclude blocks of spurious
52
53 matches due to data quality issues). For example, two players in Sport A and four players in
54
55 Sport C all with the same details would result in a registration count of six, and individual
56
57
58 7
59
60 URL: http://mc.manuscriptcentral.com/risp
 International Journal of Sport Policy and Politics Page 8 of 47

1
2
3 counts of one under the first option (referred to as Players 1) and four under the second
4
5 option (Players 2). The Players 1 count is always less than or equal to the Players 2 count,
6
7 which in turn is always less than or equal to the registration count.
8
9
10 Results
11
12 There were a total of 907,150 registrations, 757,564 unique individuals assuming “valid”
13
14 multiple individuals within sports as described (Players 2) and 714,054 unique individuals
15
16 assuming no valid multiple individuals (Players 1). Aggregated across all years and sports,
Fo
17
18 the Players 2 count was 16.5% lower, and the Player 1 count was 21.3% lower, than the
19
rP
20 count of registrations. These are indicative of the proportion of sports participants who
21
22 played multiple sports in 2015.
23
ee

24
25
Figure 1a presents the number of registrations and estimated numbers of individual players
26
rR

27
(under the two exclusion rules) across the lifespan, and Figure 1b shows the corresponding
28
participation rates per the Victorian population in standard age cohorts. The pattern within
29
ev

30
both figures is the same. The key result is that the excess of registrations over individual
31
32
iew

players declines with increasing age. That is, the effect of participants playing multiple
33
34
sports is highest for ages 5-14, and then we see a clear specialisation effect where sports
35
36
participants play one or fewer sports as they age through 15-29. From age 30 the player and
37
On

38
registration rates are almost identical, indicating that these older sports participants are
39
40
mostly playing one of the sports only.
41
ly

42
43 Insert Figure 1 and Table 1 about here
44
45
46 Figure 2a and 2b present the participation rates for metropolitan and non-metropolitan
47
48 region. Whilst the participation rate across the lifespan is generally higher in the non-
49
50 metropolitan region compared to the metropolitan region, the proportional difference
51
52 between registrations and actual players is very similar.
53
54 Insert Figure 2 and Table 2 about here
55
56
57
58 8
59
60 URL: http://mc.manuscriptcentral.com/risp
Page 9 of 47 International Journal of Sport Policy and Politics

1
2
3 Figure 3a and 3b present the participation rates for males and females. The overall
4
5 participation rate is much higher for males than females, and peaking at a registration rate of
6
7 81% for males aged 5-9 years and for females of 52% for ages 10-14 years. This
8
9 corresponds to an actual player participation rate of 51% for males aged 5-14 and 51-52%
10
11 for females aged 5-14 years. Therefore, whilst the actual player participation rate is very
12
13 similar for these children for both males and females, males are much more likely than
14
15 females to play multiple sports during childhood and early adolescence. That is, the
16
Fo
17 sampling effect is much greater for males than for females.
18
19 Insert Figure 3 and Table 3 about here
rP
20
21
22
23
ee

24
25
Discussion
26
rR

27 The existence of the sampling effect, whereby many younger sport participants play multiple
28
29 sports in the same year, has often been reported (Fraser-Thomas, Coté et al. 2008, Delorme,
ev

30
31 Chalabaev et al. 2011), but this is the first study to explore how this actually relates to player
32
iew

33 numbers and participation rates. This study is unique in quantifying the extent to which
34
35 younger sports participants sample multiple sports in a single year, identifying the age range
36
37 over which this behaviour diminishes with increasing age, and demonstrating differences
On

38
39 between these patterns for males and females and between metropolitan and non-
40
41 metropolitan sport participants. Furthermore, the majority of research on sampling and
ly

42
43 drop-out has been on elite or sub -elite athletes (Fraser-Thomas, Coté et al. 2008, Delorme,
44
45 Chalabaev et al. 2011, Bridge and Toms 2013), whereas this study examined a population of
46
47 community-based players.
48
49
50 We have previously speculated that the considerable drop-off in registrations during
51
52 adolescence, from 15-19 years, could conceivably be due to the sampling effect, and not
53
54
truly indicative of dropping out of sport altogether (Eime, Harvey and Charity 2016, Eime,
55
56
Harvey, Charity and Payne 2016). We conjectured that the considerable decline during
57
58 9
59
60 URL: http://mc.manuscriptcentral.com/risp
 International Journal of Sport Policy and Politics Page 10 of 47

1
2
3 adolescence might be attributable to children playing multiple sports when younger, and
4
5 thereby being counted multiply in aggregated multi-sport participation data, and then during
6
7 adolescence choosing one sport to specialise in (Eime, Harvey and Charity 2016). The
8
9 results of the present study show that while sampling does amplify the magnitude of the
10
11 reduction in total registrations throughout the adolescent and early adult years, it explains
12
13 only about 25% of the drop-off in the overall registration rate, 40% of the drop-off among
14
15 males, 20% of the drop-off among females, and 25% for each of metropolitan and non-
16
Fo
17 metropolitan residents.
18
19 When examining the estimated player numbers and participation rates only, and not the
rP
20
21 registrations, it is clear that there is a significant decline in sport participation from age 15-
22
23
ee

19. Player participation rates peak at 45-46% ages 5-14 before a drop to only 23% from 15-
24
25 19 years. The participation rate halves from the 10-14 age group (46%) to 23% (15-19%)
26
rR

27 and then halves again for 20-24 year-olds (12%).

28
29
ev

30 This current study highlights that the drop-off in sport participation during adolescence and
31
32 into early adulthood is real, and is not simply the effect of sampling. Sampling itself confers
iew

33
34 a range of benefits, contributing to a range of diverse skills and to improved fitness and
35
36 motor coordination (Fransen, Pion et al. 2012, Myer, Jayanthi et al. 2015). Fransen et al.
37
On

38 showed that for boys, those participating in more than one sport were exposed to a greater
39
40 number of physical, cognitive, affective, and psycho-social environments than boys playing
41
ly

42 a single sport only (Fransen, Pion et al. 2012), and concluded that it is important to be
43
44 sampling and playing multiple sports before the age of 12 rather than early specialisation
45
46 (Fransen, Pion et al. 2012). There is evidence that a higher degree of sampling for ages 11-
47
48 15 is significantly associated with increased performance, that is, competing at a national
49
50 compared to community club level (Bridge and Toms 2013 4724). We know that
51
52 participation in sport during childhood and adolescence has a lasting positive effect on
53
54 physical activity (Murphy, Rowe et al. 2016). Furthermore, sports participants are often
55
56 more active and fitter than participants in non-sport physical activity; however these
57
58 10
59
60 URL: http://mc.manuscriptcentral.com/risp
Page 11 of 47 International Journal of Sport Policy and Politics

1
2
3 associated benefits of sport participation can diminish during adolescence and especially for
4
5 girls (Telford, Telford et al. 2015). Therefore strategies are needed to keep children and
6
7 adolescents engaged in sport for health benefits, not only physical but also psychological
8
9 and social health (Eime, Young et al. 2013).
10
11 The sampling and participation rate patterns were broadly similar for both the metropolitan
12
13 and non-metropolitan regions, and this is consistent with other research (Eime, Harvey,
14
15 Charity and Payne 2016, Hoekman, Breedveld et al. 2017), however the decline is not as
16
Fo
17 pronounced for the non-metropolitan regions. A higher proportion of people living in non-
18
19 metropolitan regions remain active through sport during adolescence. This may relate to the
rP
20
21 culture of community sport in rural and regional communities (Eime, Charity et al. 2015).
22
23
ee

Further, as adolescents age and become more autonomous in their decisions regarding
24
25 physical activity there are also more opportunities to participate in a wider range of leisure
26
rR

27 activities in metropolitan compared to non-metropolitan regions (Craike, Symons et al.

28
29 2011).
ev

30
31
32 A concerning pattern is the more severe decline during adolescence for females compared to
iew

33
34 males. The participation rate for females for ages 15-19 is less than half of that for those
35
36 aged 1-14 years. It is also very clear from the registration numbers that young males sample
37
On

38 or play more sports than females. Firstly, there is evidence of a competency difference
39
40 between young females compared to males, and we know that competency is a major factor
41
ly

42 relating to participation in sport (Veldman, Palmer et al. 2017). There is evidence that young
43
44 girls are less competent at ball skills than young boys (Veldman, Palmer et al. 2017).
45
46 Further, other research amongst 6-12 year olds reports that playing multiple sports and
47
48 spending more time playing sport contributes to improved gross motor coordination
49
50 (Fransen, Pion et al. 2012). Quite simply, by sampling multiple sports children are exposed
51
52 to a greater number of physical, cognitive, affect and psycho-social environments than those
53
54 playing only one sport (Fransen, Pion et al. 2012). Secondly these gender differences may
55
56 also be related to increased opportunities for boys to play multiple sports, relative to girls,
57
58 11
59
60 URL: http://mc.manuscriptcentral.com/risp
 International Journal of Sport Policy and Politics Page 12 of 47

1
2
3 and to parents encouraging boys to play sport more than girls (Wheeler 2012, Eime, Harvey,
4
5 Charity, Casey et al. 2016). Further to this, in the case of this study higher participation for
6
7 males compared to females also relates to the traditional gender bias of Australian club-
8
9 based sport, whereby for several sports, until very recently, females were not able to play in
10
11 club-based competitions (Eime, Harvey, Charity and Payne 2016).
12
13 The main correlates of youth sport attrition include competency and social factors such as
14
15 enjoyment, support from parents, peers and coaches and a positive social club environment
16
Fo
17 (Eime, Harvey et al. 2013, Balish, McLaren et al. 2014, Henrique, Re et al. 2016, Casey,
18
19 Eime et al. 2017, Gardner, Magee et al. 2017). Higher sport competency in young childhood
rP
20
21 is associated with continued sport participation across childhood (Henrique, Re et al. 2016).
22
23
ee

Competency is a cumulative and relatively permanent phenomenon, which is in contrast to

24
25 physical activity in general (Henrique, Re et al. 2016). Children exposed to more sports
26
rR

