Regional differences in sports participation: the case of local authorities in England.
Kokolakakis, Themis ; Lera-Lopez, Fernando ; Castellanos, Pablo 等
Introduction
Between the 1960s and the 1990s there was a significant increase in
the number of people taking part in sports in Europe (Gratton &
Taylor, 2000). Nevertheless, in recent years, this tendency has
dramatically stopped: around 40% of Europeans don't participate in
sports (European Commission, 2010). Many European countries, such as
Austria, Belgium, Finland and Portugal (Van Bottenburg, 2005), Great
Britain (Sport England, 2012) and Spain (Garda & Llopis, 2011), seem
to have reached a stagnation point, while countries such as The
Netherlands and Italy show a decline (Van Bottenburg, 2005). Beyond
Europe, Canada and the US suffer a disconcerting decrease in sports
participation (Bloom, Grant, & Watt, 2005; Barnes, 2007).
This stagnation in sports participation is of considerable concern
to health and social policy. In this respect, there is a large body of
scientific evidence regarding the positive impact of sport and physical
activity on health and wellbeing (e.g., WHO, 2010; Downward &
Rasciute, 2011). At the same time, surveys show a dramatic increase in
the numbers of overweight and obese people in developed societies.
England has not been an exception in this context. During the
period 1993-2010, as English sports participation reached a stagnation
point, obesity increased by 13% and 10% in male and female populations,
respectively. Consequently, in 2010, more than a quarter of adults and
30% of children aged 2 to 15 were classified as obese (NHS, 2010). The
expansion of obesity in England has a noteworthy regional character,
with poorer areas in the Northern regions being more affected than the
prosperous South. Within the ten-year period from 2000 to 2010, the
number of prescription items dispensed for obesity increased from
127,000 to 1.45 million items, a rise greater than 1,000%. This
relationship between sport and obesity was central to the new coalition
government's policy of public health as outlined in the White Paper
Healthy Lives, Healthy People: Our Strategy for Public Health in England
(Department of Health, 2010) , raising the question of whether public
investment in sport is able to negotiate the regional character of
obesity.
Consequently, the stagnation of sports participation in the last
ten years, coupled with evidence of health, has resulted in a strong
increase in academic interest in sports participation research in
several European countries: Belgium (Scheerder & Vos, 2011) ,
Germany (Breuer & Wicker 2008, 2009; Wicker, Breuer, &
Pawlowski, 2009), Great Britain (Downward, 2007; Downward &
Rasciute, 2011; Eberth & Smith, 2010), Scandinavian countries
(Fridberg, 2010) and Spain (Downward, Lera-Lopez, & Rasciute, 2011;
Garda, Lera-Lopez, & Suarez, 2011).
Generally speaking, these contributions have examined sports
participation using microdata obtained in various surveys. As far as we
know, in spite of the significant regional differences shown in many
studies (Sport England, 2010; Garda & Llopis, 2011), no contribution
has tried to explain differences in sports participation from a regional
perspective, considering only regional characteristics. This research
endeavors to fill this gap by developing an economic approach to
investigate sports participation among the English Local Authorities
(LAs). Using the information provided by Sport England through the
Active People Survey 5 for the period 2010-2011, we analyzed the
determinants of sports participation and regular participation
considering the regional differences in 325 English Local Authorities.
Methodologically, we use a beta model, which is a continuous
distribution, providing positive density only in a finite length
interval. This feature, together with its flexibility, makes this model
particularly appropriate for the analysis of variables that express
percentages, as in the case of sports participation, letting us
determine how participation relates to other variables through a
regression structure.
Finally, it must be underlined that in this study the examined
regional element is synonymous to Local Authority analysis. In England,
Local Government refers collectively to administrative authorities of
local areas. This may have different forms, including:
Regional Differences in Sports Participation: The Case of Local
Authorities in England
a. two-tier authorities: where district (lower-tier) councils,
responsible for council housing, leisure and recycling, interact with
single (upper-tier) councils responsible for schools, public transport
and social services;
b. unitary: where there is a single layer of administration
responsible for local services, and spanning over metropolitan district
councils, boroughs, and city, country or district councils;
c. town and parish councils, covering smaller areas than district
councils and having responsibility for allotments, public toilets,
parks, pond, war memorials and local halls or community centres; and
d. shared services across administrative bodies such as police,
fire service and public transport.
The remainder of the paper is as follows. Section 2 analyzes sports
participation in England, section 3 reviews the empirical evidence
concerning the key determinants of sports participation, section 4
describes the data set and the methodology adopted in the study, section
5 presents the main estimation results, and section 6 concludes with a
summary of the main findings, policy implications and opportunities for
further research.
Sports Participation in England: Facts and Trends
The expansion of leisure centres in England contributed towards
increasing sports participation rates steadily through the 70s and 80s,
reaching for the first time stagnation and decline in the 90s. At the
same time there was an expansion of female participation through some
sports related to keeping fit and swimming, although gender inequalities
still prevailed. Table 1 shows this decline throughout the 1990s. In the
case of "at least one activity" sports participation declined
from 64.5% in 1990 to 58.5% in 2002. Similarly, using the most
restrictive definition (i.e., excluding walking), participation declined
from 47.8% in 1990 to 43.2% in 2002.
For the period 2005-2006, in England, according to the Active
People Survey, 21% of adults were involved in sports activities at least
three times a week and 40% at least once a week. In 2006, we had a
reversal of the ten-year decline, with sports participation rising to
48.3% (at least one activity in the last four weeks, excluding walking)
and 68.2% (including walking). Nevertheless, switching from the General
Household Survey (GHS) questionnaire to the Active People Survey,
despite the consistency of the definition, may have contributed to the
sudden rise in participation. It is possible that the Active People
Survey questionnaire in its detail helps people recall more information
than in the case of GHS.
The detail of the Active People Survey enables the derivation of
several sport definitions; the one used here is the KPI1, defined as
adults (16+) participating at least three times a week for at least 30
minutes, with moderate intensity, for the purpose of leisure or
recreation. Although the policy focus has shifted to other definitions,
such as the "one million indicator," it remains an important
overall indicator of sport and active recreation at national, regional
and sub-regional levels.
After adjusting for incomplete data and seasonality, the post-2005
element in Table 1 was derived from the Active People Surveys 1-5. As it
is shown in this Table, following an initial rise in 2007, the
participation rate for the intensive KPI1 definition fluctuated with
biannual peaks at exactly 21.9%. It is remarkable that this pattern
negotiated the most important crisis of capitalism since the Second
World War without changing its structure. In part, this may be due to
the increased level of participation that results from a rising stock of
free time as unemployment increases, or to the inspiration effect on the
way to the 2012 Olympic Games (see Kokolakakis, Lera-Lopez, &
Panagouleas, 2012).
Empirical Evidence of the Determinants in Sports Participation
Our theoretical motivation is Becker's (1965) model of labor
and leisure choice, which assumes that agents derive satisfaction from
consuming "basic" commodities (such as going to the theatre, a
meal or sports participation). The production and consumption of those
commodities represents time out of work. Economic consumer choice models
of sports participation are built under which agents have to decide if
they participate at all and the amount of time spent in participation
(Humphreys & Ruseski, 2010). In this way, sports participation can
occur directly by committing goods and time in the production or
consumption of sport, or indirectly through acquiring consumption or
social capital that eventually may lead to sports participation
(Downward, Lera-Lopez, & Rasciute, 2012). This further implies that
relationships of causality are ambiguous. For example, income may be
seen as a positive determinant of sports participation, as one requires
it for the acquisition of sport equipment or membership fees. At the
same time, people with higher income are likely to have a more active
social life and civic participation (social capital), leading to a
better chance of sports participation. Yet, according to the Taking Part
Survey (DCMS, 2011), cultural participation (and participation in
general) tends to be higher among upper socioeconomic groups even in the
least deprived areas of England. This is also a pattern observed in the
case of sports participation. For a given income category, as we
introduce sport we are likely to reach higher income levels. For example
the practice of sport can provide management experience to young people
within clubs, self-discipline and an expanded social circle and social
capital, all critical factors in generating income (LIRC, 1997). Because
of the aforementioned cyclical nature in sports participation and the
cross section data used, the analysis should be seen in terms of
correlation rather than causal determination.
