fedea

Fundación de Estudios de Economía Aplicada

Some Students are Bigger than Others, Some Students’ Peers are Bigger than Other Students’ Peers by Toni Mora Joan Gil^** Documento de Trabajo 2010-18

SERIE Economía de la Salud y Hábitos de Vida CÁTEDRA Fedea – “la Caixa”

June 2010

* Universitat Internacional de Catalunya & Barcelona Institute of Economics. ** Department of Economic Theory & CAEPS (Universitat de Barcelona) and FEDEA.

Toni Mora Universitat Internacional de Catalunya & Barcelona Institute of Economics

Joan Gil Department of Economic Theory & CAEPS (Universitat de Barcelona)

Correspondence to: Toni Mora, School of Economics and Social Sciences, Universitat Internacional de Catalunya, Immaculada, 22, 08017, Barcelona (Spain) Phone 0034 932541800 (4511) Fax 0034 932541850. Email: tmora@cir.uic.es

(*) We are indebted to the Catalan Mathematics Society, and especially to Antoni Gomà, for providing us with contacts in each high school for preparing the database. We are grateful to the participants at the III Workshop FEDEA-La Caixa on Health Economics (UPF, June 2009) for their comments. Joan Gi acknowledges the financial support of the Spanish Ministry of Science and Technology (Ref.: ECO2008- 04997) and the Generalitat of Catalonia (Ref.: 2009-SGR-359) and Toni Mora acknowledges the financial support of the Generalitat of Catalonia (Ref.: 09SGR102). The authors are also indebted to Josep-Lluís Carrión-i-Silvestre for his help with the programming and to Philip Oreopoulos for his useful comments at an early research stage. The usual disclaimer applies.

Abstract:

This paper analyses the extent to which peer influence on adolescent weight differs in a typical southern European country and in the United States, two geographical areas characterised by different economic, socio-cultural and environmental patterns. Our study is based on a survey of secondary school students containing a rich set of personal data and a wide range of school characteristics and parental backgrounds. After accounting for a large set of control factors and controlling for a combination of school- and neighbourhood-specific fixed effects, instrumental variable estimation and alternative definitions of peers, our results support a more powerful positive and significant effect of friends’ mean BMI on adolescent weight than that reported in previous US-based research.

Keywords: Peer effects, Adolescent behaviour, Obesity/overweight

JEL codes: I12

Resumen:

Este artículo analiza hasta qué punto la influencia de los pares (“peers”) sobre el peso de los adolescentes en un típico país del Sur de Europa difiere del efecto hallado para los EE.UU., dos áreas geográficas caracterizadas por distintas realidades económicas, socioculturales y medioambientales. Este estudio está basado en una encuesta de microdatos sobre estudiantes de secundaria, ofreciendo un rico conjunto de datos personales, familiares y escolares. Tras considerar un amplio de controles, efectos fijos de escuela y de área de localización, estimación por variables instrumentales y distintas definiciones de efectos pares, nuestros resultados apoyan la existencia de un (más poderoso) efecto positivo y significativo del BMI medio del grupo de amigos sobre el peso de los adolescentes escolares que el efecto los hallado por la literatura para el caso de los EE.UU.

Palabras clave: Efectos pares, Comportamiento de los adolescentes, Obesidad/sobrepeso Código JEL: I12

Introduction

Obesity is a global public health problem that is not restricted to the wealthy nations. The rapid increase of the obesity epidemic in recent decades is particularly alarming in children and adolescents, as the condition may pass into adulthood and create a growing health burden for the coming generations. According to the CDC report (2010) the prevalence of overweight in US teenagers, aged 12-19 years, has more than doubled in recent decades, reaching 17.8% by 2005-2006. This upward trend is also observed in European countries. The Health Behaviour School-aged Children survey conducted in 2001-02 indicated that 24% of 13-year-old girls, 34% of 13-year-old boys, 31% of 15-year-old girls and 28% of 15- year-old boys in Europe were overweight.¹ Empirical evidence shows that overweight during childhood and adolescence is a major cause of ill health in those ages but also in adulthood, increasing the risk of hypertension, cholesterol, sleep apnoea, diabetes type 2, low self-esteem, and discrimination in education and work settings (Dietz, 1998; Reilly et al., 2003).

Research on the determinants of child/adolescent overweight has highlighted the influence of parents (e.g., genes, transmission of values and norms, food choices, education), race and ethnicity, the role of food availability and prices, and the emergence of an “obesogenic environment” through changes in home, school, transport and urban policies and commercial food activities which encourage physical inactivity and poor dietary practices (Koplan et al., 2005). In the last decade or so, a new body of literature has emerged in the field of health economics –associated with social interaction models – which stresses the influence of peers on adolescents’ health status. Several papers have suggested that friends or classmates have a significant positive impact on health-related behaviours in the young such as smoking, binge drinking or illicit-drug use (e.g., Norton et al., 1998; Gaviria and Raphael, 2001; Powell et al., 2005; Lundborg, 2006; Clark and Lohéac, 2007). However, far less is known about the effect of social networks on the obesity status of the population. Using a social network of adult people with repeated measurements over a period of 32 years, the study by Christakis and Fowler (2007) concluded that social networks (e.g. ties between friends, siblings, spouses, neighbours) facilitate the spread of obesity. Similarly, the studies by Trogdon et al. (2008), Halliday and Kwak (2009) and Renna et al. (2008) used cross-sectional surveys of health-related behaviours in a sample of American adolescents and reported a positive association between friends’ weight and adolescent body weight.²

¹ A geographical pattern can be observed in European countries. Overweight for these age groups is highest in the UK and in other southern European countries like Greece, Italy, Portugal and Spain. The Scandinavian and central European countries show lower levels (WHO, 2007).

This paper seeks to study the role of peer effects in adolescent body weight (proxied by the BMI) in a radically different socio-cultural context. Namely, our goal is to assess to what extent the influence of peers on adolescents’ weight differs in a typical southern European country and in the United States, two geographical areas characterised by specific economic, socio-cultural and environmental patterns. International data on health-related behaviour among adolescents show that peer group pressure varies greatly across countries. Interestingly, the HBSC 2001/2002 data show the existence of significant cross-country differences in the size of friendship groups, which is taken as a proxy indicator of exposure to peer influences (WHO, 2004). While adolescents in English-speaking countries (Canada, the US and the UK) and Scandinavian countries report high percentages of groups containing three or more close friends of the same gender (around 80-90%), in Mediterranean and eastern European countries such percentages are far lower.³ The data also indicate that the amount of time young people spend with their friends (a strong predictor of peer influence) varies widely across countries. In principle, one might expect to find an association between frequent contact with friends and a higher likelihood of initiating or maintaining different types of risky behaviours.⁴ Among 13-year-olds, the percentage of boys (girls) meeting with friends four or more evenings per week is 29% (25%) in the US, and 41% (32%) in the case of Spain.⁵ Note that behind this geographical variation in adolescent social networks cross-country differences persist in socio-cultural, religious, life-style and environmental patterns which influence the identity and socialization process of youth and, thus, the behaviour of peer groups.

