Participation in school sport and post-school pathways: evidence from Ireland.
Keywords: sports participation; labour market; educational choice
JEL Classifications: I21; J01; J24; J40; L83
An expanding body of evidence suggests that participation in sport is associated with better outcomes on a range of indicators of personal wellbeing (Physical Activity Guidelines Advisory Committee, 2008). While the majority of this research focuses on health outcomes, attention has also been given recently to the potential impact of participation on academic achievement and labour market outcomes. Studies to date, described in more detail below, are at least suggestive of a positive impact of participation in sport on both academic achievement and success in the labour market. If confirmed, these results have policy implications with respect to the resourcing and promotion of sport, both at the school level and more broadly.
Naturally, there are limits to the strength of inference one can make from the evidence as it stands. Almost all studies are based on surveys conducted in the United States of America. Many consider participants involved in interscholastic competition only, so some of the effects recorded relate to athletes rather than to active participants more generally. Furthermore, there are of course difficult econometric issues surrounding the validity of inferring causal relationships from survey data. As we argue below, these may be particularly severe when dealing with behaviours and outcomes that change substantially during adolescence.
In the present paper, our focus is on outcomes two years after leaving the secondary education system. We provide evidence that participation in sport at secondary school has an impact on the likelihood that after leaving secondary school a young person obtains employment, becomes unemployed or continues in education/training. The overall contribution of the present paper is fourfold. First, employing survey data from Ireland for the first time, we add to the weight of evidence that suggests a positive impact of school sport on later economic outcomes. Second, the survey we employ provides detailed records of participation across the secondary school years, which allows for a more sophisticated analysis of differential impacts according to the pattern of participation at school. Thus, although our survey is cross-sectional, it does contain a time-varying element in its design, which we exploit. In particular, this advantage allows us to focus not so much on whether individuals participate in school sport, but whether they drop out from sport before their final secondary school years.
Third, because the survey contains detailed background characteristics, indicators of the quality of students' relationship with their school and information about public examinations, we are able to control for and consider a number of potential mediating influences. Fourth, as well as standard regression techniques, we undertake Propensity Score Matching (PSM), which is increasingly employed as a way to identify treatment effects in research on educational outcomes. Although no available econometric technique is perhaps adequate to prove completely decisive as to whether participation in sport has a causal impact on later labour market outcomes, the fact that we find positive effects when employing PSM, which aims to reduce selection bias when estimating such effects, adds further to the weight of recent evidence.
The remainder of the paper is organised as follows. Section 2 reviews previous studies that offer insights into both the theoretical channels by which participation in sport might influence labour market outcomes and supporting empirical evidence. Section 3 briefly describes the Irish educational system, the data and the methods we employ. Section 4 presents our results and Section 5 concludes.
2. Previous literature
2. I Possible causal channels
Theoretical linkages between participation in school sport and labour market outcomes have been proposed in both economic and sociological literature. Since both disciplines have proposed plausible causal channels, with commonalities and contrasts, here we consider both perspectives.
Broh (2002) contrasts three approaches to the potential benefits of participation in school sport that are grounded in sociological theory, drawing in particular on the seminal work of Coleman (1961). The 'developmental model' holds that participation in sport instils in students characteristics that are likely to lead to success in other aspects of school life, including academic studies, and hence to long-term benefits beyond school. These characteristics include a strong work ethic, respect for authority and perseverance, each of which is arguably in keeping with educational values. The 'leading crowd' theory proposes that participation in sport leads to a greater likelihood that an individual's peers will be high achievers in school life, resulting in knock-on effects for success in other areas (Rehberg and Schafer, 1968). More recent sociological accounts emphasise the role of sport in increasing the 'social capital' of students and, potentially, their parents. Participation in school sport may improve social ties between students and teachers, between students and parents, between different sets of parents, and between parents and teachers, perhaps increasing social control over behaviour and assisting the dissemination of educationally beneficial information (Broh, 2002). Within this sociological literature, the primary effects proposed relate to the potential of sport to improve the relationship that the student has with the school. Benefits for labour market outcomes arise mainly via improved educational performance.
The theoretical channels identified in the relevant economic literature overlap with those channels outlined above, although more recently the precise causal mechanisms linking participation in sport to labour market outcomes have become more developed and nuanced. An important distinction concerns the ultimate source of the benefit produced. For instance, economists and sociologists tend to view social capital in different ways, notably in the extent to which social capital is created by individual investment and/or by behaviour at a group level (see Glaeser, Laibson and Sacerdote, 2002). A similar distinction can be made in relation to the concepts of human capital and health capital (Grossman, 1972). Following foundational work by Schultz (1960) and Becker (1964), the investment model of education views educational decisions as an optimisation problem involving a trade-off between present income foregone and discounted future income arising from investment in human capital. In these economic models, the individual is the primary decision-maker with respect to how much to invest in search of future benefit, while the sociological literature often views benefits to individual students who participate in sport as the objective of actions undertaken by schools, parents or policymakers. Thus, the beneficial characteristics invoked by the developmental model in sociology are echoed in economic literature's recent emphasis on non-cognitive skills (e.g., Heckman and Rubinstein, 2001), considered as a form of investment in human capital (see Lechner and Downward, 2013). The causal channel is partially the same, in that participation in sport may enhance certain individual characteristics, yet partially different, because in the latter case it is not the group responsible for young people but the person's own incentives that generate the behaviour.
Turning to more recent economic literature, Lechner (2009) distinguishes between channels that propose direct impacts of participation in sport on productivity, benefits arising from enhanced social networks, and gains from signalling to employers. The primary direct effects are the possibilities that playing sport improves cognitive abilities, non-cognitive skills and/or health and fitness, with knock-on effects for both educational attainment and later productivity in the workplace (see Lechner, 2009, and references therein). It is possible that participation in sport acts indirectly, as a signal in a labour market with imperfect information, in similar manner to that envisaged for educational attainment by Spence (1973). If employers believe that playing sport is linked to desirable characteristics, which its frequent inclusion on CVs would indicate is a possibility, then it may open up greater job opportunities for applicants who have been or remain participants (Lechner, 2009; Rooth, 2011). Furthermore, assuming the additional fitness that comes with participation in sport also has an impact on personal attractiveness and weight, participants might also benefit from the existence of the so-called 'beauty premium' in the labour market (Hamermesh and Biddle, 1994; Mobius and Rosenblat, 2006), or from avoiding a penalty for being overweight (e.g., Morris, 2006). This latter explanation speaks to one of the empirical difficulties involved: the possibility of unobservable characteristics associated with both participation in sport and performance in the workplace. For researchers and policymakers, an important issue is whether desirable characteristics are caused by participation in sport or merely make individuals more likely to participate in the first place; for the employer, all that matters is the association, assuming that it is genuine. Finally, while many human capital models do explicitly consider time constraints, many studies do not consider the argument, believed by at least some parents, that participation in sport may harm academic achievement (and hence labour market prospects) by eating into study time and providing a distraction. Below we present descriptive data that is consistent with this belief being widely held.
Most of the potential causal channels linking participation in sport to labour market outcomes are also likely to lead to improvements in academic performance. However, two channels operate independently of academic performance: the possibility that employers are keener to hire workers who are attractive and the proposal that employers believe that involvement in sport is a valid signal of a worker's quality. These theories imply that employers will be keener to hire participants in sport over non-participants, for a given level of educational qualifications.
Although it does not feature prominently in previous literature, we also consider the possibility that participation in sport improves young people's liking for educational institutions and may, therefore, influence their decisions regarding whether to continue in higher or further education. Those who enjoy sport may find it easier to continue to do so if they remain in fulltime education, with the additional time, flexibility and sporting opportunities conferred by college life relative to working life. From the perspective of human capital models that optimise consumption over the life-course, easy opportunities to play sport in the short or medium term may appear like a minor consideration. However, there is evidence that immediate consumption experiences can be strong influences on young people's educational decisions (Alstadsaeter, 2011; Jacob, McCall and Stange, 2013). Moreover, time preferences may be present biased and inconsistent (Laibson, 1997; O'Donohue and Rabin, 1999). We return to this possibility in the final section.