Since the first empirical studies dealing with leisure and sports
participation, the modeling of sports participation decision has
increased in complexity. Rather than applying Ordinary Least Squares,
logistic and two-step Heckman models opened the way for more developed
models, such as double-hurdle and zero-inflated ordered models (Downward
et al., 2011).
Due to the different approaches used, caution should be exercised
in any comparison of determinants of sports participation. Firstly, the
list of sporting activities varies from one study to another and there
is no common definition agreed upon participation in the literature.
Secondly, the sports participation variable is measured in various ways:
participation or not, frequency and intensity in sports participation,
time spent in participation, etc. Thirdly, most studies use secondary
data sources with a large sample size while other studies develop ad-hoc
surveys with primary data (Breuer, Hallmann, & Wicker, 2011).
Fourthly, the comparability of estimates from different statistical
methods may be difficult in both sign and magnitude.
However, despite these problems, it is possible to make some
qualitative general assessments concerning the main determinants of
sports participation. Following Downward et al. (2011, 2012) we classify
the key determinants into three different groups: socio-demographic,
economic and sports variables.
Among the socio-demographic variables, most cross-sectional studies
reveal an agesport participation/frequency negative relationship
(Humphreys & Ruseski 2006, 2010, 2011; Downward, 2007; Breuer
&Wicker, 2008; Hovemann & Wicker, 2009; Wicker et al., 2009;
Eberth & Smith, 2010; Fridberg, 2010; Downward & Rasciute, 2011;
Scheerder & Vos, 2011; Kokolakakis et al., 2012) due to biological
and physical limitations, affecting males more than females (Breuer
& Wicker, 2009). On the other hand, using a longitudinal
perspective, Stamatakis and Chaudhury (2008) claim an increasing
participation among middle-aged and older adults. Also, in some
countries, frequency increases with age (Humphreys & Ruseski, 2006;
Lera-Lopez & Rapun-Garate, 2007; Garda et al., 2011), perhaps due to
a higher level of health awareness among older people.
Gender is another important determinant of sports participation,
with a consensus about the fact that men, in general, not only
participate in sport more than women (Humphreys & Ruseski, 2006;
Downward, 2007; Lera-Lopez & Rapun-Garate, 2007; Breuer &
Wicker, 2008; Hovemann & Wicker, 2009; Eberth & Smith, 2010;
Fridberg, 2010; Downward & Rasciute, 2011; Kokolakakis et al., 2012)
but they also show a higher frequency of participation (Humphreys &
Ruseski, 2006; Eberth & Smith, 2010). This asymmetric behavior can
be attributed to biological factors or cultural and social influences
(e.g., differences in housing and relative family responsibilities).
Similarly, a consensus about the positive effect of education has
been formed (Humphreys & Ruseski, 2006, 2010; Downward, 2007; Breuer
& Wicker, 2008; Hovemann & Wicker, 2009; Wicker et al., 2009;
Eberth & Smith, 2010; Fridberg, 2010; Downward & Rasciute, 2011;
Scheerder & Vos, 2011; Kokalakakis et al., 2012). Following Fridberg
(2010), it could be argued that a higher level of education might lead
to a greater awareness of the personal benefits and importance of sport.
Traditionally, higher educational levels are associated with higher
hourly wages and more available resources to take up sporting
activities. However, the evidence is less conclusive in terms of
frequency, with a negative relationship suggested in some studies (e.g.,
Downward & Riordan, 2007) while others, such as Humphreys and
Ruseski (2010) and Ruseski, Humphreys, Hallmann, and Breuer (2011),
reported a positive relationship.
Ethnicity has been widely included into analysis of sports
participation in countries such as England (Downward, 2007; Stamatakis
& Chaudhury, 2008) and the US (Humphreys & Ruseski, 2006).
Generally, there is a positive correlation between participation and
being white while other ethnicities could suffer from cultural barriers
(Breuer et al., 2011).
According to Downward et al. (2011, 2012), household influence on
sports participation is commonly determined through variables such as
marital status and number of both children and adults in the household.
Since time is finite, any increase in the time devoted to sport will
always be constrained by competing demands from other leisure activities
and other uses (family, etc.). In this context, it should be expected
that married people and families with more members participate less in
sport (Humphreys & Ruseski, 2006, 2010; Downward, 2007; Hovemann
& Wicker, 2009; Eberth & Smith, 2010; Garda et al., 2011;
Ruseski et al., 2011), although this negative relationship does not
happen in other studies (e.g., Downward & Rasciute, 2011).
Population size is commonly considered a proxy variable, measuring
the availability of sports facilities with ambiguous effects. On the one
hand, the empirical evidence might lead us to expect less access to
certain types of sporting facilities in rural areas than in suburbs or
cities (Hovemann & Wicker, 2009); on the other hand, in large cities
there is greater availability of a wider range of entertainment options
and consequently more substitute leisure activities for sport. In some
European countries, the latter effect is greater than the first one,
leading to a higher sports participation in rural areas (Garda et al.,
2011), while in the EU-25 it seems that individuals living in large
towns tend to be more involved in sport (Van Tuyckcom et al., 2010).
As sports participation requires consumption of some sporting goods
and services, a second group of key determinants show the influence of
economic variables. There is consensus on a positive relationship
between income and sports participation (Downward, Dawson, &
Dejonghe, 2009 and Breuer, Hallmann, Wicker, & Feiler, 2010, for a
literature review). However, among regular practitioners, income has no
influence on the frequency of sports participation (Gratton &
Taylor, 2000) or the influence is negative (Humphreys & Ruseski,
2006, 2010, 2011; Downward & Riordan, 2007). As Garda et al. (2011)
argued, this could be explained because the higher the income, the
higher the opportunity cost of time spent on any sports activity.
Professional status as a proxy income variable is positively related to
sports participation, with higher professional level groups (Van Tuyckom
& Scheerder, 2010) and white-collar jobs (Humphreys & Ruseski,
2010) having the highest level of sports involvement.
In addition, working, in general, is negatively related to sports
participation (Downward, 2007; Breuer & Wicker, 2008; Hoveman &
Wicker, 2009; Eberth & Smith, 2010), confirming a substitution
effect between sport and other leisure and work choices. However, other
authors (Downward et al., 2009; Wicker et al., 2009) found a positive
relationship, as sport is used "to compensate" for work life.
Finally, a third group of determinants included in many studies
relate to the influence of sport supply. Generally speaking, sports
participation is induced or constrained by sport supply. For example,
Downward and Rasciute (2011) report that sports facilities in general
promote participation in sports. Wicker et al. (2009) show that a poor
supply of sports facilities reduces the regularity of sports activities.
Nevertheless, the empirical evidence is far away from being clearly
conclusive. Hallmann, Wicker, Breuer, and Schonherr (2012), at
municipality level in Germany, show that sports facilities are of
importance when predicting sports participation, although the influence
depends on the type of both sport and sporting facility. Similar results
are shown by Humphreys and Ruseski (2007) in the US: the effect of
government spending on participation depends on the nature of the
sporting and physical activity. According to Hallmann, Wicker, Breuer,
and Schuttoff (2011), the impact of sport supply on sports participation
depends on the size of population where a sports facility is located. In
Spain, Pascual, Regidor, Martinez, Calle, and Dominguez (2009) found no
relationship between the provision of sporting facilities and
participation in jogging, swimming and gymnasium use.
To sum up, a positive relationship is expected between some
demographic/economic variables (such as gender [males], education,
ethnicity [white], and income) and sports participation. In addition, a
negative relationship with age and unclear effects of population size
and sports infrastructure on sports involvement should be tested more
carefully.