27 earlier in their life tend to have greater competency and therefore are more likely to continue
28
29 to participate. Further to this, there is evidence that players who drop out participated in
ev

30
31 fewer sports than those that continue playing (Fraser-Thomas, Coté et al. 2008).
32
iew

33
34 There is much research evidence that sampling is positive for skill development in children
35
36 and young people, and for continued sport participation. In Australian sport policy there is
37
On

38 no specific mention of sport sampling, with the focus being the broader aim of getting more
39
40 Australians and particular young Australians participating in sport more often, and in general
41
ly

42 trying to gain insights into participation trends. (Australian Sports Commission 2015).
43
44 There is mention of the benefits of sampling for elite athlete development (Australian Sports
45
46 Commission not dated), but nothing related to the general population. A major focus of the
47
48 national sport policy is for each sport organisations is to increase participant numbers each
49
50 year. With this “individual silos” focus, the issue of sampling versus specialisation in
51
52 children and adolescents, and the implications for overall levels of participation, are not
53
54 considered.
55
56
57
58 12
59
60 URL: http://mc.manuscriptcentral.com/risp
Page 13 of 47 International Journal of Sport Policy and Politics

1
2
3 Beyond the quantification of sampling and specialisation demonstrated in this study, there
4
5 remain important questions. Whilst we have many insights relating to sampling among elite
6
7 and sub-elite youth, we do not have evidence from community club-based sport as to
8
9 whether sampling leads to a greater likelihood of playing sport throughout life. Are the
10
11 people that maintain participation in sport, albeit only a low proportion, those who have
12
13 played and sampled many sports in childhood? Is the drop-off in sport participation
14
15 primarily attributable to ‘late sports starters’ who have less sport competency than early
16
Fo
17 sport adopters? Beside the modified sports programs for young participants, what are the
18
19 entry points to competitive sport throughout the lifespan for individuals who do not have
rP
20
21 specific sports competency? These questions await further research.
22
23
ee

24 Conclusion
25
26 In conclusion, this study confirms that the drop-off during adolescence in community sport
rR

27
28 participation, as measured by aggregated sport registrations, is real and not simply a
29
ev

30 consequence of a reduction in sampling behaviour compared to younger participants. It is

31
32 concerning that, from a peak at ages 10-14, the participation rate halves for the next age
iew

33
34 group of 15-19 years. Considering the magnitude of this drop-off, sport policy should
35
36 specifically prioritise retention in sport, and not merely focus on increasing total sport
37
On

38 numbers. This requires a longitudinal rather than cross-sectional approach to the

39
40 monitoring of participation. A policy relating to retention is needed to provide sporting
41
ly

42 organisations with the lever to make this a priority for their sport-specific strategies.
43
44 It is also recommended that national sport policy should focus on overall participation across
45
46 sports, taking into account the sampling and specialising phenomena which naturally occur
47
48 during childhood and adolescence, rather than merely asking individual sports to increase
49
50 participation numbers. From a health perspective, as long as people are regularly physically
51
52 active, it does not matter if they initially play multiple sports and then subsequently
53
54 specialise. However, from a policy and planning perspective it is important to know what
55
56 proportion of the apparent drop-off in late childhood and adolescence, both in individual
57
58 13
59
60 URL: http://mc.manuscriptcentral.com/risp
 International Journal of Sport Policy and Politics Page 14 of 47

1
2
3 sports and in data aggregated across sports, is due to increased specialisation, and what
4
5 proportion is due to drop-out from sport altogether. In the absence of a common unique
6
7 participant identifier across sports, this study has demonstrated a feasible methodology,
8
9 based on demographic characteristics, for approximate cross-linking of de-identified sport
10
11 registration data and thereby providing a solution to this important gap in the knowledge
12
13 informing policy development.
14
15
16
Fo
17
18
19
rP
20
21
22
23
ee

24
25
26
rR

27
28
29
ev

30
31
32
iew

33
34
35
36
37
On

38
39
40
41
ly

42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58 14
59
60 URL: http://mc.manuscriptcentral.com/risp
Page 15 of 47 International Journal of Sport Policy and Politics

1
2
3 Funding details
4
5 Not applicable
6
7
8
9 Disclosure statement
10
11 The authors have no conflicts of interest
12
13
14
15 Ethics approval and consent to participate
16
Fo
17 Ethics approval was granted by the Federation University, Australia Human Ethics
18
19 Committee. Project number: C13-007.
rP
20
21 This study involves secondary data analysis of de-identified data collected of sports
22
23
ee

participants collected by state sporting associations.

24
25
26
rR

27
28
29
ev

30
31
32
iew

33
34
35
36
37
On

38
39
40
41
ly

42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58 15
59
60 URL: http://mc.manuscriptcentral.com/risp
 International Journal of Sport Policy and Politics Page 16 of 47

1
2
3 References
4
5
Australian Sports Commission (2015). Play. Sport. Australia. The Australian Sports
6 Commission's participation game plan. Canberra, Australia, Australian Sports Commission:
7 26.
8
9 Australian Sports Commission (2016). AusPlay participation data for the sport sector.
10 Summary of key national findings. Canberra, Australian Sports Commission.
11
12 Australian Sports Commission. (not dated). "TThe importance of sport sampling before
13 specialising." Retrieved 30th July, 2018, from
14 https://www.ausport.gov.au/supporting/tips/tip_5.
15
16 Balish, S., C. McLaren, D. Rainham and C. Blanchard (2014). "Correlates of youth sport
Fo
17 attrition: A review and future directions." Psychology of Sport and Exercise 15: 429-439.
18
19 Breuer, C., R. Hoekman, S. Nagel and H. van der Werff, Eds. (2015). Sport clubs in Europe.
rP
20 A cross-national comparative perspective. Sports economics, management and policy.
21 London, Springer.
22
23
ee
Bridge, M. and M. Toms (2013). "The specialising or samping debate: a retrospective
24 analysis of adolescent sports participation in the UK." Journal of Sports Science 31(1): 87-
25 96.
26
rR

27 Casey, M., R. Eime, J. Harvey, N. Sawyer, M. Craike, C. Symons and W. Payne (2017).
28 "The influence of a Healthy Welcoming Environment on participation in club sport by
29 adolescent girls: a longitudinal study." BMC Sports Science, Medicine and Rehabilitation
ev

30 9(12).
31
32 Coté, J. and M. Vierimaa (2014). "The developmental model of sport particiation: 15 years
iew

33 after its first conceptualization." Review of Science and Sports 29 Supplement: S63-69.
34
35 Craike, M., C. Symons, R. Eime, W. Payne and J. Harvey (2011). "A comparative study of
36 factors influencing participation in sport and physical activity for metropolitan and rural
37 female adolescents." Annals of Leisure Research 14(4): 355-368.
On

38
39 Crane, J. and V. Temple (2015). "A systematic review of dropout from organized sport
40 among children and youth." European Physical Education Review 21(1): 114-131.
41
Delorme, N., A. Chalabaev and M. Raspaud (2011). "Relative age is associated with sport
ly

42
43
dropout: evidence from youth categories of French basketball." Scandinavian Journal of
Medicine & Science in Sports 21(1): 120-128.
44
45
Eime, R., J. Harvey and M. Charity (2016). Sport participation in Victoria, 2015.
46
Melbourne, VicHealth.
47
48 Eime, R., J. Harvey, M. Charity, M. Casey, H. Westerbeek and W. Payne (2016). "Age
49 profiles of sport participants." BMC Sports Science, Medicine and Rehabilitation 8(6).
50
51 Eime, R., J. Harvey, M. Charity and W. Payne (2016). "Population levels of sport
52 participation: implications for sport policy." BMC Public Health 16(1): 1-8.
53
54 Eime, R., W. Payne and J. Harvey (2009). "Trends in organised sport membership: Impact
55 on sustainability." Journal of Science and Medicine in Sport 12(1): 123-129.
56
57
58 16
59
60 URL: http://mc.manuscriptcentral.com/risp
Page 17 of 47 International Journal of Sport Policy and Politics

1
2
3 Eime, R., N. Sawyer, J. Harvey, M. Casey, H. Westerbeek and W. Payne (2015).
4 "Integrating Public Health and Sport Management: Sport participation trends 2001-2010."
5 Sport Management Review 18(2): 207-217.
6
7 Eime, R., J. Young, J. Harvey, M. Charity and W. Payne (2013). "A systematic review of the
8 psychological and social benefits of participation in sport for children and adolescents:
9 informing development of a conceptual model of health through sport." International Journal
10 of Behavioral Nutrition & Physical Activity 10(98).
11
12 Eime, R. M., M. M. Casey, J. T. Harvey, M. J. Charity, J. A. Young and W. R. Payne
13 (2015). "Participation in modified sports programs: a longitudinal study of children's
14 transition to club sport competition." BMC Public Health 15(649): 7.
15
16 Eime, R. M., M. M. Casey, J. T. Harvey, N. A. Sawyer, C. M. Symons and W. R. Payne
Fo
17
(2015). "Socioecological factors potentially associated with participation in physical activity
18
and sport: A longitudinal study of adolescent girls." Journal of Science and Medicine in
Sport 18(6): 684-690.
19
rP
20
Eime, R. M., M. J. Charity, J. T. Harvey and W. R. Payne (2015). "Participation in sport and
21
physical activity: associations with socio-economic status and geographical remoteness."
22
BMC public health 15(434).
23
ee

24 Eime, R. M., J. T. Harvey, M. J. Craike, C. M. Symons and W. R. Payne (2013). "Family

25 support and ease of access link socio-economic status and sports club membership in
26
rR

adolescent girls: a mediation study." Int J Behav Nutr Phys Act 10(50).
27
28 Fransen, J., J. Pion, J. Vandendriessche, B. Vandorpe, R. Vaeyens, M. Lenoir and R.
29 Philippaerts (2012). "Differences in physical fitness and gross motor coordination in boys
ev

30 aged 6-12 years specializing in one versus sampling more than one sport." Journal of Sport
31 Sciences 30(4): 379-386.
32
iew

