In recent years, research has revealed key facts about participation in sport across different countries and contexts (Van Tuyckom and Scheerder, 2010). Although there are signs that the effect of social class on practising sports has diminished in the last decades due to processes of individualisation (Yang et al, 1996; Scheerder et al., 2005a), many studies continue finding a strong correlation between both variables: the gap between social classes when practising sports still exists (Scheerder et al., 2002; Scheerder et al., 2005c; Scheerder and Vos, 2011; Kahma, 2012; Kraaykamp et al., 2013; Wheeler et al., 2017; Andersen and Bakken, 2018).

The differences in sports habits that stem from individuals’ socioeconomic status have consequences for equality. For example, we know that practising sports organized in clubs fosters friendships and consolidates social networks, i.e. increases social capital (Seippel, 2006). It has also been found that participation in sport has multiple benefits for physical and mental health (Wankel and Berger, 1990; Sallis and Owen, 1998; Bailey, 2005, 2006; Eime et al., 2013), as well as quality of life, especially among young people (Moscoso Sánchez and Moyano Estrada, 2009). Sport has even been linked to better educational performance (Bailey et al., 2009). If some social classes participate more in sports than others, this inequality can have an impact on these other fundamental areas in people’s lives. Several scientific studies have proven that one of the variables that most decisively increases people’s sports practice is that parents are physically active (Scheerder et al., 2005c). This finding corroborates others that relate people’s sports participation with parental attitudes and behaviour (Sage, 1980; Yang et al., 1996).

Sociology has repeatedly demonstrated that social class position influences the trajectories of individuals, their expectations, opportunities and, ultimately, their well-being, throughout their life cycle (Wright, 2000). In the last decade, there have been a number of attempts from social stratification researchers to identify which factors soften or intensify these class differences in areas as diverse as health (Wilkinson and Pickett, 2009; Kunts and Roskam, 2010), the labour market (Bernardi and Ballarino, 2016), the education system (Breen and Jonsson, 2005) and cultural consumption (Chan, 2010). If practising sports provides considerable benefits for people throughout their lives, it is important to analyse the possible factors that mediate class differences in this type of activity. In this article we try to shed more light on the way in which social class determines an individual’s sports practice (Andersen and Bakken, 2018; Stefansen, 2018).

The aim of this article is to analyse the extent to which parents’ sports habits have a different effect according to an individual’s social class position. Specifically, this study aims to identify to what degree having physically active parents reinforces or, on the contrary, reduces class differences among young people when practising sports. The main contribution of this work is to reveal parental influence as one of the possible levelling enhancers of inequality between social classes. In addition, this study employs the EGP social class scheme (Ganzeboom and Treiman, 2011), a classification rarely used in sports participation studies (Andersen and Bakken, 2018).

As far as we know, the present analysis is original, since it seeks to explore to what extent the association between social class and young people’s sports activity is moderated by parents’ sports activity, a relationship about which little is known. If we were to find that parents’ sports participation benefits the upper class more than the working class, it would reveal an enhancer whereby the social class gap becomes even greater. On the contrary, if parents’ sports practice benefits the working class more, this would reveal an important mediator of social equalization involving something as fundamental as sports.

The data used in this study comes from Spain’s Centre for Sociological Research (CIS), specifically, from the four Sports Habits Surveys (SHS) conducted in 1995, 2000, 2005 and 2010 (see Table A1 in Appendix), commissioned by the Higher Sports Council ¹ . These surveys were aimed at people who were 15 years of age or older. The sample is representative for all of Spain. The union of the four data files provides a total of 26,526 cases. In this analysis we focus on the sports practice of men and women between 15 and 29 years of age, which gives a final sample of 7,323 observations. The reason to limit the sample to these ages is that one of the two decisive independent variables is parents’ sports participation, so we have selected ages in which the majority of Spanish young people still live with their parents or, if they have become independent, they may continue to be under the recent influence of their parents. According to Eurostat (2007), the average age of emancipation in Spain is 29.0, an age close to that of other southern European countries such as Italy (30.1), Greece (29.4) and Portugal (28.9). The characteristics of the sample by social classes are shown in Table 1.