We use data from a single survey of secondary school students in Catalonia (Spain) conducted in 2008 which compiled a rich set of personal data and a wide range of school characteristics and parental backgrounds. Students were asked to identify their specific friends in class (without any limitation on number or gender). After accounting for a large set of control factors, controlling for a combination of school- and neighbourhood-specific fixed effects, instrumental variables estimation and alternative definitions of peers, we found a positive and significant effect of friends’ mean BMI on adolescent weight. These findings corroborate those of previous US- based research, but we observed a far stronger positive peer effect in females. Other significant positive effects were found when alternative peer pressure definitions were considered (those who do not consider themselves as leaders and those who have the same friends as the previous year). The paper is organized as follows. The next section discusses the theoretical framework of peer effects. Section 3 describes the data and presents the empirical methodology, Section 4 reports the main results, and finally section 5 concludes.

2 Using a different framework, Costa-Font and Gil (2004) found that social interactions (proxied by the intensity of meeting friends) through their impact on individual’s self-perceived image which ultimately affects weight-related behaviours, were negatively associated with adult BMI, obesity and overweight. Similar results were reported by Clímaco et al (2007) applied to a sample of Portuguese adolescents.

³ For the 13- (15-) year-old boys, the share with three or more close friends is around 83% (77%) in the US but only 63% (57%) in Spain. Among girls of the same age, the rates are 89% (81%) in the US and 60% (51%) in Spain. The data also show a cross-country differential pattern according to gender.

⁴ However, peer contact is also important for the development of protective factors against unhealthy lifestyles.

⁵ These country-specific differences are smaller among 15-year-olds, as meeting with friends increases gradually with age.

1. Social interactions within the classroom

Formatat: Numeració i pics

Adolescence is recognised to be a critical developmental period for young people, in whom they are faced with many new situations: continually adjusting to physical changes, exploring their sexuality, establishing their personal identity, seeking greater independence and increasingly relying on friendship groups. This period offers opportunities for progress but also presents risks to health and well-being (Irwin et al., 2002). It is during these years when social interactions with friends, classmates, neighbours etc. are most intense, decisively affecting adolescents’ health-related behaviours and reinforcing norms or values which ultimately shape individual behaviour. Being liked and accepted by peers is crucial for adolescents’ health status, and interactions with friends tend to improve social skills and strengthen the ability to cope with stressful events. In contrast, those who are not socially integrated are far more likely to exhibit difficulties with their physical and emotional health. The influence of peers on adolescents’ health is a complex issue that provides both protective and risk factors (Berndt, 1999).⁶

The empirical evidence on peer effects suffers from what is known as the reflection problem (Manski, 1993) which influences the identification of endogenous social effects. According to this framework individual and peers’ weight may be correlated for three different reasons: (i) the direct influence of peers’ weight on individual weight, which produces the “endogenous effect” and implies a causal interpretation; (ii) the indirect influence on individual weight caused by the exogenous characteristics of the peer group (“exogenous or contextual effects”) and (iii) the influence of a common set of unobservables on both individual and peers’ weight (“correlated effects”). According to this set up the first two effects reflect social interactions – although with different policy implications – whereas correlated effects are a statistical, non-social phenomenon (Clark and Lohéac, 2007). Given the plausibility of this hypothesis, the main empirical challenge with observational data is to discriminate between these effects and to identify a causal relationship of peer influence on adolescent weight. Another challenge when analyzing peer effects is the fact that schools, classrooms and peer groups are not formed randomly.⁷

⁶ See Cotterell (2004) for an excellent analysis of how adolescents are in general influenced by the type of social ties emerging from a variety of settings like cliques, crowds or gangs.

To identify this causal effect we first adopt the assumption made by some studies (Norton et al., 1998; Gaviria and Raphael, 2001; Lundborg 2006; Trogdon et al., 2008 and Renna et al., 2008) that only one type of social effect exists – the endogenous effect – and that contextual influences are inexistent. This is similar to hypothesizing that the influence that peers may exert on adolescent weight is channelled through the reference group weight status (Trogdon et al., 2008). Second, we control for the influence of correlated effects which appear when individuals in the same group tend to behave similarly because they are exposed to similar individual characteristics or environments (Manski, 1993). This is the case, for instance, of classmates eating the same school diet or doing the same sports at school, and of parents when choosing where to live since they are indirectly choosing their children’s peers at neighbourhood and school level. Here it could be argued that the presence of certain unobserved factors (i.e., parks, sports facilities) leads families to sort into school areas or districts which ultimately affect the bodyweight in a similar manner. Another source of correlated effects is the fact that classmates and friends tend to share some activities (e.g., physical exertion) which are unobserved by the econometrician. We address all these issues by including school- and neighbourhood-specific fixed effects. Finally, we apply instrumental variable (IV) estimation and instrument friends’ weight using information (reported by adolescents) on friends’ mothers’ years of education, friends’ mean age and the share of friends whose parents are divorced or single.

⁷ On the one hand, the usual solution has relied on exploiting experimental or quasi-experimental designs to separate social effects in the classroom (which are the combination of endogenous and exogenous peer effects), e.g. those arising from misbehaviour or satisfaction with school (see Lavy and Schlosser, 2007), individual motivation or effort from correlated effects (e.g., Sacerdote, 2001; Zimmerman, 2003; and Hanushek et al., 2003). On the other hand, using observational data, recent literature has focussed on identifying exogenous variations to explain the formation of peer groups. Following Hoxby (2000)’s strategy, some recent papers (such as Lavy and Schlosser, 2007 or Proud, 2008), have made use of the variation in the distribution of females across cohorts using the proportion of girls within a grade as a measure of the peer group.