2.2 Empirical evidence
There is a more substantial literature that addresses how participation in sport affects academic achievement than how it affects labour market outcomes, although the two are strongly linked. The association with academic outcomes has at this stage been the subject of numerous studies and several meta-analyses that approach the question from different disciplinary perspectives and, as a result, with somewhat different methodological priorities. Studies that have sought more rigorously to identify causal channels are far less common.
Systematic reviews from the public health perspective conclude that increased physical activity at school, of which participation in school sport is only one component, tends to be positive for academic performance, despite the associated reduction in time spent on academic study (Taras, 2005; Trudeau and Shephard, 2008; Centers for Disease Control and Prevention, 2010; Singh et al., 2012). Shulruf (2010), ho we ver, focuses his meta-analysis of 29 studies not on the direction of the effect, but on effect size. For the association between participation in extra-curricular sport and various learning outcomes, he concludes that while recorded effects are almost entirely positive, effect sizes are generally low (Cohen's d of less than 0.31). Contrastingly, the meta-analysis of Bird, Tripney and Newman (2013) finds substantial effect sizes for participation in organised sport, especially in relation to improved numeracy scores (Hedge's g of 0.19-0.80). This analysis includes just six studies (only one of which is common to Shulruf, 2010), which meet more stringent criteria for the identification of control comparison groups. Thus, the analysis concentrates more on the evaluation of small-scale interventions via well controlled trials. The authors conclude with "reasonable confidence" that participation in sport leads to improved educational performance.
The above analyses, while suggestive, cannot be considered conclusive. The results are not consistent; a number of studies report small or non-significant effects. Several meta-analyses in the public health literature place greater weight on controlled trials which, when properly conducted, are good methods for identifying causal effects, but which produce findings specific to the particular interventions tested. Different interventions may well have differential effects on academic performance, so the extent to which the results generalise to participation in sport is uncertain. Moreover, the studies covered are overwhelmingly based on a limited number of American data-sets, together with two British studies covered in Bird et al. (2013). Pfeifer and Cornelifien (2010) further note that the relationship between sports participation and attendance at schools and colleges in the US differs from that in most European countries. They nevertheless find significant effects of participation in school sport on educational attainment in the German Socio-Economic Panel (GSOEP). Overall, therefore, there is a need for further studies across a greater number of nations and data-sets.
Direct empirical examinations of the link between participation in sport and labour market outcomes are less common. It is well established that rates of participation in sport are higher among people with higher educational attainment, income and occupational status (e.g., Farrell and Shields, 2002; Cerin and Leslie, 2008) and that the gap in participation between those of higher and lower socio-economic status widens in adulthood, especially young adulthood (Lunn, 2010). Yet, it is of course harder to establish the direction of the underlying causal relationship, if any.
A number of studies have found positive effects of participation in sport on labour market outcomes using US longitudinal data. Long and Caudill (1991) found that introducing a dummy variable for participation in college athletics into wage equations nine years later produced a wage premium for athletes of 4 per cent. A series of studies has found generally positive impacts for men of participation in high school sport using a single wave of the National Longitudinal Survey of Youth (NLSY). Participants in sport at school (measured by recall in 1984, three years after graduation) have been found 8-11 years later to have: (i) higher wages (Barron, Ewing and Waddell, 2000); (ii) a higher likelihood of being a supervisor, union member or recipient of performance related pay (Ewing, 1998); and (iii) a higher likelihood of receiving seven fringe employment benefits, such as health insurance, profit sharing or
paid vacations (Ewing, 2007). Recent work by Cabane and Clark (2013), using the first (1994-5) wave of the National Longitudinal Study of Adolescent Health (NLSAH), finds that participation in sport is related to better maths and science grades, but also to a higher likelihood of being a supervisor or having autonomy in the job. Looking across these studies on US data, it is noteworthy that the impact of participation on the likelihood of having a job at all tends to be non-significant and that the recorded effects apply mostly to men.
In perhaps the only longitudinal study conducted on non-US data, Lechner (2009) finds that adult participation in sport increases wages in the GSOEP by some 5-10 per cent, and also that it increases the likelihood of fulltime employment for women. Lechner's study employed a measure of the level of participation in sport, rather than a simple binary indicator, and used propensity score matching to control for selection into sport. Its focus was adult rather than school sport, however.
Lastly, there is direct evidence that participation in sport may be linked to labour market outcomes via channels that do not include educational performance. Rooth (2011) conducted a correspondence test in Sweden using matched CVs that differed only by mention of sporting participation. He recorded higher call-back rates for participants, especially in occupations that required physical effort. This finding is in keeping with the idea that sport is used as a signal of something by potential employers, though it is unclear whether it signals good health, attractiveness, non-cognitive skills, or some other quality that employers value. These are all possibilities. Currie and Madrian (1999) review literature relating health to labour market outcomes and find the relationship mostly to be positive, but not that strong. Cawley (2004), Morris (2006) and Rooth (2009) confirm an obesity penalty, while evidence in Hamermesh and Biddle (1994) and Mobius and Rosenblat (2006) supports a beauty premium.
To the best of our knowledge, no previous study has looked explicitly at the influence of participation in sport on immediate post-school pathways. Other empirical results, however, do suggest that participation in sport may be linked to the propensity to stay on at school (Shulruf, 2010), to individual's aspirations to go to college (Rees and Sabia, 2010) and to expectations of obtaining a degree (Lipscomb, 2009). These findings are consistent with the theory that sport may influence educational choice and, hence, have associated impacts on labour market outcomes.
2.3 Econometric issues
While the aforementioned studies are suggestive of positive influences of participation in sport on labour market outcomes, there are difficult econometric issues surrounding a number of the studies cited. All recognise that there are problems of selection into participation and hence that participation in sport cannot be treated as exogenous with respect to models of educational attainment or labour market outcomes.
Different authors have attempted different methodologies, with the most common approaches being instrumental variables and fixed effects models. In the former case, the difficulty is to locate suitable instruments that are related to participation in sport but unrelated to the outcome of interest. These have included school characteristics (Barron et al., 2000) and height (Pfeifer and Cornelifien, 2010; Rees and Sabia, 2010), neither of which might be thought ideal for the purpose, since both may well be independently associated with the outcomes of interest. In the case of fixed effects models, as well as the usual issues surrounding the sufficiency of variability within the data, the difficulty is the possibility that unobserved heterogeneity may exist not only with respect to characteristics that can reasonably be considered consistent over time, but also to characteristics that are not time consistent. Adolescence is a period of change. Young people vary in speed of maturity with respect to intellectual, physical and social development, such that changes over time in, for instance, athleticism and educational performance, may be driven by underlying unobserved characteristics.
These methodological difficulties, acknowledged by most authors, may go some way to explaining contrasting results obtained with the same data-sets. For instance, using a fixed effects approach and different instruments with the same NLSY data as used by Ewing (1998), Kosteas (2011) failed to obtain a significant effect of participation in high-school sport on being employed later in a supervisory role. Rees and Sabia (2010) argue that the use of fixed effects and instrumental variables techniques on the NLSAH data greatly reduces or eliminates the effect of participation on academic performance, although using the same data Cabane and Clark (2013) find that the effect survives controlling for selection through school fixed effects and the comparison of siblings.
In the analysis that follows, we first estimate multinomial logit models, then employ Propensity Score Matching (PSM). Lechner (2009, p. 846) argues that PSM has advantages over fixed effects models in being more suitable for estimating influences on binary outcomes and making fewer assumptions about the absence of potentially important time varying characteristics. On the other hand, PSM has its own potential shortcomings. In particular, the reliability of any PSM estimate depends on the validity of the Conditional Independence Assumption (CIA), i.e., selection into the treatment is based solely on observable factors and that all characteristics that simultaneously impact both the treatment and outcome variable are observed. This is a strong assumption to make, because it is difficult to eliminate the possibility that estimates are influenced by one or more unobserved factors that simultaneously affect both the treatment and outcome variable. Thus, unobserved heterogeneity is an issue. We nevertheless employ PSM here as perhaps the most suitable available methodology for reducing selection bias in circumstances where we have time varying data on participation in sport and cross-sectional data on the outcome of interest. However, we also utilise the diagnostic test proposed by Rosenbaum (2002) to test the robustness of our PSM estimates to the influence of unobserved heterogeneity.