The key dependent variable of the study is to practise sport. The questionnaires used in the CIS’s four surveys collected specific information from those who reported that they exercised. They were asked about the frequency with which they participated in sports. In particular, the following four options were distinguished: i) three times or more per week, ii) once or twice a week, iii) less frequently, and iv) don’t know, or no answer. In this research, we consider someone as practising sports if they do so at least once a week, that is, if they fall into one of the first two categories. In addition, to check if the effects vary for more intensive sport practice, we have repeated the analysis with a more restrictive criterion: taking part in sports at least three times a week.

The key independent variables are social class and the sports habits of the parents. To measure social class, we have used the classification proposed by Erikson, Goldthorpe and Portocarrero (1979), which originally contained eleven categories. We have followed the operationalization proposed by Ganzeboom and Treiman (2011) and adapted it to the National Classification of Occupations. The original EGP class scheme has 11 categories: I. Large proprietors, higher professionals and managers; II. Lower professionals and managers; IIIa. Higher routine non-manual workers: IIIb. Lower routine non-manual workers; IVa. Small proprietors with employees; IVb. Small proprietors without employees; IVc. Self-employed farmers; V. Lower grade technicians and manual supervisors; VI. Skilled manual workers; VIIa. Unskilled and semiskilled manual workers; VIIb. Agricultural workers.

Given that there are only a few cases in some categories, we have found it convenient to group them into six social classes, as proposed by Andersen and Bakken (2018): 1) higher professionals (EGP class I); 2) lower professionals (II); 3) routine non-manual employees (IIIab); 4) self-employed (IVabc); 5) manual supervisors and skilled workers (V and VI); and 6) unskilled workers and farm labourers (VIIa and VIIb). The two first categories comprise Erikson and Goldthorpe’s service class.

These classes are derived from the occupation that each interviewee had at the time of the interview, or in the case of the unemployed or economically inactive, from the last occupation they had. If an individual has never worked, the social class assigned is that of the main person in the family home. In addition to these six categories, we have included a seventh, for unemployed people with previous work experience. It was not possible to construct this variable for 5% of the sample since they did not answer the corresponding questions. For this reason, another category entitled “no response” has been added to the analysis.

The question about occupation in the surveys is formulated as follows: “What is / was your current / last occupation? That is, what does / did your job specifically consist of? (Specify as much as possible the activities carried out, example: car repair mechanic, dental assistant, primary school teacher, etc.). We refer to your main occupation: the one for which you (or the head of the family) obtain the highest income”. Examples of occupations have varied over time. Occupations are coded in three digits. The occupations were originally coded using the National Classification of Occupations of 1979 and 1994. These classifications have been recoded to adapt them to the EGP scheme and to make the categories comparable over time.

The second key variable is the sports practice of parents. The original formulation of the question in the surveys is as follows: “Could you tell me if your father/mother practises or has practised (in the event that he/she no longer practises, is older or has died) any sport?”. The responses included four options: “Currently practising sports”, “Practised sports in the past”, “He/she has never practised sports”, and “No answer”. The question had to be answered both for the father’s and the mother’s activity. We have created a variable that contains the following categories: i) passive parents (none of them have practised sports), ii) Active parents in the past (at least one of them has practised some sport in the past, not in the present); iii) Active parents in the present (at least one of them practises sports at the time of the survey).

As control variables, in some models we have introduced age and age squared (to control for the potential non-linear effect of age), the educational level (primary, secondary and tertiary education), gender, municipality size, survey year and the number of sports infrastructures close to the dwelling of the interviewee (from none to more than seven).

In the sociology of sport, as in other fields, logistic regression has been the usual technique for analysing a dichotomous dependent variable (Kahma, 2012; Scheerder and Vos, 2011; Scheerder et al., 2005b). However, some methodologists have highlighted the relevance of using a linear probability model even to analyse a dichotomous dependent variable under certain conditions (Hellevik, 2009; Hippel, 2015; Breen et al., 2018). In the field of social stratification, its use is becoming more frequent (Bernardi and Cebolla, 2014a, 2014b; Martínez Pastor, 2017). Andersen and Bakken (2018) have applied it precisely in an analysis of the social stratification of sports practice.