Data and Methodology
Data
The majority of data was taken directly from the Active People
Survey 5 (2010/2011), the largest ever survey of sport and active
recreation in Europe (Sport England, 2012). This was the latest survey
available at the time this paper was written. Further, in adopting a
cross-sectional approach, we follow the practice of all the economic
impact reports in the UK, starting with the Henley Centre for
Forecasting report (1986). The Survey started in mid October 2010 and
ran continuously for 12 months until the middle of October 2011. 166,000
English adults (age 16 and over) were interviewed by telephone across
the country. The sample was randomly stratified and the results are
representative of the total adult population in the country, at regional
and local levels. From this survey, a dataset was developed, collecting
information about the 325 English Local Authorities. In this sense, all
the examined variables relate directly to the Local Authorities as
percentages, without specific references to personal information.
From the Survey, two variables were selected as dependent
variables: "sports participation" and "regular sport
participation." They correspond, respectively, to the definitions:
"percentage of adults participating at a sport activity for
training, competition or recreation, at least in one 30 minutes moderate
intensity session, during the last four weeks" and "percentage
of adults participating at a sport activity for training, competition,
or recreation, in sessions of at least 30 minutes of moderate intensity,
at least three times per week." The second definition, being the
more intensive one, is closer to the current health policy
recommendations and is more relevant in informing policy (KPI1, as
explained in the previous "Facts and Trends" section). The
first is more inclusive and remains relevant for comparisons with
previous datasets. This definitional distinction is consistent with much
of participation analyzes in Europe, examining people passing through a
hierarchy of participation stages (e.g., COMPASS Project; see for more
details Gratton, Rowe, & Neal, 2011). In both definitions above,
walking is included if it is intensive while recreational cycling for
commuting purposes is excluded. From this point onwards, the first
variable is referred as "sports participation" and the second
as "regular sports participation."
As independent variables, and following our theoretical motivation
and the empirical evidence available, four different groups have been
considered:
1. First, socio-demographics variables including gender (male), age
(three intervals: 16-34 [omitted as a base variable], 35-64, and over
64), ethnicity (white), size of households (single households, presence
of children, and households with four or more adults), educational level
(degree level, A-level, and GCSE or not finishing GCSE [omitted]),
percentage of urban population, and health limitations (longstanding
illness).
2. Second, economic variables, including: income level (with three
categories: up to [pounds sterling]20,799, [pounds
sterling]20,800_[pounds sterling]41,599, and over [pounds
sterling]41,600 [omitted]), council house, type of occupation
(professional, managerial and technical, skilled [non-manual] skilled
[manual], partly skilled, and unskilled occupations; from them, only
"skilled [non-manual]" and "unskilled" were included
in the final analysis), economic/working status (six categories:
full-time jobs, part-time jobs, retired [omitted], looking after
children, students, and unemployed), median gross weekly level of
earnings, and "live and work in the same area."
3. Third, sports variables, including not only some variables
closely related to sports participation, such as the percentage of
sports volunteers in the region, but also some others reflecting the
level of sports infrastructure. In particular, access to sporting
facilities has been incorporated into the analysis, through three
variables:
(1) the percentage of population that have 20-minute access to
three types of facility (taken from pool, hall, health & fitness,
grass pitch, synthetic turf pitch or golf course) of which at least one
has a high level of quality ("Quality Assured");
(2) the percentage of population that live within 20 minutes'
travel time from three types of facilities (without any Quality Assured
restrictions); and, finally (3) the percentage of people living within
20-minute travel of one type of facility, as defined above. We have also
taken into account variables reflecting the sport funding in the region
(Local Funding) and the quality of the regional provision of some
services (Comprehensive Performance Assessment Score), including sports
services, to the population. As local funding in sports we have included
Lottery funding and Exchequer awards (both distributed through Sport
England to the English regions) and capital expenditure in sports,
derived from Local Authority sources. As these funds could vary from one
year to another, we have considered the total amount in the period
2007-2010 (1).
4. Finally, another set of variables, such as the total area of
inland water (lake area) and the costal length, relates to participation
in water-sports. Information is included on temperature and number of
days with rain higher than 1mm to check the influence of the weather on
sports participation. Lake area and Coast length were derived using
Geographic Information System (GIS) analysis; Temperature and Rainy days
were obtained from meteorological data provided by the Met. Office, UK.
Table 2 shows the list of the variables finally selected to be
included in our empirical study and their main descriptive statistics
(mean, standard deviation and Pearson's variation coefficient).
Methodology
The selection of the modeling methodology is constrained by the
characteristics of the dependent variable. In our case, both variables
measuring sports participation are continuous proportions as they take
values in the interval (0, 1). The question that focuses our research is
how participation in sports is related to other variables through a
regression structure.
As Kieschnick and McCullough (2003) point out, although researchers
most frequently estimate the parameters of a linear regression model for
proportions using OLS, such an approach contravenes two conditions: (1)
the conditional expectation function must be nonlinear because it maps
onto a bounded interval, (2) and its variance must be heteroskedastic
because the variance will approach zero as the mean approaches either
boundary point. Due to these problems, a linear model may give
impossible predictions (out of the range 0-1), non-normal errors,
heteroskedasticity and nonlinear effects (Paolino, 2001).
From Kieschnick and McCullough's (2003) detailed analysis of
the advantages and shortcomings of some alternatives to avoid the
aforementioned problems, we opted for applying a beta regression model,
which was found to be the best option in our circumstances. In this
sense, for instance, we could use neither a Tobit model, because our
dependent variable is not censored but limited to the interval [0, 1],
nor a logit regression because this implies a discrete (binary)
dependent variable whereas proportions ("sports participation"
in our case) are continuous variables.
This distribution belongs to the family of continuous distributions
and has the characteristic of providing positive density only in a
finite length interval, (0, 1) unlike other models, such as the normal
or the exponential distributions, which have a positive density in an
infinite interval. According to the conventions of Generalized Linear
Models (GLM), the standard beta model is defined by (Ferrari &
Cribani-Neto, 2004)
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
where Var ([mu]) = [mu](1 - [mu]), [mu](location parameter) is the
mean of the response variable, and [phi] (scale parameter) can be
interpreted as a "precision" parameter in the sense that, for
fixed [mu], the larger the value of [phi], the smaller the variance of
X. The variance is therefore a function of both the mean and the
parameter [phi]. As Paolino (2001) points out, "one particularly
attractive feature of the beta distribution is its recognition of a
relationship between the mean and the variance that may occur with
proportions. A normally distributed variable can have any variance. But
a beta-distributed variable with a mean close to either 0 or 1 generally
has a smaller variance than a proportion with a mean of 0.5. Any
covariate in a quantitative model that has a large effect upon the mean
is also likely to imply a heterogeneous variance" (p. 326).
The beta distribution models heteroskedasticity in such a way that
the variance is largest when the average proportion is near 0.5 while
the mean assumes different values in different regions depending on the
values of the explanatory variables [[mu].sub.i] = f([b.sub.0] +
[b.sub.1][x.sub.1i] + [b.sub.2][x.sub.2i] + ...).
The beta distribution uses the logistic transformation
[[mu].sub.i] = exp([b.sub.0] + [b.sub.1][x.sub.1i] +
[b.sub.2][x.sub.2i] + ...)/1 + exp([b.sub.0] + [b.sub.1][x.sub.1i] +
[b.sub.2][x.sub.2i] + ...)
to ensure that p remains between 0 and 1, and the resulting
regression coefficients can be interpreted as log-odds (Smithson &
Verkuilen, 2006).
The beta density function can have quite different shapes
(symmetrical, "J," "inverted J," "U")
depending on the values of the two parameters, so it is very flexible
and versatile; this model can produce a unimodal, uniform, or bimodal
distribution of points that can be either symmetrical or skewed.
Therefore, its empirical use comprises a wide range of applications.
Modeling proportions is just one of them (see more details in Gupta
& Nadarajah, 2004).
Results and Discussion
Before modeling sports participation in the English LAs, the degree
of correlation of various potential explanatory variables was analyzed.
In this case, since the beta regression is a non-linear model estimated
by ML (Maximum Likelihood), multicollinearity has not the serious
implications associated with the traditional linear models estimated by
OLS (Ordinary Least Squares). Following the parsimony principle,
redundant variables were eliminated, deriving a model with a virtually
equal explanatory power to the one with all variables included, but with
a shorter set of regressors. For this reason, variables showing high
correlations levels (Pearson's coefficient higher than 0.7) were
examined, rejecting variables which (1) showed significant correlations
with a large number of covariates, and (2) their elimination avoided
deleting a large number of other variables. This process implied the
elimination of eighteen variables from the preliminary list of potential
candidates (2).
