In this study we found that NSC was significantly associated with LT sports participation, whereas the objectively observed availability and density of parks or sports facilities was not. The association of NSC with LT sports participation remained significant when adjusted for availability/density of sports facilities or parks. In addition we found that when NSC was high, availability of parks was associated with LT sports participation, which was in line with our hypotheses.
It is often argued that NSC has positive influences on health behaviour. However, it is theoretically also possible that negative norms that exist in a neighborhood may be “disseminated” through the community with detrimental influences on behaviour. Interesting is that higher levels of social capital may also unintentionally lead to negative influences on physical activity. In a study by Altschuler et al.  it was found that in one neighbourhood successfully lobbied in preventing installation of street lighting – so the natural scenery of the neighbourhood was saved. This lead to a situation which is less attractive for pedestrians and hence, walking. However, in this study, NSC was consistently positively associated with LT sports participation. This finding fits with our hypotheses, various theoretical models [15, 16] that assume this association and with the results of previous research that have used a measure of perceived NSC (i.e. not aggregated to the neighbourhood level) [10, 11]. The present study extends on the existing evidence, as we used an aggregated measure of NSC.
With regard to the physical environment, we did not find an association of availability and density of sports facilities or parks with LT sports participation in both areas that we studied. These results add to the current, equivocal, literature on the associations of the physical environment with adolescent PA . It may be that these inconclusive findings are due to various conceptualizations of the neighbourhood. Therefore various authors have called upon reporting results in various environmental scales (i.e. buffer sizes) [47, 48]. We have conceptualized the physical environment in two scales; one that is evidence based (1600 meter buffer size) and one that matches the NSC measure (neighbourhood level), but found no differences in associations between these two conceptualizations. It may be that adolescents are active in other places than the neighbourhood in which they live. It is very well possible that the environment around their school, or environments where their friends live, is more important for some adolescents. For instance, Jones et al. found that MVPA takes place outside commonly used buffers . A mismatch in the environment where PA facilities are assessed and the environment where PA is actually performed may have also contributed to lack of associations found. Future research should consider using advanced technology, such as GPS with integrated accelerometry in determining the environment where adolescents are active and to determine associations between the physical environments and PA. Some preliminary research indicates that using these advances in GPS technology is feasible and promising [49–53]. However, it may also be that the physical environment in itself is not enough to promote PA, but may act more as a barrier or facilitator of PA behaviour . In this study we may have found some evidence that points in that direction.
We showed that density of parks at the neighbourhood level interacted with NSC in such a way that when both NSC and density of parks were high, adolescents were most likely to participate in LT sports. Thus, indeed, density of parks can be important for sports participation, but only in combination with high levels of NSC. Our results that the combination of physical and social environmental factors results in the highest likelihood for sports participation, are in line with findings of a systematic review of qualitative studies by McCormack et al. They found that the presence of a park gives important opportunities to engage in PA, but other features of that park such as the social environment may promote park usage . Seaman et al. also found in a qualitative study that for promoting park use not only the presence of a park, but also factors like social cohesion are of importance . Finally Broyles et al., who found that in parks with more “Park level social capital” the overall intensity of activities performed in that park were higher . However, it should be noted that the social capital on a park level is not equal to NSC as we have conceptualized it.
For promoting a healthy lifestyle (e.g. sports participation) on a population level, it is important to identify important and changeable determinants of that lifestyle . We have identified NSC as a potentially important factor associated with LT sports participation, especially when density of parks is high. If our results can be replicated by others, NSC may be an important factor in relation to sports promotion. In addition, to be suitable to incorporate in interventions and policy to promote LT sports participation, it is also important to know how NSC can be promoted. Wood et al. found that various factors, such as availability of facilities like nearby shops, are positively associated with higher levels of NSC . This may give valuable target points for developing interventions and policies that aim to promote NSC. In addition, various studies suggest that PA can successfully be promoted by building NSC by cooperating closely with citizens in developing policies and strategies [57, 58].