A further problem outlined by Manski (1993) is that researchers rarely know who exactly constitutes the peer group and have to impose their own ad-hoc definition of it. Treating peer effects as a consequence solely of predetermined peer features such as ability, only captures part of the situation: peer group may come into being through unobservables such as leisure activities or lifestyle conditions. Whilst school and class composition is determined by neighbourhood characteristics such as average income per capita, the makeup of the real reference group (peers within classroom) relies on other sociological factors. Bishop et al. (2004) affirms that teenagers interact with each other based on time allocation between activities according to clique norms (such as extracurricular activities and socializing). Therefore, it is important to estimate peers’ influence by means of databases that account for nominated friendship relationships. Fortunately, our cross-sectional dataset permits us to use nominations of classroom friends to form reference groups.

2. Data and Method

Formatat: Numeració i pics

We use data from a single representative survey of secondary school students in Catalonia, one of Spain’s richest and most densely populated regions, collected between February and June 2008. The survey was targeted at secondary students in four specific academic years: the last two years of compulsory secondary education (14-16 years old) and the two years of higher secondary education (16-18 years old). The questionnaires were completed by students and maths teachers were approached to participate in the survey and help with student data collection. The questionnaire (supplied on-line with questions presented randomly) contained six blocks of questions: personal data (including anthropometric information), school characteristics, maths teaching questions, parental background information, awareness and motivation, and lifestyle habits.⁸ The final sample contains information on more than 3,000 students from 91 high schools.

⁸ Since not all high schools had computer room facilities or enough time schedules, some schools received the questionnaire in paper format. None of the students had access to the questionnaire prior to responding, which avoids attrition effects. Students were free not to respond to some questions.

⁹ As the under-representation of some specific areas or schools due to their managerial characteristics (public, semi-private or private) might have given rise to sample selectivity some administrative information was obtained from the Catalan Ministry of Education to ensure the representativeness of the sample.

Our cross-sectional dataset presents a number of advantages over the data used in previous studies. First, we define peer groups using nominated friends within the classroom. Few other studies have been able to define peer pressure at such a fine level.¹⁰ Second, in contrast to the Add Health data, in which students were limited to listing up to 10 friends (5 male and 5 female), in our dataset students were asked to identify or nominate as many close classroom friends as they wished. Third, as we have information on all listed friends we can consider more sophisticated social networks such as asymmetries in classmates’ relationships. On the one hand, following Christakis and Fowler (2007) we can define the “all nominated friends” peer effect when the group is formed by all the listed friends chosen by each adolescent. A narrower definition could be derived (“mutual friends” peer contact) where the reference group is composed only by those who also reciprocate the friendship. However, we prefer the first peer definition because in general students are affected not only by their mutual relationships; even if some peers do not correspond reciprocally, a student is actually influenced by those who he/she has nominated. Fourth, in contrast to the Framingham Heart Study (but similar to the Add Health) our social networks are more connected and less varied as they take place within the same classroom (Cohen-Cole and Fletcher, 2008). Fifth, the data allow us to address the issue of endogenous sorting. While students cannot decide which class they will belong to and so cannot sort themselves into classes with other pupils who are similar to them, sorting across schools may take place through parents’ decisions regarding where to live based on the quality of schools in the area (Lundborg, 2006).¹¹ In order to account for this effect, our estimations will include school- and neighbourhood-specific fixed effects. Finally, as noted above, the data include information regarding juvenile behaviour and social networks from a southern country in the EU characterised by a socio-cultural and environmental context which favours an intense social life and which ultimately affects the formation and dynamics of peer group pressure

The dependent variable of this study is adolescent BMI, calculated as the ratio of individual weight (measured in kilograms) to squared height (measured in metres). Since weight and height information is self-reported by each interviewee, there is a potential for measurement error, as the literature acknowledges.¹² We therefore adjusted self-reports of height and weight by applying a standard correction procedure (e.g. Cawley, 2000; Chou et al., 2004; Cawley and Burkhauser, 2008). To do so we estimated the relationships between measured and self-declared weight (height) found using a sample of adolescents of the same age group (based on the Catalan Health and Examination Surveys, 2006)¹³ and transferred these values to our dataset.

¹⁰ The exceptions are Clark and Lohéac (2007), Trogdon et al. (2008), Renna et al. (2008) and Halliday and Kwak (2009) who used the same dataset, the National Longitudinal Study of Adolescent Health (Add Health), and Christakis and Fowler (2007) who used the Framingham Heart Study. ¹¹ In Spain schools do not sort students according to ability.

The empirical model for (corrected) BMI of pupil i with peer group j in class c and in school s is,

\[B M I _ {i j c s} = \beta_ {0} + \beta_ {1} B M I _ {j c s} + \beta_ {2} X _ {i c s} + \lambda_ {s} + \lambda_ {n} + \varepsilon_ {i c s}\tag{1}\]

where our variable of interest is the average BMI among peers attending the same classroom,¹⁴ represents the covariates considered, and are school- and neighbourhood-specific fixed effects and is the individual-specific error term. The baseline estimation considers the average BMI of “all nominated friends” as the central peer influence; however, other alternative peer effects will also be investigated in the sensitivity analysis. Although this definition of friends’ weight is arguably a better measure of peer pressure, it is likely to be endogenous. Thus in addition to instrumental variable estimation, we define the average BMI of all classmates since this is considered to be a more exogenous social network effect provided that the assignment to a certain class or grade level is determined exogenously by year of birth. Of course, this broader definition of peer group is not expected to generate the same type of social influence as friend-level peer pressure, which may perhaps operate through the imposition of a BMI standard or social norm (e.g., Burke and Heiland, 2007).

As controls, we use a long list of covariates including adolescent characteristics (age, gender, immigrant status, smoking status, alcohol consumption, sleeping time, number of hours reading books and watching TV), family characteristics (family type, difference in years between the mother and the adolescent, mother’s education and health status, parental involvement in homework)¹⁵ and some class-level characteristics (percentage of mothers with university education; percentage of female students and percentage of female students within cliques).

¹² In general, high school students of both sexes consider themselves to be taller and thinner than they actually are (e.g. Farré-Rovira et al., 2002; Danubio et al., 2008).

¹³ OLS regressions of measured height (weight) on self-reported height (weight) gender, age and age square were derived.

¹⁴ Note that adolescent BMI was subtracted from the computation of peers’ weight.

In accordance with the thesis of Heckman et al. (2006), which proposed that (latent) noncognitive skills and cognitive skills are equally important factors for a successful social and economic life and for the adoption of less risky behaviour, our econometric specification includes a measure of individual awareness. As the questionnaire asked for information on awareness (one of the five components of personality, the one related to ability), this variable was constructed through a factor analysis.¹⁶ Although other measures can be affected by peers, personality traits are specific to each individual and are not so likely to be influenced. So we are able to include this covariate as a substitute for student fixed effects, and it constitutes a different source of ability. Table 1 shows the definition of the list of the covariates considered in all our estimations.