The descriptive results collected in Table 2 show a relatively low
dispersion level in sports participation, which is more significant in
the case of the variable regular sports participation. On average,
sports participation is about 46% in the English regions, while regular
sports participation is less than half this value at 22%.
The behavior of the explanatory variables is much more
heterogeneous in terms of dispersion. Most of them show low or moderate
levels, but there are exceptions. For example (in decreasing order with
Pearson's variation coefficients in parentheses): Local funding
2007-2010 (166.74), Coast length (128.31), 20 min 1 facility quality
assured (85.71) or Council house (73.30).
In relation to the socio-demographic variables, it is important to
point out that almost one in five households is a single household,
children are present in less than a third of them, and the household
size is generally small, with only 11% of them consisting of four or
more adults. In terms of educational levels, we can highlight that more
than a quarter of people have a higher education degree or equivalent.
Almost 25% of people suffer from a limiting longlasting illness,
disability or infirmity, and males represent only about 40% of the total
sample. Most of the individuals surveyed belong to the age interval of
more than 35 years old (about 84%), are white (almost 93%) and live in
urban areas (about 85%).
Regarding the economic variables, only about 6% of people live in
council houses, and the income distribution is almost identical among
the three income categories (i.e., a 33% share for each). Most of the
individuals are qualified workers (unskilled occupations account for
only 3%) that hold full-time jobs (almost 45%) and have a median gross
weekly earning of [pounds sterling]412. The significance of commuting is
clear, as less than 60% of people live and work in the same area.
In relation to the sports variables, about 6% of people have been
volunteers in sport. There are very high levels of proximity to sports
facilities: almost 99% and 80% of people live within 20 minutes'
travel time of one or three types of facilities, respectively.
After the descriptive statistics, in Tables 3 and 4, the beta
regression model results are reported. On the one hand, in the case of
Sports Participation, 11 explanatory variables are significant: five at
the 1% level (Higher Education, Longstanding illness, White, Urban and
Sport volunteers), three at the 5% level (A-Levels, Income up to 20,799
[pounds sterling] and Temperature) and, finally, three at the 10% level
(Age over 64, Unskilled and Student). On the other hand, in the case of
the Regular Sports Participation 10 regressors show statistical
significance: three at the 1% level (Higher Education, White and
Full-time), four at the 5% level (Male, Urban, Sport volunteers and
Temperature) and, lastly, three variables at the 10% level (Children,
Income 20,800-41,599 [pounds sterling] and Rainy days).
Because the beta model is nonlinear, cross-regression comparisons
for the effect of explanatory variables on sport participation cannot be
made. For this purpose, as in the case of other nonlinear models (such
as the logit model), we have to use the marginal effects ([partial
derivative]y/[partial derivative][x.sub.i]), which help us understand
the impact of each covariate. Table 5 reports the marginal effects of
the beta regression models with respect to the different regressors,
evaluated at their sample means. For the sake of simplicity, only the
marginal effects of the statistically significant explanatory variables
are included. The interpretation of the marginal effects is
straightforward.
The variable Sports Participation (SP) depends on some
socio-demographic variables such as education, age, ethnicity, and
population size. As it is shown in Tables 3 and 5, it depends positively
on the educational levels, confirming the empirical evidence in previous
studies in England (Downward, 2007; Downward & Rasciute, 2011;
Kokolakakis et al., 2012). On average, an increase of 1% in the
percentages of people with A-Levels or Higher Education implies an
increase of 0.14% and 0.25%, respectively, in sports participation.
As it is expected, longstanding illnesses and old age (over 64
years) have a negative impact on SP: an increase of 1% in those
regressors involves a reduction in sports participation of 0.18% and
0.15%, respectively. Consequently, physical limitations due to illness
or age are clearly constraints on participation. Finally, SP has a
positive relationship with being white and urban living: an increase of
1% in these variables causes an increase of 0.18% and 0.05%,
respectively, in SP. In the English LAs, as in the case of the EU-15
(Van Tuyckom & Scheerder, 2010), it seems that the sport supply
effect is higher than the substitute effect in urban areas.
Significantly, neither gender nor household characteristics have an
important influence on explaining differences in sports participation.
Among the economic variables, low income levels or low-skilled jobs
imply a reduction in SP: on the one hand, an increase of 1% in the
percentage of people with income lower than 20,799 [pounds sterling]
involves a reduction of 0.09% in SP and, on the other hand, a similar
increase in the proportion of unskilled people is associated with a
lessening of 0.23% in the dependent variable. Consequently, in the
English LAs, low income is a factor explaining differences in regional
participation in sports, confirming the empirical evidence analyzed by
Downward et al. (2009) and Breuer et al. (2010). With regard to the
occupation variables, only "being a student" is statistically
significant, confirming that the substitution effect between sport and
work neutralizes the compensation effect. In other words, the negative
effect of lesser income among students is neutralized by the positive
influence of free time available to commit in sports participation. An
increase of 1% in students leads, on average, to an increase of 0.17% in
SP. This means that differences in local sports participation are
partially explained in terms of the distribution of student population.
Unexpectedly, neither sports expenditure in the region nor the
sports infrastructure are determinants of sports participation. A
possible explanation for the absence of impact of expenditure in sport
is that we examined only a three-year period of investment. Note,
however, that this is in line with major research economic impact
studies (such as Henley Centre, 1986), where while adopting a
cross-sectional context, investment is treated as characteristic
investment over a three-year period. Other more general studies (Sport
England, 2010) have found Lottery funding to be significant, but only
after going back to 1995. Hence, when considering a medium term horizon,
spending on its own cannot explain regional variations in sports
participation. In the case of the lack of influence of sports
infrastructure, various explanations could be considered. This is partly
because public policy targets worse-off areas and partly because the
private sector looks for business opportunities in places where there is
no saturation in terms of health and fitness clubs and leisure centres.
For example, the health and fitness operator "Pure Gym" has
targeted the low-cost sector, offering facilities without a monthly
contract. More significantly, the business plan of another operator,
"Anytime Fitness," involves identifying and expanding on areas
without health and fitness facilities. These trends ensure that there is
enough sports infrastructure in place to absorb the current demand both
nationally and regionally. Because of that, in the medium term, although
spending and infrastructure may be important determinants in national
policy (or in a single region), they are not sufficient to explain
variations of participations across regions. Public policy usually seeks
to help disadvantaged areas (to increase participation), tending to
resolve regional differences.
The level of sports volunteers in the regions has a positive and
statistically significant impact on sports participation: a raise of 1%
in the proportion of sport volunteers in a local area implies an
increment of 0.35% in SP. Sports volunteers normally develop their work
in voluntary sports clubs, confirming the importance of these clubs to
increase sports participation, as it has been shown by Reid (2012) in
Scotland.
Finally, weather conditions explain differences in SP; temperature
variation has a negative effect: each 1% increase in the difference
between the maximum and minimum temperatures registered within a local
area involves a decrease of 0.04% in participation.
The Regular Sports Participation (RSP) model and its marginal
effects are shown in Tables 4 and 5, respectively. As in the previous
model, significant variables include education, ethnicity and population
size. In addition, gender and the presence of children in the household
are also significant. For example, having children has a negative impact
on regular sports participation: an increase of 1% in people who have
children implies a reduction of 0.09% in RSP, endorsing the results of
Humphreys and Ruseski (2010) in the US and Ruseski et al. (2011) in
Germany. Gender also plays a role; in this sense, an increase of 1% in
the percentage of male individuals is associated with a 0.12% of
increase in RSP, confirming empirical evidence shown by Eberth and Smith
(2010) for Scotland, and Humphreys and Ruseski (2006) for the US.
Educational level has, again, a positive impact, although slightly
lower than in the case of Sports Participation. In this case, A-level is
not statistically significant and only higher education has a positive
impact: on average, an increase of 1% in that explanatory variable
implies an increase of 0.15% in RSP. A more general conclusion is that
as we switch to more intensive forms of sports participation only high
levels of education become relevant in explaining regional variations.