33 Fraser-Thomas, J., J. Coté and J. Deakin (2008). "Examining adolescent sport dropout and
34 prolonged engagement from a developmental perspective." Journal of applied sport
35 psychology 20(3): 318-333.
36
37 Gardner, L., C. Magee and S. Vella (2017). "Enjoyment and behavioral intention predict
On

38 organized youth sport participation." Journal of Physical Activity and Health Ahead of
39 Print.
40
41 Harris, S., K. Mori and M. Collins (2009). "Great Expectations: Voluntary Sports Clubs and
ly

42 Their Role in Delivering National Policy for English Sport." VOLUNTAS: International
43 Journal of Voluntary and Nonprofit Organizations 20(4): 405.
44
45 Harris, S., G. Nichols and M. Taylor (2017). "Bowling even more along: trends towards
46 individual participation in sport." European Sport Management Quarterly 17(3): 290-311.
47
48 Harvey, J. T., M. J. Charity, N. A. Sawyer and R. M. Eime (2018). "Non-response bias in
49 estimates of prevalence of club-based sport participation from an Australian national
50 physical activity, recreation and sport survey." BMC Public Health 18(1): 895.
51
52 Henrique, R., A. Re, D. Stodden, J. Fransen, C. Campos, D. Querioz and M. Cattuzzo
53 (2016). "Association between sports participation, motor competence and weight status: A
54 longitudinal study." Journal of Science and Medicine in Sport 9(10): 825-829.
55
56
57
58 17
59
60 URL: http://mc.manuscriptcentral.com/risp
 International Journal of Sport Policy and Politics Page 18 of 47

1
2
3 Hoekman, R., K. Breedveld and G. Kraaykamp (2017). "Sport participation and the social
4 and physical environment: explaining differences between urban and rural areas in the
5 Netherlands." Leisure Studies 36(3): 357-370.
6
7 Murphy, M., D. Rowe and C. Woods (2016). "Sports participation in youth as a predictor of
8 physical activity: A 5-year longitudinal study." Journal of Physical Activity and Health 13:
9 704-711.
10
11 Myer, G. D., N. Jayanthi, J. P. Difiori, A. D. Faigenbaum, A. W. Kiefer, D. Logerstedt and
12 L. J. Micheli (2015). "Sport Specialization, Part I Does Early Sports Specialization Increase
13 Negative Outcomes and Reduce the Opportunity for Success in Young Athletes?" Sports
14 Health 7(5): 437-442.
15
16 Robinson, L. and J. Goodway (2009). "Instructional climates in preschool children who are
Fo
17
at-risk. Part 1: object-control skill development." Research Quarterly for Exercise and Sport
18
80(3): 533-542.
19
Telford, R. M., R. D. Telford, T. Cochrane, R. B. Cunningham, L. S. Olive and R. Davey
rP
20
(2015). "The influence of sport club participation on physical activity, fitness and body fat
21
during childhood and adolescence: the LOOK Longitudinal Study." Journal of Science and
22
Medicine in Sport 19(5): 400-406.
23
ee

24 Veldman, S., K. Palmer, A. Okely and L. Robinson (2017). "Promoting ball skills in
25 preschool-age girls." Journal of Science and Medicine in Sport 20(1): 50-54.
26
rR

27 VicHealth. (2018). "Physical activity strategy 2018-2023." Retrieved March 9th, 2018,
28 from
29 file:///C:/Users/reime/Downloads/VicHealth%20Physical%20Activity%20Strategy%20(2).p
ev

30 df.
31
32 Wheeler, S. (2012). "The significance of family culture for sports participation."
iew

33 International Review for the Sociology of Sport 47(2): 235-252.

34
35 Wong, H., T. Olds, R. Maddison and L. Foley (2016). "A great sporting nation? Sport
36 participation in New Zealand youth." The New Zealand Medical Student Journal 23: 18-23.
37
On

38
39
40
41
ly

42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58 18
59
60 URL: http://mc.manuscriptcentral.com/risp
Page 19 of 47 International Journal of Sport Policy and Politics

1
2
3
4
5 Table 1: Age-specific player numbers by player count estimate
6
7
Age
8

Fo
9 range
10
11

rP
Counts of: Region 4 5-9 10-14 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70+ Total
12
13 Registrations Victoria 24,303 243,264 222,265 103,026 60,982 43,807 32,526 25,112 25,575 21,834 18,894 16,631 17,169 18,883 32,879 907,150

ee
14
15 Players 2 Victoria 20,396 182,393 169,974 88,464 54,725 40,583 30,402 23,844 24,380 20,866 18,218 16,139 16,693 18,373 32,114 757,564
16

rR
17 Players 1 Victoria 18,905 165,022 155,845 83,807 52,874 39,535 29,790 23,432 23,966 20,536 17,978 15,942 16,498 18,132 31,792 714,054
18

ev
19
20

iew
21
22
23
24

On
25
26
27

ly
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45 URL: http://mc.manuscriptcentral.com/risp
46
47
 International Journal of Sport Policy and Politics Page 20 of 47

1
2
3
4
5 Table 2: Age-specific player numbers by player count estimate and region
6
7
Age range
8

Fo
9
Counts of: Region 4 5-9 10-14 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70+ Total
10
11

rP
Registrations Metropolitan 17,757 166,930 153,977 68,667 41,685 30,251 22,395 16,853 17,828 15,688 13,494 11,439 11,104 11,780 19,271 619,119
12
13 Players 2 Metropolitan 14,893 125,487 117,788 59,542 37,651 28,371 21,075 16,158 17,073 15,006 13,000 11,078 10,783 11,441 18,762 518,108

ee
14
15 Players 1 Metropolitan 13,791 113,546 107,738 56,042 36,183 27,532 20,588 15,858 16,758 14,751 12,806 10,943 10,663 11,285 18,567 487,051
16

rR
17 Registrations Non-metropolitan 6,546 76,334 68,288 34,359 19,297 13,556 10,131 8,259 7,747 6,146 5,400 5,192 6,065 7,103 13,608 288,031
18

ev
Players 2 Non-metropolitan 5,503 56,906 52,186 28,922 17,074 12,212 9,327 7,686 7,307 5,860 5,218 5,061 5,910 6,932 13,352 239,456
19
20 Players 1 Non-metropolitan 5,114 51,476 48,107 27,765 16,691 12,003 9,202 7,574 7,208 5,785 5,172 4,999 5,835 6,847 13,225 227,003

iew
21
22
23
24

On
25
26
27

ly
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45 URL: http://mc.manuscriptcentral.com/risp
46
47
Page 21 of 47 International Journal of Sport Policy and Politics

1
2
3
4
5
6 Table 3: Age-specific player numbers by player count estimate and sex
7
8 Age range

Fo
9
10 Counts of: Sex 4 5-9 10-14 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70+ Total
11

rP
Registrations Male 15,837 152,170 136,768 67,657 44,053 32,388 24,011 17,571 17,611 15,401 13,553 11,913 11,904 12,998 22,678 596,513
12
13 Players 2 Male 12,699 107,872 98,722 56,403 38,714 29,669 22,249 16,601 16,696 14,631 13,014 11,526 11,561 12,606 22,088 485,051

ee
14 Players 1 Male 11,757 96,567 89,822 53,179 37,281 28,882 21,786 16,325 16,421 14,416 12,826 11,380 11,419 12,428 21,848 456,337
15
Registrations Female 8,466 91,094 85,497 35,369 16,929 11,419 8,515 7,541 7,964 6,433 5,341 4,718 5,265 5,885 10,201 310,637
16

rR
17 Players 2 Female 7,697 74,521 71,252 32,061 16,011 10,914 8,153 7,243 7,684 6,235 5,204 4,613 5,132 5,767 10,026 272,513
18

ev
Players 1 Female 7,148 68,455 66,023 30,628 15,593 10,653 8,004 7,107 7,545 6,120 5,152 4,562 5,079 5,704 9,944 257,717
19
20

iew
21
22
23
24

On
25
26
27

ly
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45 URL: http://mc.manuscriptcentral.com/risp
46
47
 International Journal of Sport Policy and Politics Page 22 of 47

1
2
3
4
5 250,000 70
6

Participation rate (% of age cohort)

7 60
200,000
8 50

Fo
9
150,000
Number

10 40
11

rP
Registrations Registrations
100,000 30
12 Players 2 Players 2
13 20

ee
50,000 Players 1 Players 1
14 10
15
16 0 0

rR
4
5-9
10-14
15-19
20-24
25-29
30-34
35-39
40-44
45-49
50-54
55-59
60-64
65-69
70+

4
5-9
10-14
15-19
20-24
25-29
30-34
35-39
40-44
45-49
50-54
55-59
60-64
65-69
70+
17
18

ev
19 Age range Age range
20
a) Number of players

iew
21 b) Participation rate
22 Figure 1.: Age-specific pParticipation numbers and rates: by age and player count estimateestimation method
23
24

On
25
26
27

ly
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45 URL: http://mc.manuscriptcentral.com/risp
46
47
Page 23 of 47 International Journal of Sport Policy and Politics

1
2
3
4
5 90 90
6

Participation rate (% of age cohort)

Participation rate (% of age cohort)

80 80
7
8 70 70

Fo
9 60 60
10 50 50
11

rP
40 Registrations 40 Registrations
12 30 Players 2 30 Players 2
13
20

ee
Players 1 20 Players 1
14
10 10
15
16 0 0

rR
4
5-9
10-14
15-19
20-24
25-29
30-34
35-39
40-44
45-49
50-54
55-59
60-64
65-69
70+

70+
4
5-9
10-14
15-19
20-24
25-29
30-34
35-39
40-44
45-49
50-54
55-59
60-64
65-69
17
18

ev
19 Age range Age range
20
a) Metropolitan region

iew
21 b) Non-metropolitan region
22 Figure 2. Participation rates: by region, age and estimation method
23
24

On
25
26
27

ly
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45 URL: http://mc.manuscriptcentral.com/risp
46
47
 International Journal of Sport Policy and Politics Page 24 of 47

1
2
3
4
5 90 90
6

Participation rate (% per age cohort)

Participation rate (% per age cohort)