[Insert Table 1 about here]

As noted above, to control for the presence of correlated effects we use school-specific fixed effects, (e.g., Lundborg, 2006; Clark and Lohéac, 2007; Trogdon et al., 2008 and Renna et al., 2008) and also neighbourhood fixed effects, (captured by the residential postal code and the school’s postal code).¹⁷ This set of fixed effects would eliminate any unobserved school or neighbourhood characteristic that might influence the weight of subjects and peers who are in the same classroom and school or who are exposed to the same local environment. For instance, school policies on nutrition or physical activities in the curriculum shared by all students (although the latter might be controlled by classmates’ average physical activity) will lead to correlated effects.¹⁸

¹⁵ Other covariates such as physical exercise, health status and extracurricular activities or father’s health status were not included due to their non-statistical significance. These and other covariates were finally not considered for efficiency reasons. For instance, we controlled by means of the number of nominations as in Calvó-Armengol et al. (2009) which allows us to control for unobserved network-specific components.

¹⁶ Due to time constraints for applying the survey, we conducted several interviews with psychologists in order to ensure that the relevant questions were included. We followed Alonso-Tapia and Arce-Sáez (1992), which is specific for Spanish teenagers. We computed Cronbach's alpha statistic for the scale formed from the pairs of variables (0.76). A factor analysis allowed us to construct two factors related to personality. The Kaiser-Meyer-Olkin measure of sampling adequacy was satisfactory (0.81). Accordingly, the factor scores were re-scaled to variables ranging from 0 to 1 so indicating the degree of personal consciousness.

¹⁷ Weight and BMI display dissimilarities based on the district of residence - see Mora (2010) for observed differences within the city of Barcelona, which is representative for the Catalan case.

¹⁸ Lunch diets in Catalonia are controlled by the regional government through regular inspections.

To address the issue of the potential endogeneity or bi-directionality of the peer relationship we estimated equation (1) using IV estimation. Following standard practice (Case and Katz, 1991; Gaviria and Raphael, 2001; Lundborg, 2006; Clark and Lohéac, 2007; Trogdon et al., 2008 and Renna et al., 2008) we assume that contextual effects are non-existent and therefore that average background characteristics of peers can be used as instruments. While some studies have used friends’ parental obesity as an instrument for peer influence (Cawley 2000, 2004; Brunello and D’Hombres, 2007; Trogdon et al. 2008 or Renna et al. 2008), others have relied on the characteristics of other friends or classmates (e.g., living in single-parent families or in an apartment, parents born outside the country, parental health status or college education). We instrument friends’ BMI by using the average of friends’ mothers’ years of education, friends’ average age, the share of single/divorced friends’ parents and the share of their peers who smoke. These variables are presumed to be valid instruments as long as we assume that the influence of peers’ background characteristics on an adolescent’s weight are not direct but indirect through their impact on the BMI of the peer group. Obviously, we conducted a series of tests to ensure the validity of the instruments.

3. Empirical results

3.1 Descriptive data

The average BMI of adolescents in our sample was 21.81 (25.45 among the overweight group), and the prevalence of overweight and obesity was 16.5% - close to the average Spanish levels (17%) but below US levels (17.6% for adolescents aged 12-19).¹⁹ Interestingly, the average BMI of the “all nominated friends” (21.78) peer group was slightly lower than the individual BMI of the whole sample; the difference is statistically significant. This suggests that fatter students are less likely to be nominated by the rest of students, and so peer groups have a slightly lower body weight.²⁰

¹⁹ The overweight category was defined using the age- and gender-specific international cut-off points calculated by Cole et al. (2000). Following the recommendations of the International Obesity Task Force, by pooling cross-sectional data on BMI for children from six countries (Brazil, Great Britain, Hong-Kong, the Netherlands, Singapore and the United States) and using the centile based method (ensuring that at age 18 they matched the adult cut off of 30 kg/m²) these authors were able to calculate BMI cut-off points for overweight and obesity for children aged 2-18 years.

²⁰ This is consistent with evidence showing that 10- to 11-year-old children prefer as friends other children with a wide variety of handicaps to children who are overweight (Dietz, 1998).

Note that the mean number of “all nominated friends” in our dataset is 6.59 (6.36 among males and 6.81 among females). These figures are notably higher than those reported by studies based on the Add Health database. As the above mean difference by gender is statistically significant (p-value=0.00) we split the sample by gender in order to explore peer impact on individual BMI. However, about 10.8% of students do not list anyone as a friend. To have a precise idea of the variability of our peer pressure covariate, Figure 1 presents the distribution of friends’ nominations by BMI status. The inspection of the figure shows that overweight students tend to nominate fewer close friends (6.73) than their normal-weight counterparts (7.08) and, although the data are not shown, tend to receive less reciprocity from their nominations than their normal-weight counterparts. The ratio between the number of students’ nominations as friends and the number of times nominated is 0.98 for overweight adolescents but considerably lower, 0.75, for normalweight teens. This result is likely to condition inference results when asymmetries are considered. Thus, the overweight do not tend to nominate as much as they are nominated by their classmates.

[Insert Figure 1 about here]

In general, students tend to self-select their relationships within the classroom according either to their academic skills or to other more general factors like sports or music activities, but not according to their anthropometric measures. Kang (2007) has shown that weaker students have closer interaction with other weak students than with strong students, thus inducing sorting out effects. Our argument is that students base their nominations not on BMI but on academic and/or non-academic skills. In this regard, the share of overweight students nominating and receiving reciprocity from those similar to them differed little from the percentage found in their normal weight counterparts.

Finally, the mean age of our sample of adolescents was around 16 years old and 52% were females (Table 1). Almost 40% stated that they had drunk alcohol during the previous weekend and 15% reported smoking on a daily basis. Interestingly, around 30% of adolescents’ mothers had completed university studies; 14% of adolescents stated that their mothers had poor health status and almost 50% declared no family involvement in homework.