The current structure of general education (up to A-levels) contributes
effectively towards a basic structure of sports participation (once a
month), but is less relevant when we focus on regular participation.
As in Sports Participation, white ethnicity has a positive impact,
although now this impact is slightly lower: an increase of 1% in the
proportion of white people implies an increase of 0.12% in RSP. Unlike
SP, living in urban areas has a negative effect on RSP: an increase of
1% in that explanatory variable involves a reduction of 0.03% in RSP. In
other words, when considering regular sports participation, the
substitution effect between sport and other leisure activities is higher
than the positive sport supply effect in urban areas.
Among the economic variables, the role of individuals' income
is important: an increase of 1% in the proportion of people included in
the "middle class" (i.e., income between 20,800 [pounds
sterling] and 41,599 [pounds sterling]) implies a reduction of 0.08% in
RSP. Consequently, low-income level is a barrier to participate in
sport, while regular sports participation is more constrained by
middle-income levels (which are likely to be more restricted in terms of
free time). With regard to occupation, people with full-time jobs are
more likely to be regular sports participants: more precisely, a raise
of 1% in full- time occupations involves a 0.18% increase in RSP.
As mentioned before, neither the sports expenditure in the region
nor the sports infrastructure are determinants of regular sports
participation. Only the level of sports volunteers in the LAs seems to
have a positive and statistically significant impact on regular
participation, although slightly lower than in the former case: an
increase of 1% in the proportion of sports volunteers is associated with
a rise of 0.27% in RSP.
Finally, weather conditions have a significant impact on regular
sports participation. On the one hand, temperature variation has a
similar effect in sign and quantity compared to the previous model; on
the other hand, rainfall has a positive and unexpected effect on RSP: an
increase of 1% in the number of rainy days implies an increase of 0.03%
in RSP.
Conclusions and Policy Implications
In this paper the determinants of sports participation among the
English LAs are examined under two different definitions: a basic
participation in the last four weeks and a more intensive participation
of at least three times per week (regular). The focus of our analysis is
not on the significant factors determining sports participation within a
region, but on the factors that can help us explain the differences
among them. As far as we know, this is the first time that a regional
approach is exclusively followed to explain sports participation, with a
huge sample of individuals and regions (LAs). Using the information
provided by Sport England through the Active People Survey 5 for the
period 2010-2011, we analyzed the determinants of sports participation
and regular participation considering the regional differences in 325
English Local Authorities.
As the dependent variables are measured in terms of proportions of
regional population (aged 16 and over), we developed beta regression
models. These models have significant advantages compared to the
traditional linear regression models (OLS). Beta distribution is a
continuous model with the characteristic of providing positive density
only in a finite length interval. This feature and, also, its
flexibility make this model particularly appropriate for variables
representing proportions, as in the case of sports participation. We
identified four groups of regional characteristics: socio-economic
variables, economic determinants, sports variables and other regional
characteristics (weather conditions, etc.).
The results show the importance of some socio-demographic
variables, such as educational level, ethnicity, and size of population;
economic variables (income levels and occupations); sports volunteering;
and weather conditions. Nevertheless, sports infrastructure is not
significant in explaining differences in sports participation among the
English LAs. The important implication here is that public sports policy
is well-targeted, and private investment in sports facilities is
effective in balancing demand and supply across regions. It is important
to emphasize that this does not mean that sports infrastructure is
unimportant in general; however, in this case it cannot explain the
existing variations among LAs.
Sports participation (SP) is positively affected by the educational
level in the region, the percentage of white population, the percentage
of people living in urban areas, and the percentage of students and
sports volunteers in the region. At the same time, the percentages of
aged population and people suffering from longstanding illnesses as well
as the percentages of unskilled workers (the lowest level of
professional status) in the regions and the local temperature variation
have a negative impact on SP. Finally, a low-income level is a clear
constraint on participation across LAs.
Regular sports participation (RSP) is, on the one hand, positively
affected by higher educational and income levels, being male and white,
and working full-time; on the other hand, it is negatively influenced by
having children and living in urban areas. Additionally, the percentage
of sports volunteers and the number of rainy days have a positive impact
on regular participation, while the impact of regional temperature
variation is negative.
These results suggest alternative regional sports policies to boost
sports participation in England in the context of stagnation of sports
participation rates and a drastic increase in obesity and health
problems in the population. In addition, sports policies have to draw a
distinction between the decision to participate in sport and the
decision about the frequency of regular participation (Downward et al.,
2011). If the final policy aim is to improve the health level of
population, special attention should be focused on the determinants of
regular sports participation.
Firstly, some implications could be highlighted comparing both
decisions. There are some common variables in both models, such as
educational and income levels, ethnicity, sports volunteers and regional
temperature variation. In general, the impact of these variables is
greater (comparing marginal effects) when explaining SP than RSP.
In particular, low-income level is clearly a barrier to participate
in sport, but its effect on RSP is not relevant. Consequently, in order
to attract new participants with incomes up to [pounds sterling]20,799,
partially or totally subsidized regional sports programs should be
developed. However, marginal variations in price are not going to be
very effective in transforming a basic participation into a regular one.
If the middle class participation lacks at this level it is not because
they cannot afford it, but because they might not have sufficient free
time or inclination to transform radically their lifestyle.
Similarly, some conclusions could be drawn from the impact of
education. Having middle and high educational levels helps people to be
involved in sport, but only higher education has a significant impact on
regular sports participation. Information about the health benefits
associated with regular sports participation should be underlined in
secondary education. Informative and persuasive messages through the
mass media (e.g., about the positive impacts on health and appearance,
the opportunities of socialisation, etc.) could attract new participants
among people having only GCSEs or nothing at all.
Moreover, any attempt to promote sports volunteering in sports
clubs and other sports institutions could have a positive effect on
sports rates, especially in general participation.
Secondly, the fact that some variables affect exclusively one
participation definition and not the other allows us to develop some
additional policy conclusions. For example, the percentage of people
aged over 64 years has a negative impact on sports participation. Public
policy should counteract this trend utilising resources and targeting
participation within this age group. The fact that this group has a lot
of free time should make it more responsive to policy from both the
public and private sectors. Furthermore, the negative impact of having
children and living in urban areas together with the positive impact of
being male in regular participation suggest the development of regular
family sporting activities in urban areas including adults and children
participating together in activities such as swimming, paddle, tennis,
etc.
Additionally, the results suggest the possibility, in terms of
participation, of a regional gap between rural and urban areas as well
as between males and females. Consequently, regional sports programs
could be carried out in rural areas to boost some sports practices, such
as cycling, mountaineering, climbing, etc. Similarly, emphasis should be
given to regular participation among women, paying special attention to
possible motivations (health, physical appearance, etc.) and to
developing sport activities for the whole family.
Thirdly, the results show that the general problem of stagnation in
sports participation rates could not be solved simply by pouring
increasing levels of public funds into the sport industry. It is
extremely important to fine tune policy according to regional
characteristics and intervene on education and youth sport. Finally, it
was shown that, in the medium term, government funding policy is not
effective in terms of boosting sports participation. Any financial plan
will need more than a single government term to achieve the desired
results. Hence, any short-term decisions designed to bring imminent
gains in sports participation are likely to be counterproductive. There
is a need for cross-party consensus on sports policy, especially with
regards to school sports, and policy objectives to safeguard continuity
beyond the four-year term of a government.
Some limitations of this study could provide new ideas for further
research. For example, longitudinal analysis could be used to study
dynamic clusters of regions and to evaluate possible changes in the
impact of the independent variables on the two participation definitions
(basic participation and regular participation) over time.
References
Barnes, P. (2007). Physical activity among adults: United States,
2000 and 2005. Hyattsville, MD: US Department of Health and Human
Services, CDC.
Becker, G. (1965). A Theory of the Allocation of Time. The Economic
Journal, 75, 493-517. Breuer, C., Hallmann, K., & Wicker (2011).
Determinants of sport participation in different sports. Managing
Leisure, 16, 269-286.
Breuer, C., Hallmann, K., Wicker, P., & Feiler, S. (2010).