80 80
7
8 70 70

Fo
9 60 60
10 50 50
11

rP
40 Registrations 40 Registrations
12 30 Players 2 30 Players 2
13
20 20

ee
Players 1 Players 1
14
10 10
15
16 0 0

rR
4
5-9
10-14
15-19
20-24
25-29
30-34
35-39
40-44
45-49
50-54
55-59
60-64
65-69

4
5-9
10-14
15-19
20-24
25-29
30-34
35-39
40-44
45-49
50-54
55-59
60-64
65-69
70+

70+
17
18

ev
19 Age range Age range
20
a) Male

iew
21 b) Female
22 Figure 3. Participation rates: by sex, age and estimation method
23
24

On
25
26
27

ly
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45 URL: http://mc.manuscriptcentral.com/risp
46
47
Page 25 of 47 International Journal of Sport Policy and Politics

1
2
3 Sport drop-out during adolescence: Is it real, or an artefact of
4
5 sampling behaviour?
6
7
8
9
10 R. M Eime.,a,b* J. T Harvey a,b and M. J Charity a,b
11
12 a
13 Faculty of HealthSchool of Health and Life Sciences, Federation University
14
Australia; bInstitute for Health and Sport, Victoria University.
15
16
Fo
17 Provide full correspondence details here including e-mail for the *corresponding
18
19 author
rP
20
21 Associate Professor Rochelle Eime
22
23
ee

24
[email protected]
25
26 Faculty of HealthSchool of Health and Life Science, Federation University Australia
rR

27
28
PO Box 663, Ballarat, Victoria, Australia, 3353. Ph: +61 3 5327 9687
29
ev

30
31 Institute for Health and Sport, Victoria University, Australia
32
iew

33
34 Dr Jack Harvey
35
36 [email protected]
37
On

38 School of Health and Life ScienceFaculty of Health, Federation University Australia

39
40
41 PO Box 663, Ballarat, Victoria, Australia, 3353. Ph: +61 3 5327 9065
ly

42
43 Institute for Health and Sport, Victoria University, Australia
44
45
46 Ms Melanie Charity
47
48 [email protected]
49
50 School of Health and Life ScienceFaculty of Health, Federation University Australia
51
52
53 PO Box 663, Ballarat, Victoria, Australia, 3353. Ph: +61 3 5327 6211
54
55 Institute for Health and Sport, Victoria University, Australia
56
57
58 1
59
60 URL: http://mc.manuscriptcentral.com/risp
 International Journal of Sport Policy and Politics Page 26 of 47

1
2
3 Sport drop-out during adolescence: Is it real, or an artefact of
4
5 sampling behaviour?
6
7
8
9
10 Abstract
11
12 Understanding sport participation and drop-out is important for sport management. Sport
13
14 policy focuses on increasing participation overall, and not specifically on retention. It is
15
16 well-known that Mmany children sample or play multiple sports and then specialise.
Fo
17
18 However, quantifying these behaviors is challenging. Sport registration databases are
19
rP
20 potentially useful for this purpose. However, given privacy and data security issues,
21
22 identification and direct linking of data records of individual participants across sports is not
23
ee

24 possible. This study demonstrates a feasible methodology for approximate cross-linking of

25
26 de-identified data and thereby quantifying the extent of sampling behaviour, and hence
rR

27
28 investigateding whether or not, and to what degree, the decline in community club-based
29
ev

30 sport participation observed during adolescence is attributable to a “sampling to

31
32 specialisation” effect as opposed to drop-out from sport altogether. Participants were
iew

33
34 registered members of one of eleven state sporting associations in 2015. Individuals may
35
36 play more than one sport, and therefore be registered with more than one sport in any given
37
year. For this analysis, data (907,150 player records) for the eleven sports were
On

38
39
40 amalgamated, and registered Victorian players were categorised by gendersex, age group
41
and residential postcode. In the absence of a common unique identifier across sports,
ly

42
43
44
nNumbers of individual players were estimated using demographic matching, to allow
45
comparison of comparing numbers of registrations and numbers of individual participants
46
47
across age, gendersex and region. Results showed that the effect of participants playing
48
49
multiple sports is highest for ages 5-14, and then participants tend to play fewer sports. from
50
51
ages 15-29. Nevertheless, while this effect contributes to the reduction in registrations, it
52
53
does not completely explain it. This study confirms that the drop-off in community sport
54
55
participation during adolescence is real and not simply an artefact of sampling behaviour
56
57
58 2
59
60 URL: http://mc.manuscriptcentral.com/risp
Page 27 of 47 International Journal of Sport Policy and Politics

1
2
3 among younger participants. It is recommended that national sport policy focuses on overall
4
5 participation across sports, taking into account the sampling and specialising phenomena
6
7 which naturally occur during childhood and adolescence, rather than merely asking
8
9 individual sports to increase participation numbers.
10
11
12
13
14 Key words: sport, participation, drop-out, sampling, specialising
15
16
Fo
17
18
19
rP
20
21
22
23
ee

24
25
26
rR

27
28
29
ev

30
31
32
iew

33
34
35
36
37
On

38
39
40
41
ly

42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58 3
59
60 URL: http://mc.manuscriptcentral.com/risp
 International Journal of Sport Policy and Politics Page 28 of 47

1
2
3 Introduction
4
5 It is well known that participation in sport is popular among children, and that sport
6
7 participation rates peak during childhood to early adolescence (Australian Sports
8
9 Commission 2016, Eime, Harvey, Charity and Payne 2016, Wong, Olds et al. 2016). A
10
11 recent survey of children aged 0-14 years reported sport participation peaking at ages 9-11
12
13 (Australian Sports Commission 2016), and other studies of sport participants have reported
14
15 peak participation at ages 10-14 years (Eime, Harvey, Charity and Payne 2016, Wong, Olds
16
Fo
17 et al. 2016). Recent research reports that nearly a third of sports participants are aged 10-14
18
19 years (Eime, Harvey, Charity, Casey et al. 2016). However many people drop out of sport
rP
20
21
during childhood and adolescence, especially females (Australian Sports Commission 2016,
22
Eime, Harvey, Charity, Casey et al. 2016, Eime, Harvey, Charity and Payne 2016, Wong,
23
ee

24
Olds et al. 2016) . The sport participation decline is further evident throughout the lifespan,
25
26
rR

when there is evidence of a shift away from competitive club-based sport towards non-
27
28
competitive and non-organised forms of leisure-time physical activity (Eime, Sawyer et al.
29
ev

30
2015, Australian Sports Commission 2016, Eime, Harvey, Charity and Payne 2016, Harris,
31
32
iew

Nichols et al. 2017).

33
34
35 In a large recent longitudinal study of over 200,00 children aged 4-12 it was found that most
36
37 children in a modified sports (introductory) program in the first year of the study did not
On

38
39 continue playing that particular sport for throughout a 4-year period (Eime, Casey, Harvey,
40
41 Charity et al. 2015). Furthermore, two-thirds of those children who dropped out of the sport
ly

42
43 did so after the first year; it was conjectured that many of these had only a single year or
44
45 season of participation in that particular sport (Eime, Casey, Harvey, Charity et al. 2015).
46
47
48 Whilst there are a range of intrapersonal, interpersonal, organisational, and environmental
49
50
and policy factors influencing participation in sport across the lifespan, competency is a key
51
determinant (Crane and Temple 2015, Eime, Casey, Harvey, Sawyer et al. 2015). If people
52
53
do not perceive themselves as having adequate skills to play, they do not then enjoy playing
54
55
sport and drop out (Crane and Temple 2015). Sport competency is in turn related to the
56
57
58 4
59
60 URL: http://mc.manuscriptcentral.com/risp
Page 29 of 47 International Journal of Sport Policy and Politics

1
2
3 development of fundamental motor skills (FMS). There has been much research around the
4
5 importance of FMS for participation in sport (Veldman, Palmer et al. 2017). FMS are
6
7 learned skills, and children do not develop FMS naturally as part of normal growth and
8
9 development. That is, they need to be taught, practiced and reinforced (Robinson and
10
11 Goodway 2009, Veldman, Palmer et al. 2017). Furthermore, FMS development is
12
13 cumulative and relatively permanent, and sport skills may have an effect on sport
14
15 participation that persists across the lifespan (Henrique, Re et al. 2016). There is clear
16
Fo
17 evidence that FMS are important for predicting continued sport participation or drop-out in
18
19 children (Henrique, Re et al. 2016). This is thought to be associated with those children with
rP
20
21 advanced FMS being more successful in sport and as a result having more enjoyment than
22
23 those children with lower FMS. Having fun or enjoyment in sport is also a major predictor
ee

24
25 of drop-out of sport participation (Crane and Temple 2015, Gardner, Magee et al. 2017).
26
rR

27 Whilst a lack of perceived competency and enjoyment may be contributing factors to the
28
29 drop-out in participation sport, it has been recently acknowledged that the drop-off in late
ev

30
31 childhood to early adolescence may be partly due to children sampling multiple sports, and
32
iew

33 then specialising in one (Eime, Casey, Harvey, Charity et al. 2015). This may be further
34
35 explained through the Developmental Model of Sports Participation (DMSP) (Coté and
36
37 Vierimaa 2014). Whilst this model is often used to describe the elite sport performance
On

38
39 pathway we believe also it is relevant here. The DMSP proposes two pathways to elite sport
40
41 performance: early diversification and early specialisation (Fransen, Pion et al. 2012, Coté
ly

42
43 and Vierimaa 2014). Within the DMSP the first stage is the sampling stage at 6-12 years,
44
45 followed by the specialisation stage at 12-15 years and the investment stage at 15+ years. At
46
47 age 6-12, sampling is considered beneficial for athletic development because of the exposure
48
49 to a number of different physical, cognitive, affective and psychosocial environments,
50
51 reinforcing physical, personal and mental skills required for future sport elite success
52
53 (Fransen, Pion et al. 2012, Coté and Vierimaa 2014). Conversely, early specialisation often
54
55 leads to dropout (Fraser-Thomas, Coté et al. 2008). Children who specialise in a single sport,
56
57
58 5
59
60 URL: http://mc.manuscriptcentral.com/risp
 International Journal of Sport Policy and Politics Page 30 of 47