The technique used in this article is a linear probability model (LPM) with robust standard errors. We have found that the predicted probabilities are distributed over the entire range between 0 and 1 (45% of cases between 0.20 and 0.80) - rather than concentrated at one of their extremes - which makes the application of a linear rather than a non-linear model more appropriate. One of the advantages of using an LPM is that the interpretation of the coefficients is easier and more intuitive than with the odds ratios and, unlike other non-linear techniques, allows comparison of the coefficients between different models (Mood, 2010). These types of models are increasingly applied to dichotomous variables for these reasons (Breen et al., 2018).

Table 2 presents the probability of taking part in sport at least once a week. Model 1 contains only the effect of social class (Hypothesis 1 confirmed). This model supports the social stratification in sports participation: although there are no differences between the two upper classes (higher and lower professionals), there exists a gap between these and the rest of categories. The biggest one is with non-manual workers (they practise sports 14.5 percentage points less than the high-level class), semi-unskilled workers (15.5 points less), and the unemployed (18.2 points less). Model 2 introduces, in addition to social class, the sports practice of parents. The effect of having at least one physically active parent is strong and significant (Hypothesis 2 confirmed). For those who have active parents, the probability of taking part in sports increases by 24 percentage points compared to those with passive parents. The likelihood for those with active parents in the past is also higher (16 percentage points) compared to those with parents who have never practised any sport. Although somewhat lower than in the previous model, the gaps between social classes still exist after controlling for this variable.

Model 3 repeats the previous model but controlling for many more variables such as gender, educational level, age, survey year, the size of the respondent’s municipality and the number of sports facilities in the area of ​​residence. Taking all these factors into account, the differences between the upper classes and some of the others disappear; however, not so for semi-unskilled workers and unemployed people. Even controlling for all other variables, the probability that these two groups take part in sports is still 6 points less than for the upper class. Models have been estimated by introducing each of these variables separately. In general, most of the reduction in differences between classes is due to educational level, although gender is also important in explaining the differences between the higher professionals and the non-manual classes.

Finally, model 4 includes, in addition to all the control variables, the interaction between social class and the parents’ sports participation. This model is key in clarifying the research hypothesis. As shown in the table, none of the interactions is statistically significant. That is, the fact that one of the parents takes part in sports increases the probability that the young person will also practise it. However, this positive effect is not more or less significant according to social class. According to this result, the parents’ sports practice would not have a substitute or complementary effect on the inequality of social classes; that is, we would accept the null hypothesis (H3₀).

What happens if we take more intensive sports participation as a dependent variable, for example, doing sport at least three times a week instead of once? Table 3 and Figure 1 present the results for this outcome. The logic of the models is the same as that described in Table 2. In model 1, where only the effect of social class is included, statistically significant differences are appreciated, although in this case the second highest class (lower professionals) practises more sport than the higher professionals. Again, semi-unskilled workers and unemployed people practise less sport, with a difference of around 6 percentage points. According to these results, Hypothesis 1 is confirmed.

In model 2, when parents’ sports participation is included, it can be seen that the effect of this variable is statistically very significant (Hypothesis 2 accepted): it increases the probability of an individual doing sport by 21 points. For those with active parents in the past, the likelihood also increases (by 12 percentage points) compared with those whose parents never practised any sport. In this second model the differences between the higher professionals, the semi-unskilled working class, and the unemployed disappear. When taking into account the other control variables (model 3), the gap between the upper class and the rest remains insignificant, except for the class immediately inferior to the higher professionals, that is, the lower professionals, which continues to present greater sport practice.

Finally, in model 4, the interaction between social class and parents’ sports participation is included. In contrast with ‘at least once a week’ sports practice, the effects are now very significant. The fact that one of the parents is physically active at the moment of the survey increases the likelihood of playing sports in classes that are not at the top of the social structure (Hypothesis 3a confirmed). The results of model 4 show that the effect is greater among skilled manual workers and self-employed people. But it also has a stronger positive effect, albeit to a lesser extent, for semi-unskilled workers (all compared to the higher professionals).