3.2 Peer influence

Table 2 reports a set of OLS estimations of equation (1) accounting for several econometric specifications. Inference analysis is based on robust standard errors, taking into account clustering of observations at the classroom level. The first column, which excludes any controls and fixed effects, shows a positive relation between adolescent BMI and mean peer BMI in our sample of Spanish teenagers. An increase of 1 unit in the average BMI of “all nominated friends” was associated with a 0.17 point increase in the respondent’s BMI, which corresponds to a marginal effect of 17.07%. In column (2) we estimate the model including both adolescent and family controls. The results indicate that friends’ weight is correlated with an adolescent’s own weight even after controlling for this list of covariates. As expected, the peer pressure falls considerably, although it remains statistically significant. Now, the estimated marginal effect is around 10.79%. Column (3) includes classroom-level characteristics as additional controls. We find that the “all nominated friends” peer effect is still positive and statistically significant, with an estimated marginal effect of 9.08%. Finally, in a further step, in columns (4) and (5) in Table 2 we estimate the model adding school- and neighbourhood-specific fixed effects under the assumption that weight or BMI may display disparities depending on the school and/or district of residence. Similarly, families might sort into school areas based on amenities (e.g., recreation areas and parks) which could be correlated with adolescent weight (Trogdon et 2008). In fact, neighbourhood characteristics condition the type of extracurricular activities that adolescents may engage in or the kind of amenities correlated with weight (i.e., sports facilities or recreation parks). We captured these effects using the residential postal code and the school centre postal code. Our results show that the “all nominated friends” peer effect is irrelevant in explaining students’ BMI.

[Insert Table 2 about here]

Results for other covariates indicated lower BMI in females but higher BMI in older adolescents. Interestingly, smoking habitually, sleeping time and having no family help with homework were negatively associated with adolescent BMI, but reading activity and watching TV presented positive associations. Finally, as expected, low maternal level of schooling correlated with higher levels of adolescent BMI. Hereafter, the peer group coefficients are the focus of the rest of the paper.

In Table 3 we present the results of the IV estimations aimed to control for the potential endogeneity of our variable of interest due to the bi-directionality of the effect. However, as we have evidence of heteroskedasticity in the data [Pagan-Hall test = 101.53, pvalue=0.0000] we opted to re-run the model (1) using the two-step General Method of Moments (GMM) estimator, which exploits the orthogonality conditions between the residuals and the instruments. This allows a more efficient estimation than the IV estimator in the presence of heteroskedasticity of unknown form. In addition, we include the complete set of controls and the two types of fixed effects and split the sample between male and female adolescents given that the amount of body fat changes with age and gender. To ensure the validity of the instruments, a number of diagnoses were performed.²¹ The high values of the Kleibergen-Paap rk LM statistic of underidentification allow rejection of the null hypothesis that the equation is under-identified. Similarly, the low values of the Hansen’s J statistics (p-value>0.05) mean that we do not reject the null hypothesis and conclude that the instruments are orthogonal to the errors, thus validating our instruments. Compared to the Stock and Yogo (2005) critical values, the Wald Fstatistics of weak identification show the bias of the GMM estimation to be less than 15% (columns 1 and 2) and 25% (column 3) of the OLS bias.

Interestingly the first column (whole sample) shows that the GMM estimate of the “all nominated friends” peer effect is positive and statistically significant, indicating that when friends’ mean BMI is one BMI unit higher, then individual BMI is 0.57 units higher (similar to the estimates in Trogdon et al., 2008). This suggests that neglecting the potential endogeneity of peer group BMI may result in a downward bias in the estimated peer effects on adolescent weight: that is, OLS estimation underestimates the real impact of peers’ influence since unobservables would be positively correlated to peers’ measurements. As in Renna et al. (2008), our results confirm that while the GMM “all nominated friends” peer effect for females is positive and highly significant (0.71), for males this influence becomes insignificant, although our sample from a typical high school system in a southern EU country provides stronger effects.²²

²¹ Regressions in Tables 3 and 4 use the instrument sets that maximize the Wald F-statistic of weak identification and minimize the Hansen statistic.

[Insert Table 3 about here]

3.3 Alternative social influences

Other alternative measures of peer contact are investigated in Table 4 to check robustness. Again, the same set of GMM estimations using the complete set of controls along with school- and neighbourhood-specific fixed effects are accounted for. As can be observed, overidentification and weak instrument tests validate our estimation results. However, note that the number of observations has been reduced, as a consequence of restricting the sample to those who form the new definition of the reference group.

In columns (1) and (2) we analyse the impact of a closer social network effect. The first column examines the influence of the “mutual nominated friends” peer pressure where the reference group is composed by only those classmates who also reciprocate the friendship. In the second column we construct the reference group taking advantage of a yes/no survey question designed to establish whether the adolescent’s habitual listed friends are mostly the same as those of the previous academic year. Interestingly, while our results do not find any significant evidence of a peer effect caused by mutual friendship (which confirms our initial suspicion that students are in general more affected by a broader reference group than by their mutual relationships alone) the GMM results reveal a larger positive and statistically significant impact of the BMI of the same last year friends on adolescent weight (0.66 vs. 0.57). The first of these results suggests that cliques are not entirely closed, tight-knit circles, while the second shows that those individuals who maintain the same peers as the previous year will be exposed to a greater influence than those who make new friends (Christakis and Fowler, 2007).

[Insert Table 4 about here]

Similarly, in columns (3) and (4) two alternative social network influences are investigated. Based on information on whether adolescents have changed residence or school in the last three years, in column (3) we measure peer influence by grouping the nominated friends who remained in the same district or school over this period. We find a positive and significant GMM impact of this peer group BMI on adolescent weight (0.51), close to the baseline estimate. In column (4) our definition of peer effect excludes any leadership role as we understand that adolescents who consider themselves as leaders in their clique are more influential over their friends than easily influenced by others. Interestingly, we find that the mean BMI of a reference group that excludes these influential adolescents positively affects individual adolescent weight, and that this influence is even larger than at baseline (0.69). Thus, as we expected, a stronger peer effect is obtained after dropping the individuals less likely to be influenced by their peers.

²² Similar results are found when the estimations are performed using the Limited-Information Maximum Likelihood (LIML) estimator.

3.4 Influence of classmates

Our identifying assumption, namely that the background characteristics of peers have no direct impact on adolescents’ weight and can therefore be used as instruments, would be violated if the selection of friends (and thus selection of friends’ parents) is correlated with BMI (Trogdon et al., 2008). In section 4.1 we found that there is no sorting, at least, based on anthropometric measures. We therefore applied an alternative strategy that defines peer groups at a broader level, namely assuming that all classmates constitute the reference group. As the assignment to class level is made by year of birth, this definition of peer effect does not suffer the endogeneity problem. Moreover, this approach allows us to collapse unobservables at the classroom level.