1
2
3 particularly with heavy training, risk burnout, overuse injuries and actual decrease in
4
5 performance due to overtraining (Myer, Jayanthi et al. 2015). However, previous research
6
7 has focused predominantly on elite athletes. We do not have knowledge about the extent of
8
9 sampling and specialisation, or its effect in general community sport participants.
10
11 Australian sport policy focuses on increasing participation generally, with no specific
12
13 mention of sampling and specialising behaviours (Australian Sports Commission 2015). The
14
15 current policy focuses on achieving annual increases in participation numbers for individual
16
Fo
17 sports, and on talent development for children and youth (Australian Sports Commission
18
19 2015). The authors contend that this strategic focus influences sports to prioritise recruitment
rP
20
21 over retention, including the development of organised modified sports programs to very
22
23
ee

young participants, in some sports as young as four years of age (Eime, Harvey, Charity and
24
25 Payne 2016). The authors have recommended a greater focus on retention strategies,
26
rR

27 specifically in relation to the sharp decreases in sport participation during adolescence

28
29 (Eime, Harvey, Charity and Payne 2016).
ev

30
31
32 Sports organisations in Australia tend to work individually, each through their own network
iew

33
34 of state-based organisations and local sport clubs, with an absence of multi-sport clubs
35
36 which are more evident in Europe (Breuer, Hoekman et al. 2015). It is conjectured that
37
On

38 multi-sport clubs may make the transition of participants across sports easier than with
39
40 individual sport clubs. It has also been reported that in European countries with larger multi-
41
ly

42 sports clubs, the sport sector is able to be more innovative and to drive public policy
43
44 programs more effectively than individual sport clubs (Harris, Mori et al. 2009).
45
46 In Australia, community sports clubs are most often run by local volunteers (Eime, Payne et
47
48 al. 2009). In the Australian state of Victoria, some sports are funded by the Victorian Health
49
50 Promotion Foundation to focus not just on traditional competitive sport participation, but
51
52 also on the development of more social and recreational forms of sport participation, in an
53
54 effort to get more people active through sport (VicHealth 2018).
55
56
57
58 6
59
60 URL: http://mc.manuscriptcentral.com/risp
Page 31 of 47 International Journal of Sport Policy and Politics

1
2
3 It is well-known that many children sample or play multiple sports and then specialise.
4
5 Sampling may involve playing a sequence of different sports or playing multiple sports in
6
7 the same year. Some of the decline we see in sport during childhood and adolescence may
8
9 be due to a decline in the number of sports played in a particular year and not necessarily a
10
11 true drop-out from sport participation (Eime, Casey, Harvey, Charity et al. 2015, Eime,
12
13 Harvey, Charity, Casey et al. 2016). However, quantifying these behaviors is a challenging
14
15 task. Population-wide sample surveys can provide direct evidence of the participation
16
Fo
17 patterns of individuals, but sample sizes are determined from a national perspective, and so
18
19 the capacity to produce estimates with acceptable levels of sampling error for smaller
rP
20
21 geographical areas or population segments is limited. Also, historically the scope of such
22
23 surveys in Australia has generally been limited to persons 15+ years of age (Eime, Sawyer et
ee

24
25 al. 2015). Furthermore, non-response and refusal rates are high, and recent research has
26
rR

27 suggested that, in addition to sampling error, estimates of participation counts and rates from
28
29 such surveys are subject to substantial non-response bias (Harvey, Charity et al. 2018).
ev

30
31 Sport registration data are prima facie much less susceptible to bias, and as they purport to
32
iew

33
be complete enumerations of participants in each sport, are not subject to sampling error
34
35
(Harvey, Charity et al. 2018). When registration data from multiple sports are
36
amalgamated, the aggregated counts are inflated by individuals playing multiple sports in the
37
On

38
same year.
39
40
41
ly

42
43
44 Ideally, to quantify this phenomenon would require the capacity to identify individual
45
46 participants across multiple sports. However, particularly given current awareness of
47
48 privacy and data security issues, identification and direct linking of data records of
49
50 individual participants across sports is not possible. The aim of this study was to
51
52 demonstrate a feasible methodology, based on demographic characteristics, for approximate
53
54 cross-linking of de-identified data and thereby The aim of this study was to use sport
55
56 registration data to quantify, for each age level, the extent of participation in multiple sports
57
58 7
59
60 URL: http://mc.manuscriptcentral.com/risp
 International Journal of Sport Policy and Politics Page 32 of 47

1
2
3 in a given year, and hence to investigate whether or not, and to what degree, the decline
4
5 observed in the per capita rate of aggregated registrations in community club-based sport
6
7 participation during later childhood and adolescence is attributable to a reduction in the
8
9 extent of participation in multiple sports, i.e. a “sampling to specialisation” effect.
10
11
12 Methods
13
14 Data for this study were collected as part of the Sport and Recreation Spatial project, a
15
16 research project funded by government and public health agencies in the Australian state of
Fo
17
18 Victoria to monitor participation in sport and active recreation, for the purpose of informing
19
rP
20 policy development and program planning in the sport and recreation sector. The data and
21
22 have been previously described in detail (Eime, Harvey, Charity, Casey et al. 2016, Eime,
23
ee

24 Harvey, Charity and Payne 2016). Briefly, participants were registered members
25
26 ofparticipants in one of eleven state sporting associations (Australian football, basketball,
rR

27
28 bowls, cricket, football (soccer), golf, gymnastics, hockey, netball, sailing and tennis) in the
29
ev

30 Australian state of Victoria in 2015. Individuals may play more than one sport, and therefore
31
32 it is possible a player could be registered with more than one sport in any given year.
iew

33
34
For this analysis, data for the eleven sports were amalgamated, and registered Victorian
35
36
players aged 4-100 in 2015 were categorised by gendersex, age group and residential
37
On

38
postcode. Individuals may play more than one sport, and therefore it is possible a player
39
40
could be registered with more than one sport in any given year. Actual numbers of
41
ly

42
registrations and estimated numbers of individual players were calculated, to allow
43
44 comparison of these two indicators across age, gendersex and region.
45
46
47
48 The number of registrations in each year were direct counts of the total of all registrations
49
50 for the 11 sports. As the registrations for each sport were de-identified, it was not possible
51
52 to directly link all data records for a particularly individual. An approximate count of
53
54 individuals was generated by assuming that two registrations for different sports that were
55
56 matched on year, gendersex, date of birth and residential postcode were the same person.
57
58 8
59
60 URL: http://mc.manuscriptcentral.com/risp
Page 33 of 47 International Journal of Sport Policy and Politics

1
2
3 Hence, by excluding all but the first occurrence of a particular combination of these four
4
5 characteristics, estimated counts of individual players were obtained. This was done for each
6
7 sport separately to provide an indication of ‘multiples’ per sport, and then for the sample as a
8
9 whole.
10
11 A second player count estimate was also generated by assuming that multiple matching
12
13 registrations within a sport genuinely represented different individuals (i.e. same-sex twins
14
15 or children fortuitously matched on demographic characteristics). The ‘Multiple’
16
Fo
17 combinations of DOB, postcode and gendersex within each sport were retained, up to a
18
19 maximum of five occurrences per sport (arbitrarily chosen in order to exclude blocks of
rP
20
21 spurious matches due to data quality issues). For example, two players in Sport A and four
22
23 players in Sport C all with the same details would result in a registration count of six, and
ee

24
25 individual counts of one under the first option (referred to as Players 1) and four under the
26
rR

27 second option (Players 2). The Players 1 count is always less than or equal to the Players 2
28
29 count, which in turn is always less than or equal to the registration count.
ev

30
31
32 Results
iew

33
34 There were a total of 907,150 registrations, 757,564 unique individuals assuming “valid”
35
36 multiple individuals within sports as described (Players 2) and 714,054 unique individuals
37
On

38 assuming no valid multiple individuals (Players 1). Aggregated across all years and sports,
39
40 the Players 2 count was 16.5% lower, and the Player 1 count was 21.3% lower, than the
41
ly

42 count of registrations. These are indicative of the proportion of sports participants who
43
44 played multiple sports in 2015.
45
46 Figure 1a presents the number of registrations and estimated numbers of individual players
47
48 (under the two exclusion rules) across the lifespan, and Figure 1b shows the corresponding
49
50 participation rates per the Victorian population in standard age cohorts. The pattern within
51
52 both figures is the same. The key result is that the excess of registrations over individual
53
54 players declines with increasing age. That is, the effect of participants playing multiple
55
56 sports is highest for ages 5-14, and then we see a clear specialisation effect where sports
57
58 9
59
60 URL: http://mc.manuscriptcentral.com/risp
 International Journal of Sport Policy and Politics Page 34 of 47

1
2
3 participants play one or fewer sports as they age through 15-29. From age 30 the player and
4
5 registration rates are almost identical, indicating that these older sports participants are
6
7 mostly playing one of the sports only.
8
9
Insert Figure 1 and Table 1 about here
10
11
12 Figure 2a and 2b present the participation rates for metropolitan and non-metropolitan
13
14 region. Whilst the participation rate across the lifespan is generally higher in the non-
15
16 metropolitan region compared to the metropolitan region, the proportional difference
Fo
17
18 between registrations and actual players is very similar.
19
rP
20
21 Insert Figure 2 and Table 2 about here
22
23
ee

Figure 3a and 3b present the participation rates for males and females. The overall
24
25 participation rate is much higher for males than females, and peaking at a registration rate of
26
rR

27 81% for males aged 5-9 years and for females of 52% for ages 10-14 years. This
28
29 corresponds to an actual player participation rate of 51% for males aged 5-14 and 51-52%
ev

30
31 for females aged 5-14 years. Therefore, whilst the actual player participation rate is very
32
iew

33 similar for these children for both males and females, males are much more likely than
34
35 females to play multiple sports during childhood and early adolescence. That is, the
36
37 sampling effect is much greater for males than for females.
On