3. The Irish education system, data and methodology
3.1 The Irish education system
The Irish education system consists of five main components: (i) early childhood, (ii) primary, (iii) secondary (referred to in Ireland as 'second level' or 'post-primary'), (iv) further education and training, and (v) higher education. Students are usually twelve years of age when they transfer from primary to secondary education. The secondary system consists of a three-year lower secondary programme, at the end of which students sit a national public examination, the 'Junior Certificate'. Students are usually aged around fourteen when they complete this examination. This lower secondary programme is followed by an optional one-year 'Transition Year' (TY) programme (1) and then a two-year upper secondary programme culminating in a second national public examination, the 'Leaving Certificate'. Performance in this exam determines entry to the higher education system, which covers undergraduate and postgraduate courses at universities, institutes of technology, colleges of education and some specialist third-level institutions. The further education and training sector covers educational courses undertaken after completing secondary education, but which are not part of the higher education system.
In this paper, our analysis covers the impact of sports participation at secondary school on students' progression into further and higher education or training, the labour market, or unemployment/ inactivity. Hereafter, we adopt the useful shorthand from the Irish system of the 'Junior Cycle' and 'Senior Cycle', which refer to the first three years of study for Junior Certificate (plus TY) and the last two years of study for the Leaving Certificate respectively.
Our data come from the 2007 Irish School Leavers' Survey (SLS), conducted by the Economic and Social Research Institute (ESRI) for the Department of Education and Skills (DES). Administered in the Spring/ Summer of 2007, the survey captured information on school experiences and post-school pathways of a nationally representative sample who finished secondary education in the academic year 2004/5. The effective sample size was 2,025 respondents. Data were collected through a multi-mode response method, where students could complete the survey online, via post or via face-to-face or phone interviews.
The SLS collected rich information on students' personal characteristics (e.g., gender, age, ethnicity, health status, birth location and residential location) and social backgrounds (parents' socio-economic status, employment status and educational attainment), along with comprehensive data on educational attainment, examination results, (2) school experiences (e.g., teacher interaction, class context, work experience and/ or part-time work while at school), Transition Year participation, receipt of private tuition and truancy in the final school year. The SLS also recorded information on post-school pathways, including employment, unemployment, apprenticeships, training, further education, higher education and/or inactivity (i.e., home duties or illness/disability). This information was captured in calendar format from June 2005 to July 2007. In addition, the SLS collected information on school characteristics, including school type and size, gender mix and whether the school had designated disadvantaged status. (3)
As part of the questionnaire, students were asked whether during their secondary education they were "involved in any regular sporting activities, not counting PE lessons?". They were instructed to include exercise activities such as dance or gym, in addition to traditional sports such as soccer, rugby, Gaelic games, (4) swimming, etc. It is important to note that the questions related to sport undertaken at school (i.e., extra-curricular sport) only and not extra-school sport (i.e., participation outside of school in sports clubs, leisure centres, etc.). For those who indicated that they did participate, subsequent questions obtained the types of activities undertaken, the school years they participated in and whether the students were still involved in the specified activities. One of the main advantages of the SLS is that it recorded participation across school years, allowing us to exploit individual-level transitions into and out of sport during secondary education.
Thus, our main measures of participation in sport rely on an individual's ability to recall their participation by school year, two years after leaving secondary school. The use of such recall data for measuring participation in sport is not uncommon (see, for example, Barron, Ewing and Waddell, 2000; Ewing, 2007). Here, however, we not only asked whether individuals participated, but also in what years. Hence, we decided to perform a check on the reliability of the recall data by comparing it to a nationally representative survey of schoolchildren's sport conducted in 2009, two years following the SLS. The pattern of participation in extra-curricular sport recorded in the Children's Survey of Sports Participation and Physical Activity (CSPPA) is compared with that recorded by the SLS in figure 1. The match is reasonably close and suggests that the recall data are reliable.
This pattern of participation has two other important properties. First, substantial numbers of participants drop out of sport across secondary school years. Second, this pattern is strongly related to the years (three and six) of public examinations. This latter effect suggests that students, parents and/or teachers may believe that participation in sport in exam years can detract from academic performance. Both factors mean that students differ considerably in patterns of participation across secondary school years. Potentially, this variation may be important to later outcomes. In the studies cited above, participation in school sport is generally recorded as a binary variable. In contrast, we split the participants into two groups: those who participated at some point in only the first four years (Junior Cycle) but dropped out, and those who participated in their last two years (Senior Cycle).
The primary aim was to identify the impact of sports participation on whether those who left school during the academic year 2004/5 had (i) entered employment (including apprenticeships), (ii) gone on to further education and training or higher education, or (iii) become unemployed or inactive, two years after finishing secondary school. We first estimated Multinomial Logit (MNL) models, where our dependent variable consisted of these three outcomes. MNL models were employed because the decision being assessed consisted of three mutually exclusive outcomes with potential correlation between the error terms. A clear drawback of MNL, however, is that it does not address treatment group selection, where 'treatment' refers to sports participation. (5) In an attempt to address this, in addition to estimating MNL models, we employed Propensity Score Matching (PSM), which controls for selection into participation in sport by matching treatment group individuals with a control group of individuals with similar observable characteristics who did not undergo the treatment. We report both MNL and PSM analyses, since each requires assumptions the other does not and because, while the PSM approach is designed to reduce selection bias when sports participation is determined by observable factors, the MNL analysis provides additional insights into the roles of other covariates.
PSM is a non-parametric, two-step procedure. Firstly, each individual's probability (or propensity score) of receiving the treatment (participating in sport at school) is examined conditional on a set of explanatory variables, allowing treatment and control group individuals to be matched by propensity score. This is equivalent to matching on key characteristics of the treatment group (Rosenbaum and Ruben, 1983). Secondly, the outcome variable (employment; further or higher education or training; unemployed/inactive) is compared across treatment and control groups.
Results reported below derive from one of the most common PSM estimators--nearest neighbour with replacement. There are alternative matching estimators available when undertaking PSM, e.g., radius matching, kernel matching, etc. As sample size increases, all PSM estimators should yield similar results, as each estimator nears precise matching (Caliendo and Kopeinig, 2005). With smaller samples, the choice of matching estimator is usually based on a trade-off between bias and variance. Given our relatively modest sample, we focus on minimising bias and report single nearest neighbour matching without caliper. Nevertheless, we ran sensitivity checks on the matching procedure by comparing results across kernel (normal, biweight, uniform and tricube), radius and multiple neighbour matching estimators. In most cases the magnitude of the estimated treatment effect remained the same, but statistical significance increased relative to single nearest neighbour matching. In particular, the 'unemployment/inactivity' outcome became significant with other matching estimators. This is unsurprising, because the alternative estimators use more information (i.e., control observations) to match on. (6) The result may warrant further investigation, but because it is sensitive to the matching procedure employed and our sample is relatively small, we focus on the results that are consistent across estimators and we report those from single nearest neighbour matching.
We also employed several diagnostic tests in relation to the PSM models. We checked that our data was balanced through the 'ptesf procedure in Stata. The post balancing test did not generate a wholly insignificant result for the PSM model comparing drop-outs from sport to those who continued to participate in Senior Cycle (see Section 4). However, the pseudo [R.sup.2] statistic fell from 0.150 to 0.035 under the nearest neighbour with replacement estimate, which suggests that remaining differences between treated and control group populations within the matched sample were quite small. The other PSM models passed this test. We also employed a test for sensitivity to unobserved heterogeneity (Rosenbaum, 2002), via Stata's 'mhbounds' procedure (Becker and Caliendo, 2007). We began by assuming zero bias, then introduced an unobserved factor that simultaneously affected the outcome variable and the likelihood of allocation to the treatment group by 10 per cent, to see if our effects remained statistically reliable. In models where the observed impact was negative, we tested the sensitivity of our results to negative selection bias; in models where we derived positive effects, we assessed sensitivity to positive selection bias. This analysis revealed that unobserved effects would have to be substantial for our estimated effects to become questionable. (7)
In relation to covariates included in the models, we estimated three distinct specifications to identify the incremental impact of additional variables, check the stability of coefficients and deal with potential problems of colinearity. The basic specification consisted of individual and background characteristics; specifically, gender, age, illness/disability status, socio-economic background, (8) mother's education, Transition Year participation, receipt of private tuition in final school year, employment during and out of term-time (9) and school type. The second specification included 'school mediating factors', as discussed further in the next paragraph. The final specification introduced the last examination sat by a student before they left secondary school as a control for academic achievement.