Socio-economic patterns of sport demand and ageing. European Revue Aging
Physical Activity, 7, 61-70.
Breuer, C., & Wicker, P. (2008). Demographic and economic
factors influencing inclusion in the German sport system. A
microanalysis of the years 1985 to 2005. European Journal for Sport and
Society, 5, 33-42.
Breuer, C., & Wicker, P. (2009). Decreasing sports activity
with increasing age? Findings from a 20-year longitudinal and cohort
sequence analysis. Research Quarterly for Exercise and Sport, 80, 22-31.
DCMS. (2011). Taking Part: The National Survey of Culture, Leisure
and Sport. Jan-Dec 2010 Statistical Release, DCMS.
Department of Health (2010). White Paper: Healthy Lives, Healthy
People: Our Strategy for Public Health in England. Retrieved October 4,
2012 from http://www.dh.gov.uk/en/Publications
andstatistics/Publications/PublicationsPolicyAndGuidance/DH_121941
Downward, P., Dawson, A., & Dejonghe, T. (2009). Sports economics:
theory, evidence and policy. Oxford, UK: Elsevier.
Downward, P., Lera-Lopez, F., & Rasciute, S. (2011). The
zero-inflated ordered probit approach to modelling sports participation.
Economic Modelling, 28, 2469-2477.
Downward, P., Lera-Lopez, F., & Rasciute, S. (2012). The
economic analyses of sport participation. In L. Robinson, P.
Chelladurai, G. Bodet, & P. Downward (Eds.), Routledge Handbook of
Sport Management (pp. 331-353). London, UK: Routledge.
Downward, P. (2007). Exploring the economic choice to participate
in sport: Results from the 2002 General Household Survey. International
Review of Applied Economics, 21, 633-653.
Downward, P., & Rasciute, S. (2011). Does sport make you happy?
An analysis of the well-being derived from sports participation.
International Review of Applied Economics, 25(3), 331-348.
Downward, P., & Riordan, J. (2007). Social interactions and the
demand for sport: An economic analysis. Contemporary Economic Policy,
25, 518-537.
Eberth, B., & Smith, M. (2010). Modelling the participation
decision and duration of sporting activity in Scotland. Economic
Modelling, 27(4), 822-834.
European Commission. (2010). Eurobarometer 72.3. Sport and Physical
Activity., Retrieved from http://www.mhfi.org/sportandphysical.pdf
Ferrari, J.L.P., & Cribani-Neto, F. (2004). Beta regression for
modeling rates and proportions. Journal of Applied Statistics, 10, 1-18.
Fridberg, T. (2010). Sport and exercise in Denmark, Scandinavia and
Europe. Sport in Society, 13(4), 583-592.
Garda, J., Lera-Lopez, F., & Suarez, M.J. (2011). Estimation of
a structural model of the determinants of the time spent on physical
activity and sport: Evidence for Spain. Journal of Sports Economics,
12(5), 515-537.
Garda, M., & Llopis, R. (2011). Ideal democratico y bienestar
social. Encuesta sobre los habitos deportivos en Espana 2010. Madrid:
CIS and CSD.
Gratton, C., Rowe, N., & Neal, A. (2011). International
comparisons of sports participation in European Countries: An update of
the COMPASS Project. European Journal for Sport and Society, 8(1-2),
99-116.
Gratton, C., &Taylor, P. (2000). Economics of sport and
recreation. London: Spon Press.
Gupta, A.K., & Nadarajah, S. (Eds.) (2004). Handbook of beta
distribution and its applications. New York: Marcel Dekker.
Hallmann, K., Wicker, P., Breuer, C., & Schonherr, L. (2012).
Understanding the importance of sport infrastructure for participation
in different sports--findings from multilevel modelling. European Sport
Management Quarterly, 12(5), 525-544.
Hallmann, K., Wicker, P., Breuer, C., & Schuttoff, U. (2011).
Interdependency of sport supply and sport demand in German metropolitan
and medium-sized municipalities-- findings from multi-level analyses.
European Journal for Sport and Society, 8, 65-84.
Henley Centre for Forecasting. (1986). The Economic Impact and
Impoartance of Sport in the UK. International Review for the Sociology
of Sport, 22(3), 149-170.
Hovemann, G., & Wicker, P. (2009). Determinants of sport
participation in the European Union. European Journal for Sport and
Society, 6, 51-59.
Humphreys, B., & Ruseski, J. E. (2006). Economic determinants
of participation in physical activity and sport. (IASE Working Paper No.
06-13). International Association of Sports Economics.
Humphreys, B., & Ruseski, J. E. (2007). Participation in
physical activity and government spending on parks and recreation.
Contemporary Economic Policy, 25, 538-552.
Humphreys B., & Ruseski, J. E. (2010). The economic choice of
participation and time spent in physical activity and sport in Canada
(Working Paper No 2010-14). University of Alberta.
Humphreys, B., & Ruseski, J. E. (2011). An economic analysis of
participation and time spent in physical activity. The B.E. Journal of
Economic Analysis & Policy, 11, 1-36.
Kieschnick, R., & McCullough, B. D. (2003). Regression analysis
of variates observed on (0, 1): Percentages, proportions and fractions.
Statistical Modelling, 3, 193-213.
Kokalakakis, T., Lera-Lopez, F., & Panagouleas, T. (2012).
Analysis of the determinants of sports participation in Spain and
England. Applied Economics, 44(21), 2785-2798.
Lechner, M. (2009). Long-run labour market and health effects of
individual sports activities. Health Economics, 28, 839-854.
Lera-Lopez, F., & Rapun-Garate, M. (2007). The demand for
sport: Sport consumption and participation models. Journal of Sport
Management, 21, 103-122.
LIRC (1997). A review of the economic impact of sport in the United
Kingdom. Final report for the Sports Council, London.
National Health Service (NHS) (2011). Statistics for obesity,
physical activity and diet. England 2011. Retrieved from
http://www.ic.nhs.uk/webfiles/publications/003_Health_Lifestyles/
opad11/Statistics_on_Obesity_Physical_Actfvity_and_Diet_England_201Lpdf
Paolino, P. (2001). Maximum likelihood estimation of models with beta-
distributed dependent variables. Political Analysis, 9, 325-346.
Pascual, C., Regidor, E., Martmez, D., Calle, E., & Dommguez,
V. (2009). Socieconomic environment, availability of sports facilities,
and jogging, swimming and gym use. Health & Place, 15, 553-561.
Reid. F. (2012). Increasing sports participation in Scotland: Are
voluntary sports clubs the answer? International Journal of Sport Policy
and Politics, 4(2), 221-241.
Ruseski, J., Humphreys, B., Hallmann, K., & Breuer, C. (2011).
Family structure, time constraints and sport participation. European
Revue of Aging Physical Activity, 8, 57-66.
Scheerder, J., & Vos, S. (2011). Social stratification in
adults' sports participation from a timetrend perspective. Results
from a 40-year household study. European Journal for Sport and Society,
8, 31-44.
Smithson, M., & Verkuilen, J. (2006). A better lemon squeezer?
Maximum- likelihood regression with beta-distributed dependent
variables, Psychological Methods, 11, 54-71.
Sport England (2010). Understanding Variations in Sports
Participation. London: Sport England. Sport England (2012). Active
People Survey 5. London: Sport England. Retrieved from
http://www.sportengland.org/research/active_people_survey/active_people_survey_5 .aspx Stamatakis, E., & Chaudhury, M. (2008). Temporal
trends in adults' sports participation patterns in England between
1997 and 2006: The Health Survey for England. British Journal of Sports
Medicine, 42, 901-908.
Van Bottenburg, M. (2005). Sport participation in the EU: Trends
and differences. Retrieved from http://www.mulierinstituut.nl
Van Tuyckom, C., & Scheerder, J. (2010). A multilevel analysis
of social stratification patterns of leisure-time physical activity
among Europeans. Science & Sports, 25, 304-311.
Van Tuyckom, C., Scheerder, J., & Bracke, P. (2010). Gender and
age inequalities in regular sports participation: A cross-national study
of 25 European countries. Journal of Sports Sciences, 28(10), 1077-1084.