38
39
40 Insert Figure 3 and Table 3 about here
41
ly

42
43
44
45
46 Discussion
47
The existence of the sampling effect, whereby many younger sport participants play multiple
48
49
sports in the same year, has often been reported (Fraser-Thomas, Coté et al. 2008, Delorme,
50
51
Chalabaev et al. 2011), but this is the first study to explore how this actually relates to player
52
53
numbers and participation rates. This study is unique in quantifying the extent to which
54
55
younger sports participants sample multiple sports in a single year, identifying the age range
56
57
58 10
59
60 URL: http://mc.manuscriptcentral.com/risp
Page 35 of 47 International Journal of Sport Policy and Politics

1
2
3 over which this behaviour diminishes with increasing age, and demonstrating differences
4
5 between these patterns for males and females and between metropolitan and non-
6
7 metropolitan sport participants. Furthermore, the majority of research on sampling and
8
9 drop-out has been on elite or sub -elite athletes (Fraser-Thomas, Coté et al. 2008, Delorme,
10
11 Chalabaev et al. 2011, Bridge and Toms 2013), whereas this study examined a population of
12
13 community-based players.
14
15 We have previously speculated that the considerable drop-off in registrations during
16
Fo
17 adolescence, from 15-19 years, may could conceivably be due to the sampling effect, and
18
19 not a truly significantly indicative issue in terms of dropping out of sport altogether‘all
rP
20
21 sport’ (Eime, Harvey and Charity 2016, Eime, Harvey, Charity and Payne 2016). We
22
23
ee

conjectured that the considerable decline during adolescence might be attributable to

24
25 children playing multiple sports when younger, and thereby being counted multiply in
26
rR

27 aggregated multi-sport participation data, and then during adolescence choosing one sport to
28
29 specialise in during adolescence (Eime, Harvey and Charity 2016). The results of the present
ev

30
31 study show that while sampling does amplify the magnitude of the reduction in total
32
iew

33 registrations throughout the adolescent and early adult years, it explains only about 25% of
34
35 the drop-off in the overall registration rate, 40% of the drop-off among males, 20% of the
36
37 drop-off among females, and 25% for each of metropolitan and non-metropolitan residents.
On

38
39
40 When examining the estimated player numbers and participation rates only, and not the
41
ly

42 registrations, it is clear that there is a significant decline in sport participation from age 15-
43
44 19. Player participation rates peak at 45-46% ages 5-14 before a drop to only 23% from 15-
45
46 19 years. The participation rate halves from the 10-14 age group (46%) to 23% (15-19%)
47
48 and then halves again for 20-24 year-olds (12%).
49
50 This current study highlights that the drop-off in sport participation during adolescence and
51
52 into early adulthood is real, and is not simply the effect of sampling. Sampling itself confers
53
54 a range of benefits, contributing to a range of diverse skills and to improved fitness and
55
56 motor coordination (Fransen, Pion et al. 2012, Myer, Jayanthi et al. 2015). Fransen et al.
57
58 11
59
60 URL: http://mc.manuscriptcentral.com/risp
 International Journal of Sport Policy and Politics Page 36 of 47

1
2
3 showed that for boys, those participating in more than one sport were exposed to a greater
4
5 number of physical, cognitive, affective, and psycho-social environments than boys playing
6
7 a single sport only (Fransen, Pion et al. 2012), and concluded that it is important to be
8
9 sampling and playing multiple sports before the age of 12 rather than early specialisation
10
11 (Fransen, Pion et al. 2012). There is evidence that a higher degree of sampling for ages 11-
12
13 15 is significantly associated with increased performance, that is, competing at a national
14
15 compared to community club level (Bridge and Toms 2013 4724). We know that
16
Fo
17 participation in sport during childhood and adolescence has a lasting positive effect on
18
19 physical activity (Murphy, Rowe et al. 2016). Furthermore, sports participants are often
rP
20
21 more active and fitter than participants in non-sport physical activity; however these
22
23 associated benefits of sport participation can diminish during adolescence and especially for
ee

24
25 girls (Telford, Telford et al. 2015). Therefore strategies are needed to keep children and
26
rR

27 adolescents engaged in sport for health benefits, not only physical but also psychological
28
29 and social health (Eime, Young et al. 2013).
ev

30
31 The sampling and participation rate patterns were broadly similar for both the metropolitan
32
iew

33 and non-metropolitan regions, and this is consistent with other research (Eime, Harvey,
34
35 Charity and Payne 2016, Hoekman, Breedveld et al. 2017), however the decline is not as
36
37 pronounced for the non-metropolitan regions. A higher proportion of people living in non-
On

38
39 metropolitan regions remain active through sport during adolescence. This may relate to the
40
41 culture of community sport in rural and regional communities (Eime, Charity et al. 2015).
ly

42
43 Further, as adolescents age and become more autonomous in their decisions regarding
44
45 physical activity there are also more opportunities to participate in a wider range of leisure
46
47 activities in metropolitan compared to non-metropolitan regions (Craike, Symons et al.
48
49 2011).
50
51
52 A concerning pattern is the more severe decline during adolescence for females compared to
53
54 males. The participation rate for females for ages 15-19 is less than half of that for those
55
56 aged 1-14 years. It is also very clear from the registration numbers that young males sample
57
58 12
59
60 URL: http://mc.manuscriptcentral.com/risp
Page 37 of 47 International Journal of Sport Policy and Politics

1
2
3 or play more sports than females. Firstly, there is evidence of a competency difference
4
5 between young females compared to males, and we know that competency is a major factor
6
7 relating to participation in sport (Veldman, Palmer et al. 2017). There is evidence that young
8
9 girls are less competent at ball skills than young boys (Veldman, Palmer et al. 2017).
10
11 Further, other research amongst 6-12 year olds reports that playing multiple sports and
12
13 spending more time playing sport contributes to improved gross motor coordination
14
15 (Fransen, Pion et al. 2012). Quite simply, by sampling multiple sports children are exposed
16
Fo
17 to a greater number of physical, cognitive, affect and psycho-social environments than those
18
19 playing only one sport (Fransen, Pion et al. 2012). Secondly these gender differences may
rP
20
21 also be related to increased opportunities for boys to play multiple sports, relative to girls,
22
23 and to parents encouraging boys to play sport more than girls (Wheeler 2012, Eime, Harvey,
ee

24
25 Charity, Casey et al. 2016). Further to this, in the case of this study higher participation for
26
rR

27 males compared to females also relates to the traditional gender bias of Australian club-
28
29 based sport, whereby for several sports, until very recently, females were not able to play in
ev

30
31 club-based competitions (Eime, Harvey, Charity and Payne 2016).
32
iew

33 The main correlates of youth sport attrition include competency and social factors such as
34
35 enjoyment, support from parents, peers and coaches and a positive social club environment
36
37 (Eime, Harvey et al. 2013, Balish, McLaren et al. 2014, Henrique, Re et al. 2016, Casey,
On

38
39 Eime et al. 2017, Gardner, Magee et al. 2017). Higher sport competency in young childhood
40
41 is associated with continued sport participation across childhood (Henrique, Re et al. 2016).
ly

42
43 Competency is a cumulative and relatively permanent phenomenon, which is in contrast to
44
45 physical activity in general (Henrique, Re et al. 2016). Children exposed to more sports
46
47 earlier in their life tend to have greater competency and therefore are more likely to continue
48
49 to participate. Further to this, there is evidence that players who drop out participated in
50
51 fewer sports than those that continue playing (Fraser-Thomas, Coté et al. 2008).
52
53
54 There is much research evidence that sampling is positive for skill development in children
55
56 and young people, and for continued sport participation. In Australian sport policy there is
57
58 13
59
60 URL: http://mc.manuscriptcentral.com/risp
 International Journal of Sport Policy and Politics Page 38 of 47

1
2
3 no specific mention of sport sampling, with the focus being the broader aim of getting more
4
5 Australians and particular young Australians participating in sport more often, and in general
6
7 trying to gain insights into participation trends. (Australian Sports Commission 2015).
8
9 There is mention of the benefits of sampling for elite athlete development (Australian Sports
10
11 Commission not dated), but nothing related to the general population. A major focus of the
12
13 national sport policy is for each sport organisations is to increase participant numbers each
14
15 year. With this “individual silos” focus, the issue of sampling versus specialisation in
16
Fo
17 children and adolescents, and the implications for overall levels of participation, are not
18
19 considered.
rP
20
21 Beyond the quantification of sampling and specialisation demonstrated in this study, there
22
23
ee

remain important questions. Whilst we have many insights relating to sampling among elite
24
25 and sub-elite youth, we do not have evidence from community club-based sport as to
26
rR

27 whether sampling leads to a greater likelihood of playing sport throughout life. Are those
28
29 the people that maintain participation in sport, albeit only a low proportion, those who have
ev

30
31 played and sampled many sports in childhood? Is the drop-off in sport participation
32
iew

33 primarily attributable to ‘late sports starters’ who have less sport competency than early
34
35 sport adopters? Besides the modified sports programs for young participants, what are the
36
37 entry points to competitive sport throughout the lifespan for individuals who do not have
On

38
39 specific sports competency? These questions await further research.
40
41
ly

42
43 Conclusion
44 In conclusion, this study confirms that the drop-off during adolescence in community sport
45
46 participation, as measured by aggregated sport registrations, during adolescence is real and
47
48 not simply an artefacta consequence of a reduction in sampling behaviour compared
49
50 toamong younger participants. It is concerning that, from a peak at ages 10-14, the
51
52 participation rate halves for the next age group of 15-19 years. Considering the magnitude of
53
54 this drop-off, sSport policy should specifically prioritise retention in sport, and not merely
55
56 focus on increasing total sport numbers. This requires a longitudinal rather than cross-
57
58 14
59
60 URL: http://mc.manuscriptcentral.com/risp
Page 39 of 47 International Journal of Sport Policy and Politics

1
2
3 sectional approach to the monitoring of participation. A policy relating to retention is
4
5 needed to provide sporting organisations with the lever to make this a priority for their sport-
6
7 specific strategies.
8
9
It is also recommended that national sport policy should focus on overall participation across
10
11 sports, taking into account the sampling and specialising phenomena which naturally occur
12
13 during childhood and adolescence, rather than merely asking individual sports to increase
14
15 participation numbers. From a health perspective, as long as people are regularly physically
16
Fo
17 active, it does not matter if they initially play multiple sports and then subsequently
18
19 specialise. However, from a policy and planning perspective it is important to know what
rP
20
21 proportion of the apparent drop-off in late childhood and adolescence, both in individual
22
23
ee

sports and in data aggregated across sports, is due to increased specialisation, and what
24
25 proportion is due to drop-out from sport altogether. In the absence of a common unique
26
rR

27 participant identifier across sports, this study has demonstrated a feasible methodology,
28
29 based on demographic characteristics, for approximate cross-linking of de-identified sport
ev

30
31 registration data and thereby providing a solution to this important gap in the knowledge
32
iew

33 informing policy development.