Since one of the effects we were interested in was the degree to which participation in sport might promote an attachment to an individual's school or to educational institutions generally, we made use of a set of questions available in the SLS that gauged how much students liked their school experience. We measured individual motivation at school by whether students agreed that "school work was worth doing", opinions of teachers by similar responses to the statement "my teachers didn't care about me", and opinions of the classroom environment by responses to the statement "there were too many troublemakers in my classes". To these attitudinal measures of attachment to the school we added a behavioural one: the frequency with which students played truant. We employed this collection of variables to control for positive disposition towards the school and we refer to them as "school mediating factors".
As indicated in section 3, we first modelled economic status two years after leaving school by Multinomial Logistic (MNL) regression. The results are presented in tables 1 and 2. The base category for the dependent variable is being in employment and, hence, the reported coefficients relate to the influence of each covariate on the likelihood of being unemployed or inactive (table 1) and continuing in education or training (table 2) following completion of secondary education. This basic specification contains controls for the background characteristics outlined above, including parental occupational class and maternal educational attainment, as well as a range of school characteristics.
The pattern of participation in extracurricular sport is associated with post-school pathways, especially the likelihood of progressing to further education and training or higher education (table 2). Individuals who participated in sport during Senior Cycle (the final two-year secondary school programme) are significantly more likely than non-participants to progress to further education and training or higher education rather than join the labour market. This result does not hold for those who played sport at school during Junior Cycle (first three years plus TY) but then dropped out: the estimated coefficient for this group is, in fact, negative. The implication of this result is that there is a difference between those who merely participate in sport at school and those who participate but drop out during their time at school. The relationship between sports participation in Senior Cycle and the likelihood of being unemployed is also marginally significant in the basic specification (table 1, p = 0.053).
The second model introduces variables intended to capture the potential mediating effects of the quality of the relationship between the student and their secondary school, including measures of truancy, motivation and negative or positive perceptions regarding the teachers and classroom. The introduction of these variables has a clear but modest effect on the sports participation coefficients. Specifically, the estimated impact of participation in sport in Senior Cycle on the likelihood of unemployment ceases to be significant; while its impact on the likelihood of progression to further education and training or higher education continues to be significant at the 1 per cent level, but is reduced by around one fifth. The third specification adds variables to control for the highest level of examination completed at secondary school, which leads to a further reduction in the estimated coefficients, but the impact of sports participation in Senior Cycle on the likelihood of progressing to further education and training or higher education continues to be statistically significant at the 5 per cent level. (10)
Since our sample size is relatively modest for an analysis of this type, it is important also to consider the estimated effect sizes and their confidence limits, which are provided as marginal effects estimated at the mean level of the covariates in table 3. In the basic specification, without the inclusion of mediating factors indicating the quality of the relationship with the school or the last official exam sat before leaving secondary education, participation in sport during Senior Cycle is associated with an increase in the likelihood of continuing in education of more than 10 per cent, with reductions in the probability of employment and unemployment of 5 per cent and 6 per cent respectively. Once the variables for the quality of the student-school relationship are introduced, these figures fall to 8 per cent (significant), 4 per cent (non-significant) and 4 per cent (significant), then to 6 per cent (significant), 3 per cent (non-significant) and 3 per cent (non-significant) when the level of final examinations is specified.
Some previous studies have reported differential effects of participation in sport by gender and by type of sport played (e.g., team versus individual, competitive versus non-competitive). We reran the above models with interaction terms included for participation in sport by gender and, separately, with the categorical participation variable split by type of sport. No robust and statistically significant effects were obtained. (11)
While the results presented thus far are indicative of a relationship between participation in sport and an individual's post-school pathway, there are issues of selection into participation in sport and associated problems regarding inferring causality. The findings above imply that those who participate in sport in Senior Cycle are more likely to continue in education or training, even after controlling for the quality of their relationship with the school and the level of final examinations sat. Similar effects are not found for those who participated in sport during Junior Cycle but then dropped out of sport. Yet it is possible that this pattern does not reflect a causal role for ongoing participation in sport. Those who opt to continue to play in Senior Cycle may already be more likely than those who do not to go onto further education and training or higher education for other reasons, including perhaps greater enjoyment of the educational environment.
As described in the previous section, we attempt to reduce the potential influence of selection bias by employing PSM, where different patterns of participation in sport correspond to different 'treatments'. Participation in extra-curricular sport at secondary school in Ireland is voluntary not compulsory and, hence, no intervention could hope to apply this treatment broadly across the whole student population. We are consequently interested in illuminating the potential for marginal gain from increasing voluntary participation in sport at secondary school, rather than trying to estimate an effect that might apply to the full population of students (i.e., students who did and did not play sport). Thus, we focus on the average treatment effect on the treated (ATT) as opposed to the average treatment effect (ATE). We conduct all three possible matches between the three groups, i.e., non-participants, participants who dropped out and participants in Senior Cycle. Where the non-participants are involved in the match, participation is taken to be the treatment and non-participation the control. Where ongoing participants (i.e., Senior Cycle participants) are matched with those who dropped out, dropping out is the treatment and ongoing participation the control. We also examine two specifications of the matching equation: the basic specification and one that includes variables for the quality of the student-school relationship.
The PSM estimates are provided in tables 4, 5 and 6. Regardless of which matching variables are employed, we estimate that compared with individuals who do not participate in secondary school sport, those who participate in their final secondary school years are significantly more likely to continue in education after leaving secondary school, with an ATT of 11 percentage points (table 4). This is, clearly, a substantial effect size.
The effect diminishes and becomes non-significant, however, when we match also based on variables that measure the quality of the student-school relationship. While a negative result is derived for the impact of Senior Cycle sports participation on employment, the effect is only marginally significant and becomes insignificant when we control for school mediating factors. Matching those who participated in sport only earlier in the school career with non-participants confirms no significant effect on post-school pathways (table 5). Lastly, we matched individuals who participated only early in their school career with those who participated in Senior Cycle to examine drop out from sport as a treatment (table 6). The match indicates a reduction of 9 percentage points on the likelihood of continuing in education beyond secondary school.
The present analysis adds to the weight of evidence that participation in sport may have beneficial effects on labour market outcomes. To previous studies conducted almost exclusively on data collected in the USA, we add evidence based on a survey of recent secondary school leavers in Ireland and evidence that the pattern of participation in sport across school years may be important. We find that participants in extracurricular sport are more likely to continue their formal education after leaving secondary school, as opposed to joining the labour market, provided they continue to participate in their final school years. Based on the point estimates across our models, which control for selection based on a range of variables that includes parental occupational class, mother's education and school characteristics, we estimate the main effect size to be in the region of 9-14 percentage points (tables 4 and 6). There is some suggestion in our findings that the effect may be mediated by the overall quality of the student-school relationship and by exam performance, consistent with potential causal channels identified in the economic and sociological literature described in section 2.
Our study is unusual in its focus on post-school pathways and the decision to continue in education beyond secondary level. The findings may relate to those that suggest participants in sport go on to enjoy better jobs and higher earnings (Barron et al., 2000; Ewing, 2007), assuming that staying on in education leads to improved labour market prospects afterwards. Implicit in this suggestion is the idea that young people may not make educational choices as described by traditional models based on optimal levels of human capital accumulation for the generation of utility derived from (time-consistent) discounted future consumption. Rather, the attachment to educational institutions and the extent to which it appears to be linked to participation in sport is at least suggestive that going to college provides consumption value and that the immediate school and college experience may be an important element of educational choice. This possibility does not negate other channels linking participation in sport to labour market outcomes, such as through its potential effects on health, attractiveness or non-cognitive skills (Lechner, 2009; Rooth, 2011).