The last column in Table 5 shows the results. The OLS estimation indicates a high positive and statistical significant impact of the mean “all classmates” BMI on individual weight (0.85), being robust to the inclusion of school- and neighbourhood-specific fixed effects and a wide set of controls for other confounders. Interestingly, this influence of peer contact is the highest reported in this study. As observed by Eisenberg et al. (2005), this evidence is consistent with the fact that around twenty-five students share classrooms for almost thirty-five hours a week; therefore, it is highly likely that although students share their leisure time and study with a few specific students within their clique, they will also be influenced by the rest of their classmates. Similarly students are unlikely to restrict themselves to their tightly-knit cliques and reject all other relationships at classroom level. Granovetter (2005) argues that smaller networks are connected to other networks by the strength of weak ties, and that this also occurs in classrooms. Therefore, information flows in a way that overlaps cliques.

4. Concluding remarks

This paper investigates the influence of peer effects on adolescents’ body weight (proxied by the BMI) for a sample of Spanish students in which, compared to the most analysed environment in the literature (the US), adolescents’ social influences emerge and develop inside a clearly different socio-cultural context. We use data from a single survey of secondary school students (14 to 18 years-old) in Catalonia (Spain) which contains information regarding friendships inside classrooms. Our approach improves on previous reports because we are able to use complete information regarding all the friends nominated by high school students without any limit on numbers. After controlling for a wide set of controls, school and neighbourhood fixed effects, GMM estimation and alternative definitions of peers, we corroborate the previous findings of a significant positive effect of friends’ mean BMI on adolescent weight. Stratifying the sample by gender, we find evidence of a significant (insignificant) positive estimated effect of nominated friends’ BMI on female (male) body weight. These results support other evidence suggesting that adolescent girls are in general more aware of and concerned with, their own body weight and corporal image than boys and are therefore more likely to be receptive or influenced by their peers. Some research even suggests that some of these social influences may degenerate and be associated with unhealthy weight-control behaviours like self-induced vomiting, laxatives, diet pills or fasting (Eisenberg et al., 2005).

While the impact of our “all nominated friends” peer group influence on adolescent weight is similar to the estimate found by Trogdon et al. (2008) based on US data (Wave 1 of the Add Health), we report a more powerful impact of BMI of this peer group on the weight of the Spanish adolescent females. Our data also indicate a comparatively stronger effect of “all classmates” peer pressure, which is considered a more exogenous peer measure, on adolescent body weight. In addition, the peer effect estimates derived from our sample of Spanish adolescents are greater than those reported by Renna et al. (2008) using the abovementioned Add Health dataset.

In all, the results reported in this study can contribute to explain, along with other factors, the rapid increase of adolescents’ overweight in the last decades. Moreover, the findings are consistent with the strand of the literature that supports the premise that overweight/obesity -as other health-related behaviours- are a social phenomenon. As such, not only the family and the community affect weight-related behaviours of children and adolescents, but also peer contact by shaping self-perceived corporal image affects ultimately adolescents’ health status. The extent to which adolescents respond to social pressure may have important policy implications. As such, the presence of social multiplier (spillover) effects contributes to amplify the impacts of policy interventions aimed at reducing overweight/obesity levels among adolescents. Undoubtedly, public health interventions to curb the epidemic will require a co-ordinated effort between the different actors involved (e.g., health professionals and administrators, teachers, parents, food producers, retailers, caterers, advertisers, sport planners, urban architects, politicians, legislators, etc.) which demonstrates the complexity of the issue.

Some caveats should be mentioned when interpreting the results. First, our analysis is based on a cross-section of schools and the understanding of the dynamics of peer influence, which may change as adolescents’ age, needs to be investigated with the use of a panel dataset. Second, our measure of peer pressure is based on listed school friends, a definition that may ignore the role played by other influential friend or partner networks outside the school. Third, the design of our dataset limits the use of other family background characteristics which might provide more relevant instruments. Finally, the paper does not analyze the mechanisms operating behind the influence of peers on adolescent’s weight.

References

1. Alonso-Tapia, J. and Arce-Sáez, E. (1992) Expectativas de control y motivación: el cuestionario ECO. In Alonso Tapia, J. (ed.) Motivar en la adolescencia: Teoría, evaluación e intervención. pp. 135-175. Madrid: Servicio de Publicaciones de la Universidad Autónoma.

References

1. Akerlof, G.A. and Kranton, R.E. (2002) Identity and schooling: some lessons for the Economics of Education. Journal of Economic Literature, 40, 1167-1201.

References

1. Berndt, T. (1999) Friendship and friend’s influence in adolescence, in: Muus R., Porton H. (eds.) Adolescent Behaviour and Society, MacGraw-Hill, Boston, 1999.

References

1. Bishop, J.H., Bishop, M., Bishop, M., Gelbwasser, L., Green, S., Peterson, E., Rubinsztaj, A. and Zuckerman, A. (2004) Why we harass nerds and freaks: a formal theory of student culture and norms. Journal of School Health, 74(7), 235-251.

References

1. Brock, W.A. and Durlauf, S. N. (2001) Discrete choice with social interactions, The Review of Economic Studies, 68(2), 235-260.

References

1. Brunello, G. and D’Hombres, B. (2007) Does body weight affect wages? Evidence from Europe. Economics and Human Biology 5: 1-19.

References

1. Burke, M.A. and Heiland, F. (2007) Social dynamics of obesity, Economic Inquiry, 45(3), 571-591.

References

1. Calvó-Armengol, A., Patacchini, E. and Y. Zenou (2009) Peer effects and social networks in education, Review of Economic Studies, 76: 1239-1267.

References

1. Cawley, J. (2000) An instrumental variables approach to measuring the effect of body weight on employment disability. Health Services Research, 35 (5, Part II): 1159-1179.

References

1. Cawley, J. and Burkhauser, R.V. (2008) Beyond the BMI: the value of more accurate measures of fatness and obesity in social science research, Journal of Health Economics, 27: 519-529.

References

1. Centers for Disease Control and Prevention CDC, National Center for Health Statistics, 2010. Health, United States, 2009: With Special Feature on Medical Technology. Hyattsville, MD. 2010.

References

1. Christakis, N.A. and Fowler, J.H. (2007) The spread of obesity in a large social network over 32 years, The New England Journal of Medicine, 357: 370-379.

References

1. Chou, S-Y., Grossman, M. and Saffer, H. (2004) An economic analysis of adult obesity: Results from the Behavioural Risk Surveillance System, Journal of Health Economics, 23: 565-587.

References

1. Clark, A.E. and Loheac, Y. (2007) It wasn’t me, it was them! Social influence in risky behavior by adolescents, Journal of Health Economics, 26, 763-784.

References

1. Climaco, M.I., Barros, P.P. and Lourenço, O. (2007) Can adolescents’ obesity be a consequence of peers lifestyles?, IHEA 2007 6^th World Congress: Explorations in Health Economics Paper (available at SSRN: http://ssrn.com/abstract=992964).