Although the aim of this study was not to fully explain neighbourhood variance in LT sports participation, even in the most extended models there was unexplained neighbourhood variance. This indicates that other factors may be of interest that may explain neighbourhood variation in LT sports participation; for instance environmental factors such as presence of sidewalks and bicycle lanes (increasing opportunities to access parks or sports facilities), quality of these facilities, crime or aesthetics. Studies that are aimed at explaining neighbourhood differences in LT sports participation should thus, additionally, incorporate other factors than we did.
There is some criticism on the concept of social capital in that it is a fuzzy concept. According to Kawachi et al. this may have to do with two conceptualizations of social capital. On the one side there is conceptualization that is more individual level social capital (e.g. networks) and on the other side social capital may be conceptualized as resources available to the community . We used the latter definition and therefore aggregated data to the neighbourhood level. To aggregate individual data to a neighbourhood level, it is most common to calculate the average of the items measured at the individual level per neighbourhood . However, this aggregation procedure has several drawbacks. First, variables measuring social capital are based on individual perception, and it is likely that this perception is influenced by the characteristics of the respondent (e.g. time living in a neighbourhood, cultural values). Second, since the number of respondents differs per neighbourhood, the reliability of the aggregated social capital variable also differs between the neighbourhoods. Finally, the items that measure social capital are not independent of each other but nested within respondents; that is, answers on one item are likely to be associated with answers on the other item. In summary, an approach that accounts for individual differences in response to certain items, for differences in numbers of respondents on which the estimation is based, and for dependency among the items that measure social capital is needed. The ecometrics methodology is an approach that does this. And we have applied this in our study, which can be regarded as a major strength in this study.
Another strength of this study is that the behaviour (LT sports participation) studied is relevant for its context (the neighbourhood environment), because it is less likely that the home neighbourhood affects inside school sports participation than LT sports participation. In contrast to other studies, the neighbourhood perception of adolescents instead of adults was used to estimate NSC, which is another strength of this study. However, as this is the first study to use adolescent perceptions of NSC, our results need to be replicated to draw more definitive conclusions. Future studies may, given the promising results of our study also wish to further explore the construction of a more child-specific measure of NSC, as the measure that we used was derived from an instrument that was originally developed for adults. Our study has also some limitations. Firstly, the current study relies on self-reported measures of PA. Objective measures, such as accelerometers, may give more valid data on minutes spend in PA. However, accelerometers do not measure important information on the type of activity (e.g. cycling, walking, sports), which is relevant to study the influence of environmental factors on behaviour, as different factors in the environment, such as availability of facilities are likely to be associated with different types of PA behaviour specific (e.g. sports facilities may be of importance for sports, but not for walking to school) . Secondly, there may be self-selection bias in our sample, as suggested by Boone-Heinonen et al. ; physically active families may select supportive neighbourhoods for their PA to reside. This may have altered the associations found in this study. However, a study among Dutch adults did not find evidence for self-selection bias in the Netherlands  and parents are most likely to choose the residential neighbourhood and not the adolescents. Therefore, we expect that self-selection bias was not a major cause of bias in this study. Another limitation is the cross-sectional design and no conclusions on causality can be drawn. To draw conclusions on causal relationships future research, using longitudinal and experimental designs need to be carried out. Moreover, in replicating our findings these studies should be specifically designed to study neighbourhood influences on sports participation; by sampling multiple neighbourhoods with maximal neighbourhood variation and small clustered samples of individuals in each neighbourhood. In addition studies may also profit from inducing changes in the environmental factors to study the effects of changing physical and social environmental factors in their relation to sports participation. Finally, the data were retrieved from the YouRAction trial, which aimed to promote MVPA among adolescents. In the evaluation study, an effect of the intervention could not be established. Furthermore, we have checked whether the interventions were associated with the outcome of interest in this study and seen that there was no effect of the interventions and have adjusted for intervention group. Therefore, it is not likely that the interventions have affected the results of this study.
To conclude, in this study we found evidence for NSC as a potentially important and robust correlate of LT sports participation among adolescents. We did not find a direct association between availability of sports facilities or parks with LT sports participation among adolescents. The interaction between density of parks in the neighbourhood and NSC showed that when NSC was high, presence of parks was stronger associated with LT sports participations. In sum, the combination of high NSC and high density of parks is associated with the highest likelihood of LT sports participation among adolescents.