We do not obtain reliable statistically significant effects for the association between participation in sport and the likelihood of having a job. While our point estimates are positive and a larger sample might find such an effect, this non-significant finding is nevertheless in line with some previous studies (Barron et al, 2000; Cabane and Clark, 2013). Our analysis is based on a much shorter duration between the measure of participation in sport and the measure of employment status, which is taken just two years after leaving school.
Although our sample size is modest for studies of this nature, the estimated effect sizes are substantial and the main results statistically significant at conventional levels. This may reflect our strategy of isolating participants who continued to play sport in their later secondary school years, but another possibility worth considering is that the timing and location of the present survey may have had an influence. Specifically, the behaviours and decisions that generate our data occurred in Ireland between 2000 and 2007, which was a period of economic growth that was extraordinary by the standards of developed economies. One potential side-effect of Ireland's mighty boom was that labour market opportunities for young people were unusually attractive, in terms of both the likelihood of employment and wages on offer. It is therefore possible that the effects we record are affected by additional variation in our dependent variable and its relationships with covariates. We have no way presently to test this. Future research might estimate the impacts of participation in sport at secondary school on post-school pathways elsewhere. Our data-set also precluded us from taking into account sport played outside of school. Taking this omitted variable into account might provide a more detailed picture of the potential causal channels underpinning the effects we report. Future research might test whether the effects, which suggest that sporting opportunities may increase the attractiveness of continuing in education, are influenced by the extent to which students engage in sport outside of the education system prior to leaving secondary school.
Lastly, we turn to policy implications. It is not uncommon that studies reporting beneficial effects of participation in sport conclude by proposing increased levels of public investment in sporting facilities and sporting opportunities. Indeed, we concur that the present analysis suggests the potential for returns from such investment, since it adds further weight to evidence that participation in sport may bring not only public health benefits, but economic ones too. However, it is important to note, first, that econometric considerations mean that we cannot be completely confident that the effects reported in this study and others like it are causal and, second, that our findings do not relate to interventions that can be applied broadly across students. We estimate benefits associated with voluntary participation in extra-curricular sport and, hence, the treatment effect only relates to students who respond to opportunities for voluntary participation. Policy might potentially provide more such opportunities and encourage participation, on the assumption that the estimated effect would apply to marginal voluntary participants who responded to the policy. In addition to this assumption, however, it needs to be recognised that interventions designed to raise voluntary levels of participation in sport meet with mixed success (see, for example, Taskforce on Community Preventive Services, 2001; Lunn, 2007, pp. 16-19).
Appendix table Al. Impact of Senior Cycle sports participation relative to non-participation on 2004-5 school leavers' post-school pathways two years after leaving school (May 2007): PSM matching estimator sensitivity check (basic specification) Nearest-neighbour Single Caliper (5) Employment (includes Apprenticeships) Sports Participation (Ref: No) Senior Cycle -0.093 ** -0.084 *** (0.041) (0.032) Unemployment/lnactivity Sports Participation (Ref: No) Senior Cycle -0.018 -0.037 ** (0.020) (0.017) Student/Training Sports Participation (Ref: No) Senior Cycle 0.111 *** 0.121 *** (0.040) (0.032) Observations 1,528 1,528 Epan Normal Employment (includes Apprenticeships) Sports Participation (Ref: No) Senior Cycle -0.073 ** -0.080 *** (0.030) (0.029) Unemployment/lnactivity Sports Participation (Ref: No) Senior Cycle -0.038 ** -0.042 *** (0.017) (0.016) Student/Training Sports Participation (Ref: No) Senior Cycle 0.112 *** 0.122 *** (0.029) (0.029) Observations 1,528 1,528 Kernel Biweight Uniform Employment (includes Apprenticeships) Sports Participation (Ref: No) Senior Cycle -0.071 ** -0.077 *** (0.030) (0.029) Unemployment/lnactivity Sports Participation (Ref: No) Senior Cycle -0.038 ** -0.039 ** (0.017) (0.016) Student/Training Sports Participation (Ref: No) Senior Cycle 0.109 *** 0.116 *** (0.029) (0.029) Observations 1,528 1,528 Radius Tricube Caliper (5) Employment (includes Apprenticeships) Sports Participation (Ref: No) Senior Cycle -0.070 ** -0.071 *** (0.030) (0.026) Unemployment/lnactivity Sports Participation (Ref: No) Senior Cycle -0.038 ** -0.102 *** (0.017) (0.014) Student/Training Sports Participation (Ref: No) Senior Cycle 0.108 *** 0.173 *** (0.029) (0.026) Observations 1,528 1,528
Alstadsaeter, A. (2011), 'Measuring the consumption value of higher education', CESifo Economic Studies, 57, pp. 458-79.
Barron, J.M., Ewing, B.T. and Waddell, G.R. (2000), 'The effects of high school athletic participation on education and labor market outcomes', Review of Economics and Statistics, 82, pp. 409-21.
Becker, G.S. (1964), Human Capital. A Theoretical and Empirical Analysis with Special Reference to Education, Chicago, University of Chicago Press.
Becker, S. and Caliendo, M. (2007), 'Sensitivity analysis for average treatment effects', The Stata Journal, 7, pp. 71-83.
Bird, K.S., Tripney, J. and Newman, M. (2013), 'The educational impacts of young people's participation in organised sport: a systematic review 'Journal of Children's Services, 8, pp. 264-75.
Broh, B.A. (2002), 'Linking extracurricular programming to academic achievement: who benefits and why?', Sociology of Education, 75, pp. 69-95.
Cabane, C. and Clark, A.E. (2013), 'Childhood sporting activities and adult labour-market outcomes', CEP Discussion Paper No. 1253.
Caliendo, M. and Kopeinig, S. (2005), 'Some practical guidance for the implementation of propensity score matching', IZA Discussion Paper No. 1588.
Cawley, J. (2004), 'The impact of obesity on wages', Journal of Human Resources, 39, pp. 451-74.
Centers for Disease Control and Prevention (2010), The Association Between School Based Physical Activity, Including Physical Education, and Academic Performance, Atlanta, GA, U.S. Department of Health and Human Services.
Cerin, E. and Leslie, E. (2008), 'How socio-economic status contributes to participation in leisure-time physical activity', Social Science and Medicine, 66, pp. 2596-609.
Coleman, J.S. (1961), The Adolescent Society, Glencoe, IL, Free Press.
Currie, J. and Madrian, B. (1999), 'Health, health insurance and the labor market', in Ashenfelter, O. and Card, D. (eds), Handbook of Labor Economics, Amsterdam, Elsevier, pp. 3309-416.
Ewing, B.T. (1998), 'Athletes and work', Economics Letters, 59, pp. I 13-7.
--(2007), 'The labor market effects of high school athletic participation: evidence from wage and fringe benefit differentials', Journal of Sports Economics, 8, pp. 255-65.
Farrell, L. and Shields, M.A. (2002), 'Investigating the economic and demographic determinants of sporting participation in England', Journal of the Royal Statistical Society: Series A, 165, pp. 335-48.
Glaeser, E.L., Laibson, D. and Sacerdote, B. (2002), 'An economic approach to social capital', Economic Journal, I 12, F437-58.
Grossman, M. (1972), 'On the concept of health capital and the demand for health', Journal of Political Economy, 80, pp. 223-55.
Hamermesh, D.S. and Biddle, J.E. (1994), 'Beauty and the labor market', American Economic Review, 84, pp. I 174-94.
Heckman, J.J. and Rubinstein, Y. (2001), 'The importance of noncognitive skills: lessons from the GED Testing Program', American Economic Review, 91, pp. 145-9.
Jacob, B., McCall, B. and Stange, K.M. (2013), 'College as country club: do colleges cater to students' preferences for consumption?', NBER Working Paper No. 18745.
Kosteas, V.D. (2011), 'High school clubs participation and future supervisory status', British Journal of Industrial Relations, 49:S I, pp. s 181-206.
Laibson, D. (1997), 'Golden eggs and hyperbolic discounting', Quarterly Journal of Economics, I 12, pp. 443-77.
Lechner, M. (2009), 'Long-run labour market and health effects of individual sports activities', Journal of Health Economics, 28, pp. 839-54.