References

1. Cohen-Cole, E. and Fletcher, J.M. (2008) Is obesity contagious? Social networks vs. environmental factors in the obesity epidemic, Journal of Health Economics, 27(5), 1382-1387.

References

1. Cole, T. J., Bellizzi, M.C., Flegal, K.M. and Dietz, W.H. (2000) Establishing a standard definition for child overweight and obesity worldwide: international survey, British Medical Journal, 320: 1240-1243.

References

1. Costa-Font, J. and Gil, J. (2004) Social interactions and the contemporaneous determinants of individuals’ weight, Applied Economics, 36(20): 2253-2263.

References

1. Cotterell, J. (2004) Social Networks and Social Influences in Adolescence, Routledge, NY.

References

1. Danubio. M.E, Miranda, G. Vinciguerra, M. G, Vecchi, E. and Rufo, F. (2008) Comparison of self-reported and measured height and weight: Implications for obesity research among young adults, Economics and Human Biology, 6: 181-190.

References

1. Dietz, W.H. (1998) Health consequences of obesity in youth: childhood predictors of adult disease, Pediatrics, 101: 518-525.

References

1. Eisenberg, M.E., Neumark-Sztainer, D., Story, M., Perry, Ch. (2005) The role of social norms and friends’ influences on unhealthy weight-control behaviors among adolescent girls, Social Science and Medicine, 60, 1165-1173.

References

1. Etilé, F. (2008) Social norms, ideal body weight and food attitudes, Health Economics, 16(9), 945-966.

References

1. Farré-Rovira, R., Frasquet-Pons, I., Martínez-Martínez, M.I. and Romá-Sánchez, R. (2002) Self-reported versus measured height, weight and BMI in Spanish Mediterranean teenagers: Effects of gender, age and weight on perceptual measures of body image, Annals of Nutrition and Metabolism, 46: 68-72.

References

1. Gaviria, A. and Raphael, S. (2001) School-based peer effects and juvenile behaviour, Review of Economics and Statistics, 83(2): 257-268.

References

1. Gil, J. and Mora, T. (2009) The determinants of misreporting weight and height: the role of social norms, FEDEA WP Nº 01/09.

References

1. Granovetter, M. (2005) The impact of social structure on economic outcomes, Journal of Economic Perspectives, 19(1): 33-50.

References

1. Halliday, T.J. and Kwak, S. (2009) Weight gain in adolescents and their peers, Economics and Human Biology, 7(2), 181-190.

References

1. Heckman, J.J., Stixrud, J. and Urzua, S. (2006) The effects of cognitive and noncognitive abilities on labor market outcomes and social behavior, Journal of Labor Economics, 24(3), 411-482.

References

1. Hanushek, E.A., Kain, J.F., Markman, J.M., and Rivkin, S.G. (2003) Does peer ability effect student achievement? Journal of Applied Econometrics, 18(5), 527-544.

References

1. Hoxby, C.M. (2000) Peer Effects in the Classroom: Learning from Gender and Race Variation, NBER Working Papers 7867, National Bureau of Economic Research, Inc.

References

1. Irwing, C.E., Scott J.B. and Uhler Cart, C. (2002) America’s adolescents: where have we been, where are we going?, Journal of Adolescent Health 31(6): 91-121.

References

1. Kang, Ch. (2007) Classroom peer effects and academic achievement: quasirandomization evidence from South Korea. Journal of Urban Economics, 61, 458-495.

References

1. Koplan, J.P., Liverman, C. T., Kraak, V.I. (2005) Preventing Childhood Obesity: Health in the Balance. Institute of Medicine of the National Academies. The National Academic Press, Washington, DC.

References

1. Lavy, V. and Schlosser, A. (2007) Mechanisms and impacts of gender peer effects at school. NBER Working Papers 13292, National Bureau of Economic Research, Inc.

References

1. Lundborg, P. (2006) Having the wrong friends? Peer effects in adolescent substance use. Journal of Health Economics, 25: 214-233.

References

1. Manski, Ch. (1993) Identification of endogenous social effects: the reflection problem. Review of Economic Studies, 60(3), 531-542.

References

1. Norton, E.C., Lindrooth, R.C. and Ennett, S.T. (1998) Controlling for the endogeneity of peer substance use on adolescent alcohol and tobacco use, Health Economics 7(5): 439-453.

References

1. Oswald, A.J. and Powdthavee, N. (2007) Obesity, unhappiness and the challenge of affluence: theory and evidence, Economic Journal, 117, F441-454.

References

1. Proud, S. (2008) Girl power? An analysis of peer effects using exogenous changes in the gender make-up of the peer group. CMPO Working Paper Series No. 186, Centre for Market and Public Organisation, Bristol, England.

References

1. Reilly, J.J., Methven, E., McDowell, Z.C., Hacking, B., Alexander, D., Stewart, L. and Kelnar. C.J. H. (2003) Health consequences of obesity, Archives of Disease in Childhood, 88: 748-752.

References

1. Renna, F., Grafova, I.B. and Thakur, N. (2008) The effect of friends on adolescent body weight. Economics and Human Biology, 6, 377-387.

References

1. Sacerdote, B.I. (2001) Peer effects with random assignment. Quarterly Journal of Economics, 116, 681-704.

References

1. Trogdon, J.G., Nonnemaker, J. and Pais, J. (2008) Peer effects in adolescent overweight. Journal of Health Economics, 27, 1388-1399.

References

1. WHO (2004) Young people’s health in context. Health Behaviour in School-aged Children (HBSC) study: International report from the 2001/2002 survey. Edited by: C. Currie, C. Roberts, A. Morgan, R, Smith, W. Settertobulte, O. Samdal and V. Barnekow-Rasmussen, WHO Regional Office for Europe, Copenhagen, Denmark.

References

1. WHO (2007) The Challenge of Obesity in the WHO European Region and the Strategies for Response. WHO Regional Office for Europe, Copenhagen, Denmark.

References

1. Zimmerman, D.J. (2003) Peer effects in academic outcomes: evidence from a natural experiment. The Review of Economics and Statistics, 85(1), 9-23.

Figure 1 Distribution of friend’s nominations by BMI status

Table 1 Description of Control Variables

Note: standard deviations are reported into brackets

Table 2 Friend’s weight influence on Adolescent BMI (OLS Estimation)

Notes : Adj usted robust standard errors for clustering at the classroom level were computed and reported in brackets . *** ** and * denote statistical significance at 1 5 and 1 0% respectively. Regres sions include a dummy variable for outliers and a con stant term.