Wicker, P., Breuer, C., & Pawlowski, T. (2009). Promoting sport
for all age- specific target groups. The impact of sport infrastructure.
European Sport Management Quarterly, 9, 103- 118. World Health
Organization (WHO) (2010). Global Recommendations on Physical Activity
for Health. Retrieved from
http://www.who.int/dietphysicalactivity/factsheet_recommendations/en/
Themis Kokolakakis [1], Fernando Lera-Lopez [2], and Pablo
Castellanos [3]
Sheffield Hallam University [1]
Public University of Navarre [2]
University of Corunna [3]
Endnotes
(1) The data of "sports participation" are from the
period 2010/2011 whereas the data of "sports expenditure" are
from the period 2007/2010 (there were not data available for the period
2010/2011 at the time of carrying out our study); hence the variable
"sports expenditure" is lagged in relation to the variable
"sport participation." In this way, as past could affect
present but not vice versa, it is clear that in our circumstances there
is no chance for reverse causality (Humphreys & Ruseski, 2007). The
only feasible causal relationship we could study with our data is the
one which stems from "sports expenditure" to "sports
participation," so the first appears just as an explanatory
variable of the second. The potential reverse causality, shown by some
papers, between education or income and sports participation (e.g.,
Lechner, 2009) should be carefully considered in further analysis. As we
are considering exclusively regional variables, this potential reverse
causality is much less relevant.
(2) Both the full and detailed list of potential variables that
were considered in this study and their correlation coefficients are not
included in this paper for the sake of space. They can be obtained from
the authors upon request.
Themis Kokolakakis is a Senior Research Fellow at the Sport
Industry Research Centre. His research interests include sport/leisure
economics and the analysis of sport participation patterns.
Fernando Lera-Lopez is a Senior lecturer in the Department of
Economics. His research interests include sport economics and sports
participation.
Pablo Castellanos is a Senior lecturer in the Department of Applied
Economics. His research interests include sport economics, economic
policy, and social values of economic activity.
Table 1: Sports Participation. England 1987-2011, %
1987 1990 1993
At least one activity (excluding walking) once 44.7 47.8 47.3
every 4 weeks
At least one activity once every 4 weeks 60.7 64.5 63.7
At least three 30' activities per week, moderate -- -- --
intensity (KPI1)
1996 1999 2002
At least one activity (excluding walking) once 45.6 44.4 43.2
every 4 weeks
At least one activity once every 4 weeks 63.6 61.1 58.5
At least three 30' activities per week, moderate -- -- --
intensity (KPI1)
2005 2006 2007
At least one activity (excluding walking) once 47.0 48.3 --
every 4 weeks
At least one activity once every 4 weeks 65.8 68.2 --
At least three 30' activities per week, moderate 20.9 21.2 21.9
intensity (KPI1)
2008 2009 2010
At least one activity (excluding walking) once -- -- --
every 4 weeks
At least one activity once every 4 weeks -- -- --
At least three 30' activities per week, moderate 21.2 21.9 21.4
intensity (KPI1)
2011
At least one activity (excluding walking) once --
every 4 weeks
At least one activity once every 4 weeks --
At least three 30' activities per week, moderate 21.9
intensity (KPI1)
Source: General Household Survey, Active People Survey, SIRC
Note: From 1971 to 2002, sports participation figures were derived from
the General Household Survey (GHS). The main definition of sports
participation was "at least once in the last four weeks." In 2005,
Sport England developed the Active People Survey, a survey with a
bigger sample size including information about frequency and intensity
of sports participation. Despite consistency in sports definitions, the
series are not completely homogenous and comparable.
Table 2: Selected Variables and Descriptive Statistics
Mean Std. Dev. Var. Coef. (%)
1. Dependent variables
Sports participation: 46.27 4.48 9.6823
Percentage of adults (16+)
participating at a sport
activity at least 30 min.
(moderate intensity session)
during the last four weeks.
Regular sports participation: 22.27 3.29 14.7732
Percentage of adults (16+)
participating at a sport
activity at least 30 min.
(moderate intensity session) at
least three times per week.
2. Explanatory variables
2.1. Socio-demographic
variables
Single households: Percentage 19.8175 3.4139 17.2267
of one-member households in the
region.
Children: Percentage of 30.4746 3.6563 11.9979
households with at least one
child in the region.
Four or more adults: Percentage 11.4180 2.9828 26.1237
of households with four or more
adults in the region.
A-Levels: Percentage of people 17.7810 2.6088 14.6718
having an A-Level certificate
in the region.
Higher education: Percentage of 27.5323 9.0100 32.7252
people having a degree (as
their highest qualification) in
the region.
Longstanding illness: 23.6543 3.8742 16.3784
Percentage of people with a
limiting long lasting illness,
disability or infirmity.
Male: Percentage of people/ 40.4276 2.3945 5.9229
population being male in the
region.
Age 35/64: Percentage of 54.8096 3.0018 5.4768
people/population between 35
and 64 years old in the region.
Age over 64: Percentage of 28.4572 4.6292 16.2672
people/population over 64 years
in the region.
White: Percentage of people/ 92.6929 10.9147 11.7751
population being white in the
region.
Urban: Percentage of people/ 85.1043 15.9083 18.6927
population living in urban
areas in the region.
2.2. Economic variables
Council house: Percentage of 6.2961 4.6148 73.2962
people in the region living in
council houses (Council houses
are built and operated by Local
Authorities to supply
well-built homes on secure
tenancies at reasonable rents
to, primarily, working-class
people).
Income up to 20,799 [pounds 33.5227 8.0226 23.9318
sterling]: Percentage of
people with personal income
below 20,799 [pounds sterling]
per year in the region.
Income 20,800 [pounds sterling] 33.8473 4.2712 12.6190
-41,599 [pounds sterling]:
Percentage of people with
personal income between 20,800
[pounds sterling-41,599
[pounds sterling] per year.
Skilled occupation Non-manual: 16.9187 2.8327 16.7430
Percentage of people developing
skilled occupation
(non-manual) in the region.
Unskilled occupation: 2.9872 1.4558 48.7346
Percentage of people developing
unskilled occupation in the
region.
Full-time: Percentage of people 44.9449 4.4786 9.9646
working in full-time jobs in
the region.
Part-time: Percentage of people 13.4348 2.1556 16.0449
working in part-time jobs in
the region.
Looking after children: 3.9910 1.1825 29.6292
Percentage of people looking
after children and not working
in the region.
Student: Percentage of students 7.4113 2.8313 38.2025
as working status in the
region.
Unemployed: Percentage of 4.5358 1.9544 43.0883
unemployed people as working
status in the region.
Median earnings: Level of 412.4990 69.5110 16.8512
median gross weekly level of
earnings in the region.
Live and work same area: 57.2479 15.9102 27.7918
Percentage of residents in a
region that also work within
it.
2.3. Sport variables
Local funding 2007-2010: Total 6,532.9 10,892 166.7359
Lottery Funding, Exchequer
Awards and Capital Expenditure
in sport during the period
2007-2010 in LAs, 000s.
CPA: Number of stars per region 2.5670 1.0504 40.9194
of the Comprehensive
Performance Assessment (CPA).
It measures the Local Authority
performance in providing
services. The score has five
categories from zero to four
stars.
Sport volunteers: Percentage of 6.3086 1.5598 24.7250
people in the region who have
volunteered in sport-related
work in the last year (at least
one hour a week).
20 min 1 facility quality 24.5439 21.0374 85.7134
assured: Percentage of
population that have 20
minutes' access (on foot in
urban areas and by car in rural
areas) to three types of
facility (from pool, hall,
health & fitness, grass pitch,
synthetic turf pitch or golf
course) of which at least one
is Quality Assured.
20 min 3 facilities: Percentage 79.7480 10.4587 13.1147
of population that live within
20 minutes' travel time (on
foot in urban areas and by car
in rural areas) from three
types of facilities (without
any Quality Assured
restrictions).
20 min 1 facility: Percentage 98.5723 2.4218 2.4569
of population living within 20
minutes' travel time (on foot
in urban areas and bv car in
rural areas) from one sporting
facility.