Lechner, M. and Downward, P. (2013), 'Heterogeneous sports participation and labour market outcomes in England', Discussion Paper 2013-25, University of St. Gallen.
Lipscomb, S. (2007), 'Secondary school extracurricular involvement and academic achievement: a fixed effects approach', Economics of Education Review, 26, pp. 463-72.
Long, J.E. and Caudill, S. B. (1991), 'The impact of participation in intercollegiate athletics on income and graduation', Review of Economics and Statistics, 73, pp. 525-31.
Lunn, P.D. (2007), Fair Play? Sport and Social Disadvantage in Ireland. ESRI Books and Monographs, 190.
--(2010), 'The sports and exercise life-course: a survival analysis of recall data from Ireland', Social Science and Medicine, 70, pp. 711-9.
Lunn, P., Kelly, E. and Fitzpatrick, N. (2013). Keeping Them in the Came: Taking Up and Dropping Out of Sport and Exercise in Ireland. ESRI Research Series 33. Dublin: ESRI.
Mobius, M.M. and Rosenblat, T.S. (2006), 'Why beauty matters', American Economic Review, 96, pp. 222-35.
Morris, S. (2006), 'Body mass index and occupational attainment', Journal of Health Economics, 25, pp. 347-64.
O'Donoghue, T. and Rabin, M. (1999), 'Doing it now or later', American Economic Review, 89, pp. 103-24.
Pfeifer, C. and CorneliBen, T. (2010), 'The impact of participation in sports on educational attainment--new evidence from Germany', Economics of Education Review, 29, pp. 94-103.
Physical Activity Guidelines Advisory Committee (2008), Physical Activity Guidelines Advisory Committee Report, 2008, Washington, DC, U.S. Department of Health and Human Services.
Rees, D.I. and Sabia, J.J. (2010), 'Sports participation and academic performance: evidence from the National Longitudinal Study of Adolescent Health', Economics of Education Review, 29, pp. 751-9.
Rehberg, R.A. and Schafer, W.E. (1968), 'Participation in interscholastic athletics and college expectations', American Journal of Sociology, 73, pp. 732-40.
Rooth, D.O. (2009), 'Obesity, attractiveness and differential treatment in hiring--a field experiment', Journal of Human Resources, 44, pp. 710-35.
--(2011), 'Work out or out of work: the labor market return to physical fitness and leisure sport activities', Labour Economics, 18, pp. 399-409.
Rosenbaum, P.R. (2002), Observational Studies, 2nd Edition, New York, Springer.
Rosenbaum, P. and Ruben, D. (1983), 'The central role of the propensity score in observational studies for causal effects', Biometrika, 70, pp. 41-55.
Schultz, T.W. (1960), 'Capital formation by education', Journal of Political Economy, 68, pp. 571-82.
Shulruf, B. (2010), 'Do extra-curricular activities in schools improve educational outcomes? A critical review and meta-analysis of the literature', International Review of Education, 56, pp. 591-612.
Singh, A., Uijtdewilligen, L, Twisk, J.W.R., van Mechelen, W. and Chinapaw, M. (2012), 'Physical activity and performance at school: a systematic review of the literature including a methodological quality assessment', Archives of Paediatrics and Adolescent Medicine, 166, pp. 49-55.
Spence, J. (1973), 'Job market signaling', Quarterly Journal of Economics, 87, pp. 355-74.
Taras, H. (2005), 'Physical activity and student performance at school', Journal of School Health, 75, pp. 214-8.
Taskforce on Community Preventive Services (2001), Increasing Physical Activity, US Department of Health and Human Services, Centers for Disease Control and Prevention, Morbidity and Mortality Weekly Report, 50.
Trudeau, F. and Shephard, R.J. (2008), 'Physical education, school physical activity, school sports and academic performance', International Journal of Behavioral Nutrition and Physical Activity, 5(10).
(1) TY is designed to allow students to experience alternative educational inputs without formal examination. For further information on the TY programme, see http://www.education. ie/en/Schools-Colleges/Information/Curriculum-and-Syllabus/ Transition-Year-/ty_transition_year_school_guidelines.pdf).
(2) See Lunn, Kelly and Fitzpatrick (2013) for the use of the 2007 SLS data to examine the impact of sports participation on secondary education examination performance, specifically the official exam that is sat at the end of upper secondary.
(3) In Ireland, schools that are designated as being disadvantaged are provided with supplementary resources and supports in order to address the educational needs of children and young people from disadvantaged communities.
(4) For example, hurling and Gaelic football, which are two traditional Irish sports. For further information, see http://www. gaa.ie/about-the-gaa/our-games/.
(5) MNL models additionally assume that the Independence of Irrelevant Alternatives (IIA) property is satisfied. This property implies that, conditional on observed characteristics, the utility levels of any two alternatives are independent.
(6) The radius matching algorithm produced slightly larger sports participation treatment effects on the 'student/training' and 'unemployment/inactivity' outcomes. The kernel is a nonparametric estimator that uses weighted averages of all individuals in the control group as opposed to extra control cases; nevertheless, the estimator is still based on additional information in comparison with the single nearest neighbour matching estimator. The sensitivity test results for the Senior Cycle sports participation relative to Non-Participation model are presented in Appendix table Al. All other sensitivity test results are available from the authors on request.
(7) This test was conducted on all treatments and outcomes presented in tables 4 to 6--too many to present each result individually. All diagnostic tests referred to above are available from the authors on request.
(8) Based on the occupation group of the higher placed parent where both parents were in employment.
(9) The employment information during term-time related to any year of secondary school, while the employment outside of term-time data related to the respondent's last year in school.
(10) The result is very similar if a variable is instead introduced to control for when the student left school. This variable is too highly correlated with the final examination sat to employ both simultaneously.
(11) Results available from the authors on request.
Peter D. Lunn and Elish Kelly *
* Economic and Social Research Institute (ESRI) and Department of Economics, Trinity College Dublin. Corresponding author e-mail: Elish.Kelly@esri.ie. The authors would like to thank Emer Smyth and two anonymous referees for helpful comments on earlier drafts, Selina McCoy for her assistance with the School Leavers' Survey, Seamus McGuinness for econometric advice, and the Irish Sports Council for supporting a research project on transitions in sports participation, which ultimately led us to investigate the present issue.