34
35
36
37
On

38
39
40
41
ly

42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58 15
59
60 URL: http://mc.manuscriptcentral.com/risp
 International Journal of Sport Policy and Politics Page 40 of 47

1
2
3 Funding details
4
5 Not applicable
6
7
8
9 Disclosure statement
10
11 The authors have no conflicts of interest
12
13
14
15 Ethics approval and consent to participate
16
Fo
17 Ethics approval was granted by the Federation University, Australia Human Ethics
18
19 Committee. Project number: C13-007.
rP
20
21 This study involves secondary data analysis of de-identified data collected of sports
22
23
ee

participants collected by the sState sSporting aAssociations.

24
25
26
rR

27
28
29
ev

30
31
32
iew

33
34
35
36
37
On

38
39
40
41
ly

42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58 16
59
60 URL: http://mc.manuscriptcentral.com/risp
Page 41 of 47 International Journal of Sport Policy and Politics

1
2
3 References
4
5
Australian Sports Commission (2015). Play. Sport. Australia. The Australian Sports
6 Commission's participation game plan. Canberra, Australia, Australian Sports Commission:
7 26.
8
9 Australian Sports Commission (2016). AusPlay participation data for the sport sector.
10 Summary of key national findings. Canberra, Australian Sports Commission.
11
12 Australian Sports Commission. (not dated). "TThe importance of sport sampling before
13 specialising." Retrieved 30th July, 2018, from
14 https://www.ausport.gov.au/supporting/tips/tip_5.
15
16 Balish, S., C. McLaren, D. Rainham and C. Blanchard (2014). "Correlates of youth sport
Fo
17 attrition: A review and future directions." Psychology of Sport and Exercise 15: 429-439.
18
19 Breuer, C., R. Hoekman, S. Nagel and H. van der Werff, Eds. (2015). Sport clubs in Europe.
rP
20 A cross-national comparative perspective. Sports economics, management and policy.
21 London, Springer.
22
23
ee
Bridge, M. and M. Toms (2013). "The specialising or samping debate: a retrospective
24 analysis of adolescent sports participation in the UK." Journal of Sports Science 31(1): 87-
25 96.
26
rR

27 Casey, M., R. Eime, J. Harvey, N. Sawyer, M. Craike, C. Symons and W. Payne (2017).
28 "The influence of a Healthy Welcoming Environment on participation in club sport by
29 adolescent girls: a longitudinal study." BMC Sports Science, Medicine and Rehabilitation
ev

30 9(12).
31
32 Coté, J. and M. Vierimaa (2014). "The developmental model of sport particiation: 15 years
iew

33 after its first conceptualization." Review of Science and Sports 29 Supplement: S63-69.
34
35 Craike, M., C. Symons, R. Eime, W. Payne and J. Harvey (2011). "A comparative study of
36 factors influencing participation in sport and physical activity for metropolitan and rural
37 female adolescents." Annals of Leisure Research 14(4): 355-368.
On

38
39 Crane, J. and V. Temple (2015). "A systematic review of dropout from organized sport
40 among children and youth." European Physical Education Review 21(1): 114-131.
41
Delorme, N., A. Chalabaev and M. Raspaud (2011). "Relative age is associated with sport
ly

42
43
dropout: evidence from youth categories of French basketball." Scandinavian Journal of
Medicine & Science in Sports 21(1): 120-128.
44
45
Eime, R., J. Harvey and M. Charity (2016). Sport participation in Victoria, 2015.
46
Melbourne, VicHealth.
47
48 Eime, R., J. Harvey, M. Charity, M. Casey, H. Westerbeek and W. Payne (2016). "Age
49 profiles of sport participants." BMC Sports Science, Medicine and Rehabilitation 8(6).
50
51 Eime, R., J. Harvey, M. Charity and W. Payne (2016). "Population levels of sport
52 participation: implications for sport policy." BMC Public Health 16(1): 1-8.
53
54 Eime, R., W. Payne and J. Harvey (2009). "Trends in organised sport membership: Impact
55 on sustainability." Journal of Science and Medicine in Sport 12(1): 123-129.
56
57
58 17
59
60 URL: http://mc.manuscriptcentral.com/risp
 International Journal of Sport Policy and Politics Page 42 of 47

1
2
3 Eime, R., N. Sawyer, J. Harvey, M. Casey, H. Westerbeek and W. Payne (2015).
4 "Integrating Public Health and Sport Management: Sport participation trends 2001-2010."
5 Sport Management Review 18(2): 207-217.
6
7 Eime, R., J. Young, J. Harvey, M. Charity and W. Payne (2013). "A systematic review of the
8 psychological and social benefits of participation in sport for children and adolescents:
9 informing development of a conceptual model of health through sport." International Journal
10 of Behavioral Nutrition & Physical Activity 10(98).
11
12 Eime, R. M., M. M. Casey, J. T. Harvey, M. J. Charity, J. A. Young and W. R. Payne
13 (2015). "Participation in modified sports programs: a longitudinal study of children's
14 transition to club sport competition." BMC Public Health 15(649): 7.
15
16 Eime, R. M., M. M. Casey, J. T. Harvey, N. A. Sawyer, C. M. Symons and W. R. Payne
Fo
17
(2015). "Socioecological factors potentially associated with participation in physical activity
18
and sport: A longitudinal study of adolescent girls." Journal of Science and Medicine in
Sport 18(6): 684-690.
19
rP
20
Eime, R. M., M. J. Charity, J. T. Harvey and W. R. Payne (2015). "Participation in sport and
21
physical activity: associations with socio-economic status and geographical remoteness."
22
BMC public health 15(434).
23
ee

24 Eime, R. M., J. T. Harvey, M. J. Craike, C. M. Symons and W. R. Payne (2013). "Family

25 support and ease of access link socio-economic status and sports club membership in
26
rR

adolescent girls: a mediation study." Int J Behav Nutr Phys Act 10(50).
27
28 Fransen, J., J. Pion, J. Vandendriessche, B. Vandorpe, R. Vaeyens, M. Lenoir and R.
29 Philippaerts (2012). "Differences in physical fitness and gross motor coordination in boys
ev

30 aged 6-12 years specializing in one versus sampling more than one sport." Journal of Sport
31 Sciences 30(4): 379-386.
32
iew

33 Fraser-Thomas, J., J. Coté and J. Deakin (2008). "Examining adolescent sport dropout and
34 prolonged engagement from a developmental perspective." Journal of applied sport
35 psychology 20(3): 318-333.
36
37 Gardner, L., C. Magee and S. Vella (2017). "Enjoyment and behavioral intention predict
On

38 organized youth sport participation." Journal of Physical Activity and Health Ahead of
39 Print.
40
41 Harris, S., K. Mori and M. Collins (2009). "Great Expectations: Voluntary Sports Clubs and
ly

42 Their Role in Delivering National Policy for English Sport." VOLUNTAS: International
43 Journal of Voluntary and Nonprofit Organizations 20(4): 405.
44
45 Harris, S., G. Nichols and M. Taylor (2017). "Bowling even more along: trends towards
46 individual participation in sport." European Sport Management Quarterly 17(3): 290-311.
47
48 Harvey, J. T., M. J. Charity, N. A. Sawyer and R. M. Eime (2018). "Non-response bias in
49 estimates of prevalence of club-based sport participation from an Australian national
50 physical activity, recreation and sport survey." BMC Public Health 18(1): 895.
51
52 Henrique, R., A. Re, D. Stodden, J. Fransen, C. Campos, D. Querioz and M. Cattuzzo
53 (2016). "Association between sports participation, motor competence and weight status: A
54 longitudinal study." Journal of Science and Medicine in Sport 9(10): 825-829.
55
56
57
58 18
59
60 URL: http://mc.manuscriptcentral.com/risp
Page 43 of 47 International Journal of Sport Policy and Politics

1
2
3 Hoekman, R., K. Breedveld and G. Kraaykamp (2017). "Sport participation and the social
4 and physical environment: explaining differences between urban and rural areas in the
5 Netherlands." Leisure Studies 36(3): 357-370.
6
7 Murphy, M., D. Rowe and C. Woods (2016). "Sports participation in youth as a predictor of
8 physical activity: A 5-year longitudinal study." Journal of Physical Activity and Health 13:
9 704-711.
10
11 Myer, G. D., N. Jayanthi, J. P. Difiori, A. D. Faigenbaum, A. W. Kiefer, D. Logerstedt and
12 L. J. Micheli (2015). "Sport Specialization, Part I Does Early Sports Specialization Increase
13 Negative Outcomes and Reduce the Opportunity for Success in Young Athletes?" Sports
14 Health 7(5): 437-442.
15
16 Robinson, L. and J. Goodway (2009). "Instructional climates in preschool children who are
Fo
17
at-risk. Part 1: object-control skill development." Research Quarterly for Exercise and Sport
18
80(3): 533-542.
19
Telford, R. M., R. D. Telford, T. Cochrane, R. B. Cunningham, L. S. Olive and R. Davey
rP
20
(2015). "The influence of sport club participation on physical activity, fitness and body fat
21
during childhood and adolescence: the LOOK Longitudinal Study." Journal of Science and
22
Medicine in Sport 19(5): 400-406.
23
ee

24 Veldman, S., K. Palmer, A. Okely and L. Robinson (2017). "Promoting ball skills in
25 preschool-age girls." Journal of Science and Medicine in Sport 20(1): 50-54.
26
rR