2. 4. Others
Lake area: Total area of inland 9.1708 4.5567 49.6871
water in a region, including
lakes totally within its limits
and lakes that intersect its
10km radius based perimeter.
Coast length: Total length of 325.6930 417.8815 128.3053
coast in the region, including
sea and major rivers within
10km radius.
Temperature: An index on 1.9840 0.1106 5.5746
temperature variability in the
region based on maximum and
minimum values derived from 192
weather stations across the UK.
Rainy days: Number of days with 4.7960 0.1357 2.8294
a rain higher than 1mm over the
month of the interview in the
region.
Table 3: Model to Explain Sports Participation (Estimates)
Number of obs
Wald chi2(33)
Log likelihood = 806.2489 Prob > chi2
AIC
BIC
Variable Coef. Std. Err. z
Socio-demographic variables
Single households -.002422 .002999 -0.81
Children .000926 .002192 0.42
Four or more adults .000438 .002403 0.18
A-Levels .005702 .002551 2.23
Higher Education .009930 .001615 6.15
Longstanding illness -.007264 .002599 -2.80
Male .000432 .002358 0.18
Age 35-64 -.003316 .002955 -1.12
Age over 64 -.005982 .003123 -1.92
White .007052 .001206 5.85
Urban .002027 .000571 3.55
Economic variables
Council house -.001562 .001871 -0.83
Income up to -.003520 .001737 -2.03
20,799 [pounds sterling]
Income -.001543 .001754 -0.88
20,800 [pounds sterling]
-41,599 [pounds sterling]
Skilled_N on-manual .001179 .002352 0.50
Unskilled -.009241 .004947 -1.87
Full-time .003886 .003026 1.28
Part-time .002025 .004113 0.49
Looking after children .007661 .006118 1.25
Student .006798 .004049 1.68
Unemployed .003792 .004999 0.76
Live & work same area -.000196 .000491 -0.40
Sport variables (investment,
infrastructure)
Local funding 2007-2010 8.06e-07 5.51e-07 1.46
CPA .002381 .005397 0.44
Sport volunteers .014008 .005013 2.79
20 min 1 facility quality .000277 .000279 0.99
assured
20 min 3 facilities .000415 .000698 0.59
20 min 1 facility -.001226 .001397 -0.88
Other variables (weather,
etc.).
Lake area -.001017 .001366 -0.74
Coast length -8.83e-06 .000015 -0.59
Temperature -.174835 .071385 -2.45
Rainy days -.033891 .057718 -0.59
Constant -.428475 .583720 -0.73
Phi 400.5597 30.0707
= 354
= 793.29
Log likelihood = 806.2489 = 0.0000
= -1544.50
= -1412.94
Variable P>z
Socio-demographic variables
Single households 0.429
Children 0.673
Four or more adults 0.856
A-Levels 0.025 **
Higher Education 0.000 ***
Longstanding illness 0.005 ***
Male 0.855
Age 35-64 0.262
Age over 64 0.055 *
White 0.000 ***
Urban 0.000 ***
Economic variables
Council house 0.404
Income up to 0.043 **
20,799 [pounds sterling]
Income 0.379
20,800 [pounds sterling]
-41,599 [pounds sterling]
Skilled_N on-manual 0.616
Unskilled 0.062 *
Full-time 0.199
Part-time 0.623
Looking after children 0.211
Student 0.093 *
Unemployed 0.448
Live & work same area 0.690
Sport variables (investment,
infrastructure)
Local funding 2007-2010 0.143
CPA 0.659
Sport volunteers 0.005 ***
20 min 1 facility quality 0.320
assured
20 min 3 facilities 0.552
20 min 1 facility 0.380
Other variables (weather,
etc.).
Lake area 0.457
Coast length 0.554
Temperature 0.014 **
Rainy days 0.557
Constant 0.463
Phi
Notes: * Statistically significant at the 10% level
** Statistically significant at the 5% level
*** Statistically significant at the 1% level
Table 4: Model to Explain Regular Sports Participation (Estimates)
Number of obs
Wald chi2(32)
Log likelihood = 806.2489 Prob > chi2
AIC
BIC
Variable Coef. Std. Err. z
Socio-demographic variables
Single households .000303 .004597 0.07
Children -.005246 .002971 -1.77
Four or more adults -.001805 .003233 -0.56
A-Levels .004229 .003461 1.22
Higher Education .008693 .002151 4.04
Longstanding illness .004677 .003484 1.34
Male .006723 .003178 2.12
Age 35-64 .000361 .004038 0.09
Age over 64 -.000549 .004225 -0.13
White .006641 .001630 4.07
Urban -.001590 .000743 -2.14
Economic variables
Economic variables
Council house -.001760 .002526 -0.70
Income up to -.001006 .002524 -0.40
20,799 [pounds sterling]
Income -.004723 .002450 -1.93
20,800 [pounds sterling]
-41,599 [pounds sterling]
Full-time .010422 .004054 2.57
Part-time .005562 .005552 1.00
Looking after children .007676 .008213 0.93
Student .008141 .005509 1.48
Unemployed -.001628 .006815 -0.24
Median earnings .000095 .000241 0.39
Live & work same area .000673 .000676 0.99
Sport variables (investment,
infrastructure)
Local funding 2007-2010 -9.50e-07 7.54e-07 -1.26
CPA .008557 .007291 1.17
Sport volunteers .015540 .006793 2.29
20 min 1 facility quality .000261 .000378 0.69
assured
20 min 3 facilities .000328 .000930 0.35
20 min 1 facility -.001879 .003306 -0.57
Other variables (weather,
etc.)
Lake area .001479 .001806 0.82
Coast length -.000018 .000020 -0.90
Temperature -.194950 .097146 -1.98
Rainy days .142723 .076509 1.87
Constant -3.034523 .823151 -3.69
Phi 319.2353 24.0374
= 352
= 354.69
Log likelihood = 806.2489 = 0.0000
= -1586.85
= -1459.35
Variable P>z
Socio-demographic variables
Single households 0.947
Children 0.077 *
Four or more adults 0.577
A-Levels 0.222
Higher Education 0.000 ***
Longstanding illness 0.174
Male 0.034 **
Age 35-64 0.929
Age over 64 0.897
White 0.000 ***
Urban 0.032 **
Economic variables
Economic variables
Council house 0.486
Income up to 0.690
20,799 [pounds sterling]
Income 0.054
20,800 [pounds sterling]
-41,599 [pounds sterling]
Full-time 0.010
Part-time 0.316
Looking after children 0.350
Student 0.139
Unemployed 0.811
Median earnings 0.695
Live & work same area 0.320
Sport variables (investment,
infrastructure)
Local funding 2007-2010 0.208
CPA 0.241
Sport volunteers 0.022 **
20 min 1 facility quality 0.490
assured
20 min 3 facilities 0.725
20 min 1 facility 0.570
Other variables (weather,
etc.)
Lake area 0.413
Coast length 0.368
Temperature 0.048 **
Rainy days 0.062 *
Constant 0.000 ***
Phi
Notes: * Statistically significant at the 10% level
** Statistically significant at the 5% level
*** Statistically significant at the 1% level
Table 5: Models to Explain Sports Participation and Regular Sports
Participation
Sports Participation [delta]y/ Regular Sports Participation
Variable [delta]x Variable
A-Levels .001418 Children
Higher Education .002469 Higher Education
Longstanding illness -.001806 Male
Age over 64 -.001487 White
White .001753 Urban
Urban .000504 Income
20,800 [pounds sterling]
-41,599 [pounds sterling]
Income up to -.000875 Full-time
20,799 [pounds sterling]
Unskilled -.002298 Sport volunteers
Student .001690 Temperature
Sport volunteers .003482 Rainy days
Temperature -.043466
Sports Participation [delta]y/
Variable [delta]x
A-Levels -.000905
Higher Education .001500
Longstanding illness .001160
Age over 64 .001146
White -.000274
Urban -.000815
Income up to .001798
20,799 [pounds sterling]
Unskilled .002681
Student -.033114
Sport volunteers .024622
Temperature