Table 1. Multinomial logistic regression of being unemployed/inactive relative to entering employment for 2004/5 school leavers two years post leaving school (May 2007) Basic specification Coefficient Std error Sports Participation (Ref: No) Junior Cycle -0.206 (0.198) Senior Cycle -0.503 * (0.260) Male -0.482 ** (0.192) Age -0.009 ** (0.005) Illness/Disability 1.386 *** (0.340) Socio-Economic Background (Ref: Skilled/Semi/Unskilled Manual) Farmer -1.518 ** (0.626) Professional 0.031 (0.370) Employer/Manager -0.185 (0.397) Intermediate/Other non-manual 0.219 (0.224) Unemployed 0.710 ** (0.319) Other 0.874 *** (0.320) Mother's Education (Ref Primary or Less) Junior Certificate 0.023 (0.233) Leaving Certificate -0.732 *** (0.281) Post-secondary non-tertiary -0.656 (0.658) Non-degree (Cert/Diploma) -1.057 ** (0.522) Degree or higher -0.222 (0.416) Education unknown -0.204 (0.287) Transition year participation -0.375 * (0.226) Grinds in final year -0.306 (0.233) Part-time job during term -0.566 ** (0.234) Part-time job outside of term -0.610 *** (0.232) Full-time job outside of term -0.990 *** (0.300) No information on job outside of term -0.712 (0.678) School Type (Ref Vocational) Girls' secondary 0.145 (0.334) Boys' secondary 0.113 (0.310) Co-educational 0.182 (0.265) Community/Comprehensive 0.365 * (0.21 1) Truancy (Ref Never) Lesson here and there Day here and there Several days at a time Weeks at a time No truancy information available Individual motivation Non individual motivation information Teacher support No teacher support information Classroom context No classroom context information Last exam sat Established Junior Certificate Junior Cert school programme Established Leaving Certificate Leaving Cert applied Leaving Cert vocational Constant 1.631 (1.078) Observations 1,956 Pseudo R2 0.1719 School mediating factors Coefficient Std error Sports Participation (Ref: No) Junior Cycle -0.245 (0.204) Senior Cycle -0.385 (0.264) Male -0.493 ** (0.198) Age -0.005 (0.005) Illness/Disability 1.403 *** (0.348) Socio-Economic Background (Ref: Skilled/Semi/Unskilled Manual) Farmer -1.460 ** (0.628) Professional 0.087 (0.375) Employer/Manager -0.223 (0.403) Intermediate/Other non-manual 0.235 (0.229) Unemployed 0.658 ** (0.329) Other 0.895 *** (0.329) Mother's Education (Ref Primary or Less) Junior Certificate 0.053 (0.240) Leaving Certificate -0.679 ** (0.287) Post-secondary non-tertiary -0.608 (0.659) Non-degree (Cert/Diploma) -1.084 ** (0.527) Degree or higher -0.241 (0.422) Education unknown -0.135 (0.298) Transition year participation -0.395 * (0.229) Grinds in final year -0.244 (0.236) Part-time job during term -0.585 ** (0.238) Part-time job outside of term -0.540 ** (0.237) Full-time job outside of term -0.979 *** (0.305) No information on job outside of term -0.737 (0.693) School Type (Ref Vocational) Girls' secondary 0.154 (0.340) Boys' secondary 0.115 (0.316) Co-educational 0.226 (0.272) Community/Comprehensive 0.418 * (0.217) Truancy (Ref Never) Lesson here and there -0.147 (0.234) Day here and there 0.481 ** (0.217) Several days at a time 0.602 ** (0.290) Weeks at a time 0.751 (0.530) No truancy information available -1.259 (1.209) Individual motivation -0.098 (0.210) Non individual motivation information 1.422 (0.865) Teacher support -0.332 (0.298) No teacher support information -0.428 (0.534) Classroom context -0.033 (0.225) No classroom context information -0.339 (0.661) Last exam sat Established Junior Certificate Junior Cert school programme Established Leaving Certificate Leaving Cert applied Leaving Cert vocational Constant 0.840 (1.123) Observations 1,956 Pseudo R2 0.1964 Last examination sat Coefficient Std error Sports Participation (Ref: No) Junior Cycle -0.300 (0.207) Senior Cycle -0.233 (0.270 Male -0.594 *** (0.204) Age 0.005 (0.005) Illness/Disability 1.308 *** (0.352) Socio-Economic Background (Ref: Skilled/Semi/Unskilled Manual) Farmer -1.554 ** (0.643 Professional 0.083 (0.381) Employer/Manager -0.203 (0.408) Intermediate/Other non-manual 0.171 (0.233) Unemployed 0.519 (0.339) Other 0.830 ** (0.332 Mother's Education (Ref Primary or Less) Junior Certificate 0.151 (0.246) Leaving Certificate -0.546 * (0.293) Post-secondary non-tertiary -0.461 (0.662) Non-degree (Cert/Diploma) -1.055 ** (0.534) Degree or higher -0.093 (0.431) Education unknown -0.070 (0.304) Transition year participation -0.302 (0.234) Grinds in final year 0.022 (0.252) Part-time job during term -0.529 ** (0.243) Part-time job outside of term -0.455 * (0.242) Full-time job outside of term -0.867 *** (0.310) No information on job outside of term -0.791 (0.706) School Type (Ref Vocational) Girls' secondary 0.138 (0.350) Boys' secondary 0.135 (0.319) Co-educational 0.172 (0.277) Community/Comprehensive 0.405 * (0.223) Truancy (Ref Never) Lesson here and there -0.128 (0.238 Day here and there 0.444 ** (0.222) Several days at a time 0.502 * (0.293) Weeks at a time 0.491 (0.536) No truancy information available -1.283 (1.220) Individual motivation 0.003 (0.214) Non individual motivation information 1.541 * (0.882) Teacher support -0.248 (0.300) No teacher support information -0.538 (0.542) Classroom context 0.011 (0.228 No classroom context information -0.257 (0.658) Last exam sat Established Junior Certificate -0.804 *** (0.270) Junior Cert school programme -0.784 ** (0.336) Established Leaving Certificate -1.538 *** (0.350) Leaving Cert applied -0.870 * (0.477) Leaving Cert vocational -1.962 *** (0.560) Constant -0.522 (1.152) Observations 1,956 Pseudo R2 0.2245 Table 2. Multinomial logistic regression of being a student relative to entering employment for 2004/5 school leavers two years post leaving school (May 2007) Basic specification Coefficient Std error Sports Participation (Ref: No) Junior Cycle -0.194 (0.145) Senior Cycle 0.483 *** (0.130) Male -0.734 *** (0.130) Age -0.011 *** (0.003) Illness/Disability -0.604 (0.445) Socio-Economic Background (Ref: Skilled/Semi/Unskilled Manual) Farmer 0.418* (0.232) Professional 0.816 *** (0.209) Employer/Manager 0.391 * (0.217) Intermediate/Other non-manual 0.257 (0.169) Unemployed 0.639 ** (0.293) Other 0.715 *** (0.270) Mother's Education (Ref Primary or Less) Junior Certificate 0.068 (0.204) Leaving Certificate 0.177 (0.205) Post-secondary non-tertiary 0.472 (0.332) Non-degree (Cert/Diploma) 0.535 ** (0.253) Degree or higher 0.445 * (0.254) Education unknown 0.318 (0.237) Transition year participation 0.464 *** (0.122) Grinds in final year 0.707 *** (0.118) Part-time job during term -0.396 *** (0.132) Part-time job outside of term 0.104 (0.152) Full-time job outside of term -0.187 (0.174) No information on job outside of term 0.072 (0.469) School Type (Ref: Vocational) Girls' secondary 1.134 *** (0.202) Boys' secondary 0.790 *** (0.186) Co-educational Q 944 *** (0.168) Community/Comprehensive 0.495 *** (0.152) Truancy (Ref Never) Lesson here and there Day here and there Several days at a time Weeks at a time No truancy information available Individual motivation Non individual motivation information Teacher support No teacher support information Classroom context No classroom context information Last exam sat Established Junior Certificate Junior Cert school programme Established Leaving Certificate Leaving Cert applied Leaving Cert vocational Constant 1.244 (0.794) Observations 1,956 Pseudo R2 0.1719 School mediaiting factors Coefficient Std error Sports Participation (Ref: No) Junior Cycle -0.210 (0.148) Senior Cycle 0.388 *** (0.132) Male -0.690 *** (0.132) Age -0.014 *** (0.003) Illness/Disability -0.599 (0.451) Socio-Economic Background (Ref: Skilled/Semi/Unskilled Manual) Farmer 0.384 (0.237) Professional 0.811 *** (0.213) Employer/Manager 0.416 * (0.221) Intermediate/Other non-manual 0.261 (0.172) Unemployed 0.771 ** (0.301) Other 0.723 *** (0.276) Mother's Education (Ref Primary or Less) Junior Certificate 0.103 (0.207) Leaving Certificate 0.219 (0.208) Post-secondary non-tertiary 0.550 (0.340) Non-degree (Cert/Diploma) 0.601 ** (0.258) Degree or higher 0.481 * (0.258) Education unknown 0.370 (0.242) Transition year participation 0.488 *** (0.125) Grinds in final year 0.629 *** (0.121) Part-time job during term -0.343 ** (0.134) Part-time job outside of term 0.090 (0.154) Full-time job outside of term -0.129 (0.178) No information on job outside of term -0.066 (0.476) School Type (Ref: Vocational) Girls' secondary 1.020 *** (0.206) Boys' secondary 0.782 *** (0.190) Co-educational 0.946 *** (0.172) Community/Comprehensive 0.412 *** (0.155) Truancy (Ref Never) Lesson here and there -0.293 ** (0.130) Day here and there -0.711 *** (0.157) Several days at a time -0.822 *** (0.269) Weeks at a time -1.244 ** (0.610) No truancy information available -0.226 (0.769) Individual motivation 0.605 *** (0.171) Non individual motivation information 0.951 (0.704) Teacher support 0.125 (0.266) No teacher support information -0.007 (0.368) Classroom context 0.008 (0.165) No classroom context information 0.066 (0.407) Last exam sat Established Junior Certificate Junior Cert school programme Established Leaving Certificate Leaving Cert applied Leaving Cert vocational Constant 1.599 * (0.856) Observations 1,956 Pseudo R2 0.1964 Last examinition sat Coefficient Std error Sports Participation (Ref: No) Junior Cycle -0.055 (0.154) Senior Cycle 0.301 ** (0.135) Male -0.552 *** (0.136) Age -0.027 *** (0.004) Illness/Disability -0.365 (0.459) Socio-Economic Background (Ref: Skilled/Semi/Unskilled Manual) Farmer 0.353 (0.245) Professional 0.728 *** (0.217) Employer/Manager 0.342 (0.226) Intermediate/Other non-manual 0.276 (0.176) Unemployed 0.758 ** (0.310) Other 0.745 *** (0.284) Mother's Education (Ref Primary or Less) Junior Certificate 0.055 (0.213) Leaving Certificate 0.176 (0.214) Post-secondary non-tertiary 0.410 (0.348) Non-degree (Cert/Diploma) 0.598 ** (0.266) Degree or higher 0.333 (0.265) Education unknown 0.291 (0.250) Transition year participation 0.527 *** (0.129) Grinds in final year 0.421 *** (0.125) Part-time job during term -0.319 ** (0.138) Part-time job outside of term -0.018 (0.159) Full-time job outside of term -0.264 (0.184) No information on job outside of term -0.126 (0.480) School Type (Ref: Vocational) Girls' secondary 0.853 *** (0.210) Boys' secondary 0.621 *** (0.196) Co-educational 0.862 *** (0.178) Community/Comprehensive 0.353 ** (0.160) Truancy (Ref Never) Lesson here and there -0.271 ** (0.133) Day here and there -0.605 *** (0.161) Several days at a time -0.615 ** (0.277) Weeks at a time -1.027 (0.635) No truancy information available -0.110 (0.768) Individual motivation 0.477 *** (0.177) Non individual motivation information 0.751 (0.710) Teacher support -0.003 (0.275) No teacher support information 0.067 (0.377 Classroom context -0.033 (0.170) No classroom context information 0.041 (0.425) Last exam sat Established Junior Certificate -0.666 ** (0.284) Junior Cert school programme -0.951 *** (0.357) Established Leaving Certificate 0.721 ** (0.298) Leaving Cert applied -0.380 (0.458) Leaving Cert vocational 0.505 (0.341) Constant 4.570 *** (0.9 Observations 1,956 Pseudo R2 0.2245 Table 3. Impact of sports participation on 2004-5 school leavers' post-school pathways two years after leaving school (May 2007): marginal effects School Basic mediating specification factors Employment (includes Apprenticeships) Sports Participation (Ref: No) Junior Cycle 0.042 0.045 * (-0.012-0.095) (-0.008-0.099) Senior Cycle -0.052 * -0.041 (-0.104-0.000) (-0.093-0.012) Unemployment/lnactivity Sports Participation (Ref: No) Junior Cycle -0.009 -0.014 (-0.040-0.021) (-0.045-0.017) Senior Cycle -0.057 *** -0.042 ** (-0.098--0.015) (-0.082--0.001) Student/Training Sports Participation (Ref No) Junior Cycle -0.030 -0.032 (-0.083-0.022) (-0.084-0.020) Senior Cycle 0.109 *** 0.083 *** (0.062-0.156) (0.036-0.129) Observations 1,956 1,956 Last examination sat Employment (includes Apprenticeships) Sports Participation (Ref: No) Junior Cycle 0.024 (-0.030-0.078) Senior Cycle -0.033 (-0.085-0.019) Unemployment/lnactivity Sports Participation (Ref: No) Junior Cycle -0.022 (-0.052-0.009) Senior Cycle -0.026 (-0.066-0.014) Student/Training Sports Participation (Ref No) Junior Cycle -0.002 (-0.054-0.050) Senior Cycle 0.059 ** (0.013-0.104) Observations 1,956 Note: 95 per cent confidence internals in parentheses; *** p<0.0l, ** p<0.05, * p<0.l. Table 4. Impact of Senior Cycle sports participation relative to non-participation on 2004-5 school leavers' post-school pathways two years after leaving school (May 2007): MNL and PSM results (marginal effects) Basic specification MNL PSM (a) Employment (includes Apprenticeships) Sports Participation (Ref: No) Senior Cycle -0.056 ** -0.093 ** (0.026) (0.041) Unemployment/lnactivity Sports Participation (Ref No) Senior Cycle -0.052 *** -0.018 (0.020) (0.020) Student/Training Sports Participation (Ref: No) Senior Cycle 0.109 *** 0.1 1 1 *** (0.025) (0.040) Observations 1,530 1,528 School mediating factors MNL PSM Employment (includes Apprenticeships) Sports Participation (Ref: No) Senior Cycle -0.041 -0.035 (0.027) (0.039) Unemployment/lnactivity Sports Participation (Ref No) Senior Cycle -0.040 ** -0.022 (0.020) (0.019) Student/Training Sports Participation (Ref: No) Senior Cycle 0.081 *** 0.056 (0.025) (0.039) Observations 1,530 1,530 Note: Standard errors in parentheses; *** p<0.0l, ** p<0.05, * p<0.1. (a) Two observations lost through the enforcement of common support. Table 5. Impact of Junior Cycle sports participation relative to non-participation on 2004-5 school leavers' post-school pathways two years after leaving school (May 2007): MNL and PSM results (marginal effects) School mediating Basic specification factors MNL PSM MNL PS MW Employment (includes Apprenticeships) Sports Participation (Ref No) Junior Cycle 0.037 0.085 0.041 0.038 (0.028) (0.041) (0.028) (0.043) Unemployment/lnactivity Sports Participation (Ref: No) Junior Cycle -0.015 -0.031 -0.018 -0.031 (0.019) (0.030) (0.019) (0.031) Student/Training Sports Participation (Ref No) Junior Cycle -0.022 -0.054 -0.023 -0.007 (0.027) (0.040) (0.026) (0.041) Observations 1,405 1,405 1,405 1,404 Note: Standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1. (a) One observation lost through the enforcement of common support. Table 6. Impact of sports drop out after Junior Cycle relative to participation in Senior Cycle on 2004-5 school leavers' post-school pathways two years after leaving school (May 2007): MNL and PSM results (marginal effects) Basic specification MNL PS MW Employment (includes Apprenticeships) Sports Participation (Ref: Senior Cycle Participation) Junior Cycle drop out 0.087 *** 0.065 (0.032) (0.054) Unemployment/lnactivlty Sports Participation (Ref: Senior Cycle Participation) Junior Cycle drop out 0.050 ** 0.027 (0.020) (0.027) Student/Tralning Sports Participation (Ref Senior Cycle Participation) Junior Cycle drop out -0.138 *** -0.092 * (0.029) (0.054) Observations 977 922 School mediating factors MNL PSM(b) Employment (includes Apprenticeships) Sports Participation (Ref: Senior Cycle Participation) Junior Cycle drop out 0.080 ** 0.050 (0.032) (0.053) Unemployment/lnactivlty Sports Participation (Ref: Senior Cycle Participation) Junior Cycle drop out 0.030 0.038 * (0.020) (0.023) Student/Tralning Sports Participation (Ref Senior Cycle Participation) Junior Cycle drop out -0.110 *** -0.088 * (0.029) (0.053) Observations 977 890 Note: Standard errors in parentheses; *** p<0.0l, ** p<0.05, * p<0.1. (a) Fifty-five observations lost through the enforcement of common support, (b) Eighty-seven observations lost through the enforcement of common support.
|Printer friendly Cite/link Email Feedback|
|Author:||Lunn, Peter D.; Kelly, Elish|
|Publication:||National Institute Economic Review|
|Date:||May 1, 2015|
|Previous Article:||Born in the USA: national origin effects on time allocation in US and Spanish professional basketball.|
|Next Article:||Forecast overview.|