27 VicHealth. (2018). "Physical activity strategy 2018-2023." Retrieved March 9th, 2018,
28 from
29 file:///C:/Users/reime/Downloads/VicHealth%20Physical%20Activity%20Strategy%20(2).p
ev

30 df.
31
32 Wheeler, S. (2012). "The significance of family culture for sports participation."
iew

33 International Review for the Sociology of Sport 47(2): 235-252.

34
35 Wong, H., T. Olds, R. Maddison and L. Foley (2016). "A great sporting nation? Sport
36 participation in New Zealand youth." The New Zealand Medical Student Journal 23: 18-23.
37
On

38
39
40
41
ly

42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58 19
59
60 URL: http://mc.manuscriptcentral.com/risp
 International Journal of Sport Policy and Politics Page 44 of 47

1
2
3
4
5 ID RISP-2018-0027
6 Sport drop-out during adolescence: Is it real, or an artefact of sampling behaviour?
7 Rochelle Eime, Melanie Charity, Jack Harvey
8 International Journal of Sport Policy and Politics

Fo
9
We thank the reviewers for their time and their very thorough reviews and helpful comments, which have guided substantial improvements to the manuscript.
10 Our responses are tabulated below.
11

rP
12 Reviewer 1 comments Author response
13 Overall this is a well written article, with some interesting findings. Thank you

ee
14 The data is substantial and unique in respect to the opportunity to
15 access the data. However, I do feel it needs some work.
16

rR
1. I would suggest to provide more context and discussion- I got the The paper discusses quite extensively the role that player
17 impression after reading it - the main finding was quite simplistic. We development pathways may be associated with sampling,
18 wanted to do this, and we found this, but don’t really do much beyond specialising and in more general sport participation trends including

ev
19 that. For this journal I would want to see more implications to policy - retention and drop-off.
20 and reference some of the policy documents in Australia and Victoria

iew
21 - and some of the programs going on. For example, the authors We have included in both the Introduction and Discussion more
22 briefly discuss women and girls as a key focus for drop off - and discussion related to sport policy.
23 there are lots of programs happening in Victoria such as Vic Health's
24 new "This Girl Can' and the Victorian State Government's "Change

On
25 our Game". For this journal, the authors do not engage much in
26 policy implications or current policies which could be leading or
27 promoting sampling behaviour. On this note, the authors make some

ly
28 assumptions that the reader be aware of Australian context.
29 2. Australian sport policy should also be addressed I feel to give We have included more text relating to sport policy. However we
30 context - why might this be a problem around specialising and believe that the phenomenon of sampling and specialisation is more
31 sampling - and could this because of the national sport policy or in linked to the sport development and sampling/specialising
32 Victoria? I think there needs more discussion here and to tease out developmental models presented throughout this paper, rather than
33 some nuances. an outcome of sport policy.
34
35 3. I would also suggest to make reference to how the authors got the Text added (P7).
36 data - they discuss they have previously written about it but what
37 stands out to me is how they got access to such a large data set and Ethics approval is documented in the relevant declaration at the end
38 the ethics associated with it. of the paper.
39 4. The discussion and conclusion should have more links to the We have included more text relating to sport policy.
40
41 1
42
43
44
45 URL: http://mc.manuscriptcentral.com/risp
46
47
Page 45 of 47 International Journal of Sport Policy and Politics

1
2
3
4
5 Australian context and policy imperatives also - I think it’s quite
6 simplistic and basic - especially for this journal. Implications for policy
7 and the wider contextual picture would make this article better.
8 5. During the introduction and early section, I would also have more We have added more information about the Australian and Victorian

Fo
9 background to the Australian and Victorian context - it is quite context of sport.
10 different to the rest of the world - there are many things which are
11 specific to Australia and the authors on several occasions assume

rP
12 that the reader might know or be aware of the cultural context in
13 Australia.

ee
14 6. P7. Line 35 I think you mean significant rather than significantly – it Correct point. However, we have reworded this explanation.
15 reads better this way.
16

rR
I would advise to make these changes, but would suggest it does We have provided more contextual information and
17 need work before being accepted to be published in the article. How recommendations.
18 much it offers to new developments in the field is questionable too – I

ev
19 think they need to be clearer in what it means in the big picture,
20 beyond sampling behaviour causes drop out. It is important research We do not say that sampling behaviour causes drop out; rather we

iew
21 and can have significant implications for policy and localised policies show that reduction in sampling behaviour leads to dropout from
22 and participation plans for State Sport Organisations in Australia, but particular sports, but not necessarily to dropout from sport altogether.
23 this needs to be drawn out more and articulate better.
24 Reviewer 2 comments

On
25 This paper is limited in scope and it seems to me ill conceived, We have revised and extended the last two paragraphs of the
26 methodologically problematic and adds little new knowledge or Introduction (P6-7) to more clearly explain the context, rationale and
27 insight. As a consequence I am afraid that I have to conclude that I aims of the study.

ly
28 do not think as currently presented that it meets the standard
29 required for publication. This investigation into the extent to which declining registration
30 numbers can be attributed to increased specialisation versus dropout
31 from sport altogether was encouraged and supported by the sport
32 agency of the Victorian government, who regarded it as an important
33 policy issue. As far as we are aware, no quantitative evidence about
34 this issue had been presented prior to the present study.
35 There is considerable evidence around the impact of sampling and The reviewer has outlined an interesting and worthwhile research
36 specialisation - particularly as it impacts on performance and elite question, but notwithstanding the expectation of the reviewer, it is a
37 sport. However, I was encouraged by the start of this paper that said different question to the question addressed by the present study.
38 it was going to explore this more fully in the context of 'community See our previous response immediately above.
39 club-based sport'. I expected to see a study that built on the evidence
40
41 2
42
43
44
45 URL: http://mc.manuscriptcentral.com/risp
46
47
 International Journal of Sport Policy and Politics Page 46 of 47

1
2
3
4
5 to show how early sampling behaviours increased the probability for There is much discussion about the sampling and specialisation
6 sustained involvement in 'community sport' and the mechanisms for literature and the benefits of sampling for individuals. However, the
7 this - whether they are psycho-social and or competency based or as great majority of this research has related to talent development
8 seems likely a synthesis of both of these domains. I was also pathways and elite athletes. How sampling might influence sustained

Fo
9 interested to see how this might vary by social variables of gender, participation in community sport is not known, but this was not the
10 class and age. subject of this study.
11

rP
12 However the paper sought to explore drop out (or otherwise) through In the absence of access to a unique identifier (such as name and
13 what seems a convoluted process of disaggregating multiple address) to enable linking of registration records across sports, the

ee
14 registrations in club based sport to establish 'unique' individuals and process of matching based on date of birth, sex and residential
15 then by implication tracing the reduction in numbers of participants postcode is imperfect, but it is in principle quite straightforward
16

rR
and deriving a 'participation rate'. It seems to me that there are a although there is some complexity and nuance in the detail.
17 number of problems implicit in this approach - the largest of which is
18 the association that is made between club membership and In Australia, registration with a state sporting association (SSA),

ev
19 participation. This excludes any informal or semi-formal (and perhaps generally through an affiliated club, accounts for the great majority of
20 even formal non affiliated) participation outside of a club participation in competitive sports.

iew
21 environment. This point is not discussed by the authors or explained
22 or justified. There is also the issue of registration not necessarily Registration with an SSA is a good proxy for regular participation in
23 equating to participation - they are not the same thing. So we do not competitive sport during the relevant seasons.
24 know if a registered member is a frequent, infrequent or even lapsed

On
25 participant. We also do not know if an individual gives up their These unknowns are of course limitations to the study, but they do
26 'registration' as a club member but still informally participates in a not completely invalidate it or render it worthless.
27 sport or for that matter changes to a sport outside of the 11 included

ly
28 in the sample. There is no mention of what 11 sports are covered by We have now listed the 11 sports (P7).
29 the registrations - this could be important if significant participant
30 sports are excluded or where sport participation is particularly
31 socially skewed.
32 The conclusion that there really is drop out from sport into teenage This conclusion may not be surprising, but the provision of empirical
33 years (and it is not just a consequence of reduction in sampling) is evidence for it, and the quantification of the relative effects of the two
34 not a profound one. This conclusion could be drawn in much simpler mechanisms, is a worthwhile scientific endeavour with practical
35 and in more accurate ways by examining sport participation data policy implications.
36 from social surveys that include the whole population - and not just
37 club based sport and where the problems of double counting In principle, the result can be triangulated against data from
38 individuals is absent. population-wide surveys. However, there are serious limitations to
39 the data from sample surveys. First, all relevant population-wide
40
41 3
42
43
44
45 URL: http://mc.manuscriptcentral.com/risp
46
47
Page 47 of 47 International Journal of Sport Policy and Politics

1
2
3
4
5 surveys in Australia for which data were available at the time of this
6 study were limited to persons 15+ years of age, thereby omitting a
7 large proportion of the target age group, and no such survey had
8 been conducted for a number of years. While a new national sport

Fo
9 and recreation survey established more recently does include some
10 data on children and adolescents younger than 16 years of age, in
11 common with other such population-based surveys, the refusal rate

rP
12 is high, and there is evidence of non-response bias in estimates of
13 participation numbers and rates. We have now alluded to this,

ee
14 including reference to a recently published paper concerning non-
15 response bias (P6).
16

rR
It would take a major re-write for me to be convinced that this paper We hope that the reviewer is prepared to read our responses to the
17 has the potential to add useful knowledge and can address the points points above, and also the perspective of Reviewer 1, and reconsider
18 I raise above. I would, however, not want to totally close down the this assessment, which we consider to be excessively severe and

ev
19 possibility that there is something that could be published - but it unjustified.
20 would require a major re-think and re-write.

iew
21
22 Additional revisions:
23 In accordance with conventional terminology in the demographic literature, we have replaced “gender” by “sex” wherever appropriate to do so.
24 We have also revised some table headings and figure captions.

On
25
26
27

ly
28
29
30
31
32
33
34
35
36
37
38
39
40
41 4
42
43
44
45 URL: http://mc.manuscriptcentral.com/risp
46
47