Figure 1 represents the differences discussed above. This shows that parents’ sports participation increases the probability of taking part in sport for all social classes; but in in the case of those with active parents at the moment of the survey (red lines), much more strongly from class III (routine non-manual workers) to the unemployed. So much so that, surprisingly, among those whose parents practise sports, the probability of doing sports is greater for those who do not belong to the highest class. Having active parents in the past (blue lines) increases the likelihood of practising sport compared with having parents who have never practised; but it does so in such a way that the differences between social classes do not vary significantly.

In summary, if we take practising sports at least three times a week as a dependent variable, parents’ sports participation levels out inequalities between class II (lower professional) and the rest, and even opens up a gap between the rest and the higher class to the detriment of the latter. These results support the study’s hypothesis in which it was expected that parents’ sports habits would have a substitute effect, by which the low probability of practising sport among young people of middle or lower social classes would be compensated for if any of the parents were physically active.

Why does the effect of parents’ sports practice affect social classes differently only if individuals take part in sports at least three times a week instead of once? A possible explanation is based on our hypotheses about the interaction between social class and parents’ sport participation and the type of sport that is practised at least three times per week. The data from the surveys indicate that the sports practised once or twice a week are mainly team sports, such as soccer, football, basketball, handball or volleyball. These sports are characterized by not needing much financial resource to practise them, except to pay for sports facilities, which are generally public in Spain.

On the other hand, most of the sports that are practised at least three times a week have to do with gyms. These are maintenance gymnastics, aerobic or anaerobic exercises, or group exercises in gyms. Given that public gyms are scarce in Spain, practising this type of sport incurs ongoing expense, such as paying tuition or a monthly fee, something that is not so common among those who do sports at least once a week. In this sense, non-professional classes (non-service classes in the EGP scheme) have fewer resources or may be more reluctant to spend money on sports in a sustained way, so for them the effect of the parents’ sports practice can be a bigger boost than for the more professional classes, which have greater resources. This is what is known as a substitutive effect for non-service classes in our hypotheses. To verify this explanation further research will be necessary.

According to the results of the article, sporting activity in Spain is more extensive among the professional classes (higher and lower) than in the other social classes. However, the interesting thing is the effect of having physically active parents. When predicting the probability of taking part in sports at least once a week, having parents who practise sports increases that likelihood for everyone, equally. On the other hand, when we delimit the measure to practising sports at least three times a week, the parental influence in the sports practice of the individual is more decisive for the middle and lower classes. The premium received by these groups is so strong compared to the higher professionals that the initial ‘penalties’ are reversed to the detriment of the latter.

The findings of this study have revealed a variable that diminishes and even reverses the gap between social classes with respect to an activity as relevant as sports practice (measured by taking part in sports at least three times a week). The finding could well be summed up this way: if you play sports, your children will do so as much or more so than the upper classes. Young people who belong to the latter group have various advantages when it comes to practising sports, such as greater economic capacity and family support to carry out this type of activity. Individuals who are not at the top of the social structure, on the other hand, can compensate for this initial disadvantage if their parents are physically active. To date, the scientific literature has documented how parents’ sports habits constituted one of the main determinants when it comes to predicting the sports participation of young people. What was not known until now is that this variable is much more decisive for the middle and working classes.

This study does not explain why the influence of parents’ habits does not have the same equalizing effect when it comes to more moderate sport practice, that is, at least once a week. In this case, the impact of one of the parents playing sports is the same for all individuals, in such a way that the gap caused by social class is the same regardless of the parental activity. A possible reason is that people who take part in sports at least once a week are a much less selective group than those who do so at least three times a week and the premium of having physically active parents would be less decisive in predicting a moderate level of sports practice.

The implications of this study are not insignificant. Firstly, another benefit of sports participation can be added to the list of those already known: it levels the opportunity gap that exists between the social classes when it comes to young people taking part in sports. Secondly, and from a practical point of view, it provides evidence for the recommendation that sports participation be encouraged through Sport for All policies, but especially among adults who do not belong to the higher classes. According to this research, the intergenerational transmission of sports practice is stronger among the middle and working classes. Promoting sports among the adult population will ensure an impact on its youth, without the need to invest directly in this group.