Table 3 Friend’s weight influence on Adolescent BMI (GMM Estimation)

Notes : Adjusted robust standard errors for clustering at the clas sroom level were computed and reported in brackets . Then *** ** and * denote statistical significance at 1 5 and 1 0% respectively. Regres sions include a dummy variable for outliers and a constant term .

Table 4 Alternative definitions of peers

Notes : Adjusted robust standard errors for clustering at the clas sroom level were computed and reported in brackets . Then *** ** and * denote statistical significance at 1 5 and 1 0% respectively. Regres sions include a dummy variable for outliers and a constant term .

Table 5 Friend’s weight influence on Adolescent BMI (GMM Estimation)

Notes : Adjusted robust standard errors for clustering at the clas sroom level were computed and reported in brackets Then *** ** and * denote statistical significance at 1 5 and 1 0% respectively Regres sions include a dummy variable for outliers and a constant term .

References

1. 2010-18: “Some Students are Bigger than Others, Some Students’ Peers are Bigger than Other Students’ Peers”, Toni Mora y Joan Gil.

References

1. 2010-17: “Electricity generation cost in isolated system: the complementarities of natural gas and renewables in the Canary Islands”, Gustavo A. Marrero y Francisco Javier Ramos-Real.

References

1. 2010-16: “Killing by lung cancer or by diabetes? The trade-off between smoking and obesity”, Federico A. Todeschini, José María Labeaga y Sergi Jiménez-Martín.

References

1. 2010-15: “Does gender matter for academic promotion? Evidence from a randomized natural experiment”, Natalia Zinovyeva y Manuel F. Bagues.

References

1. 2010-14: “Spain, Japan, and the Dangers of Early Fiscal Tightening”, R. Anton Braun y Javier Díaz-Giménez.

References

1. 2010-13: “Competition and horizontal integration in maritime freight transport”, Pedro Cantos-Sanchez, Rafael Moner-Colonques, Jose Sempere-Monerris y Óscar Alvarez.

References

1. 2010-12: “Eppur si Muove! Spain: Growing without a Model”, Michele Boldrin, J. Ignacio Conde-Ruiz y Javier Díaz Gimenez.

References

1. 2010-11: “The Spanish Business Bankruptcy Puzzle and the Crisis”, Marco Celentani, Miguel García-Posada y Fernando Gómez.

References

1. 2010-10: “Promoting Employment of Disabled Women in Spain; Evaluating a Policy”, Judit Vall Castello.

References

1. 2010-09: “The Role of Construction in the Housing Boom and Bust in Spain”, Carlos Garriga.

References

1. 2010-08: “Did Good Cajas Extend Bad Loans? Governance, Human Capital and Loan Portfolios”, Vicente Cuñat y Luis Garicano.

References

1. 2010-07: “Unemployment and Temporary Jobs in the Crisis: Comparing France and Spain”, Samuel Bentolila, Pierre Cahuc, Juan J. Dolado y Thomas Le Barbanchon.

References

1. 2010-06: “La Subida del Impuesto Sobre el Valor Añadido en España: Demasiado Cara y Demasiado Pronto”, Juan Carlos Conesa , Javier Díaz-Giménez, Julián Díaz-Saavedra y Josep Pijoan-Mas.

References

1. 2010-05: “Off-the-peak preferences over government size”, Francisco Martínez-Mora y M. Socorro Puy.

References

1. 2010:04: “Green Shoots? Where, when and how?”, Maximo Camacho, Gabriel Perez-Quiros y Pilar Poncela.

References

1. 2010-03: “The Spanish Crisis from a Global Perspective”, Jesús Fernández-Villaverde y Lee Ohanian.

References

1. 2010-02: “Fiscal Centralization and the Political Process”, Facundo Albornoz y Antonio Cabrales.

References

1. 2010-01: “The Evolution of Adult Height Across Spanish Regions 1950-1980: A New Source of Data”, Mariano Bosch, Carlos Bozzoli y Climent Quintana-Domeque.

References

1. 2009-40: “Demographic Change and Pension Reform in Spain: An Assessment in a Two-Earner, OLG Model”, Alfonso R. Sánchez Martín y Virginia Sánchez Marcos.

References

1. 2009-39: “The new growth model: How and with Whom?”, Florentino Felgueroso y Sergi Jiménez-Martín. 2009-38: “Rain and the Democratic Window of Opportunity”, Markus Brückner and Antonio Ciccone.

References

1. 2009-37: “International Commodity Prices, Growth, and the Outbreak of Civil War in Sub-Saharan Africa”, Markus Brückner and Antonio Ciccone.

References

1. 2009-36: “Determinants of Economic Growth: Will Data Tell?”, Antonio Ciccone y Marek Jarocinski.

References

1. 2009-35: “The Timing of Work and Work-Family Conflicts in Spain: Who Has a Split Work Schedule and Why?”, Catalina Amuedo-Dorantes y Sara de la Rica

References

1. 2009-34: “Fatter Attraction: Anthropometric and Socioeconomic Characteristics in the Marriage Market”, Pierre-André Chiappori, Sonia Oreffice y Climent Quintana-Domeque.

References

1. 2009-33: “Infant disease, economic conditions at birth and adult stature in Brazil”, Víctor Hugo de Oliveira Silva y Climent Quintana-Domeque.

References

1. 2009-32: “Are Drinkers Prone to Engage in Risky Sexual Behaviors?”, Ana I. Gil Lacruz, Marta Gil Lacruz y Juan Oliva Moreno.

References

1. 2009-31: “Factors Explaining Charges in European Airports: Competition, Market Size, Private Ownership and Regulation”, Germà Bel y Xavier Fageda.

References

1. 2009-30: “Are Women Pawns in the Political Game? Evidence from Elections to the Spanish Senate”, Berta Esteve-Volart y Manuel Bagues.

References

1. 2009-29: “An Integrated Approach to Simulate the Impacts of Carbon Emissions Trading Schemes”, Xavier Labandeira, Pedro Linares y Miguel Rodríguez.

References

1. 2009-28: “Disability, Capacity for Work and the Business Cycle: An International Perspective”, Hugo-Benítez-Silva, Richard Disney y Sergi Jiménez-Martín.

References

1. 2009-27: “Infant mortality, income and adult stature in Spain”, Mariano Bosch, Carlos Bozzoli y Climent Quintana-Domeque.

References

1. 2009-26: “Immigration and Social Security in Spain”, Clara I. Gonzalez, J. IgnacioConde-Ruiz y Michele Boldrin.