1. Introduction

Adolescence is the period of human life from age 10 to 19 where the transition from childhood to adulthood [1] occurs and where alcohol and drug use is known to begin [2]. This period of life is marked by physical, psychological, and cognitive changes and the search for identity and autonomy [3].

Alcohol consumption during adolescence is associated with several harmful health and social consequences. Adolescent drinkers tend to get involved in dangerous situations such as traffic accidents and fights. They are prone to misconduct, hyperactivity, and drug use, and often get into problems with parents, schools, and the police [4, 5]. For girls, alcohol abuse is associated with falling prey to sexual crimes [6]. Alcohol use, for both boys and girls, is associated with risky sexual behaviors, such as having multiple partners and having unprotected sex [7–9]. Another important aspect is the relationship between alcohol consumption and mental health. Some authors, such as Mentzakis et al. [10], bring attention to a large amount of evidence that links alcohol abuse with mental disorders such as depression, anxiety, phobias, and personality disorders. Marschall-Lévesque et al. [11] also indicate the relationship between alcohol consumption and adolescent suicides. Cheng et al. [12] and Weitzman et al. [13] explain that since adolescent drinkers begin drinking early in life, their alcohol consumption period is longer, which leads to increases the chance of developing dependence in the future. In Brazil, alcohol consumption is becoming a major health risk to monitor due to a variety of reasons. Rehm et al. [14] analyzes the 2004 Global Burden of Disease and Injury and indicate the existence of an elevated loss of disability-adjusted life-years. According to WHO [15], the male Brazilian population is expected to lose 17.7% disability-adjusted life-years due to alcohol consumption. The World Health Organization 2018 Global Status Report on Alcohol and Health projects an 8.3% increase in Brazil’s alcohol consumption per capita until 2025.

Thus, the objective of this study is to verify factors associated with different intensities of alcohol intake by students of the ninth-grade of elementary school during the year 2015, discriminated according to student gender. The econometric model used is a continuation ratio logit.

We chose to estimate a continuation ratio logit based on Fone et al. [46] study of the effects of neighbourhood deprivation and excessive alcohol consumption and binge drinking in Wales (UK). Fone et al. [46] indicate the absence of knowledge in order to distinguish between excess consumption and binge drinking as distinct categories and how those categories may be influenced by neighbourhood deprivation. The authors observe that residents of less deprived neighbourhoods have a greater chance of excessive alcohol consumption. Despite that, Fone et al. [46] also observe that the residents of less deprived neighbourhood also present a lesser risk of binge drinking compared to the observed effect on excessive alcohol consumption. Based on Fone et al. [46] work and their indication of the absence of literature concerned with distinguishing different levels of alcohol intake, this study also aim to better comprehend whether the associations of the studied factors may vary within different population groups of alcohol intake.

The dataset from National Adolescent School-based Health Survey (PeNSE) was used, specifically that of ninth-grade elementary school students of public and private schools. It is noteworthy that in Brazil, selling alcoholic beverages to individuals under 18 years old is forbidden under law 9,294, of July 15, 1996. However, this issue becomes controversial when advertising alcoholic beverages in the media is allowed. Alcoholic beverages are legal drugs, accepted by all social levels, and are easily accessible. [16, 17]. Data from the Brazilian Institute of Geography and Statistics (IBGE) reveal that approximately 34% students enrolled in Brazilian private and public schools had already experimented with alcoholic beverages sometime in their lifetime. In 2015, these proportions increased to approximately 55% of the population in that age range [18]. The highest percentages of alcohol experimentation were observed in Rio Grande do Sul (74.2%), Santa Catarina (73.8%), Paraná (70.2%), followed by São Paulo (69.2%) and Mato Grosso do Sul (68.7%). The lowest percentages were observed in Pará (50.6%), Maranhão (50.8%), and Ceará (53.8%) (IBGE, 2021).

These data about Brazil’s alcohol consumption have important implications for the financing of the national public health system. Coutinho et al. [19] indicate that approximately 8 million dollars (PPP 2010) are spent solely on treating diseases related to alcohol consumption, a sum that is possibly greater considering its role in the spread of other diseases.

This paper contributes to the understanding of alcohol consumption among school adolescents in Brazil. Here, we analyze whether there is a difference in drinking behavior in relation to gender and age groups. Variables such as professional activity, tobacco and drug use, number of friends who drink, duration of physical activity, economic condition, family composition, and parental supervision provide information for the formulation of public health policies for this population. The consumption of alcoholic beverages is a risk factor for the population in general, as it is considered one of the main causes of non-transmissible chronic diseases, accidents, and violence.

The adolescent situation is equally concerning. Although the academic literature presents several studies describing adolescent binge drinking in Brazil (examples include [20–26]), to the best of our knowledge, this is the first work that investigates this subject integrating different types of drinking in a unique empirical framework and using a nationally representative sample. By doing so, the estimated models allow us to analyze different socioeconomic characteristics associated with specific types of drinking and even the different weighting a characteristic may have on distinguishing students who binge drink from moderate drinkers or nondrinkers.

Literature suggests that gender may play an important part in determining how some known characteristics that influence alcohol consumption operate. Nolen-Hoeksema [27] indicates that women drink less alcohol and have fewer alcohol-related problems than men. Hence, women may be less likely to develop alcohol-related problems because they are less likely to have multiple risk factors for these behaviors. Men, unlike women, tend to have characteristics associated with excessive alcohol consumption, including aggressiveness, drinking to reduce distress, lack of behavioral control, sensation-seeking, and antisociality, among others. On the other hand, women are more likely to experience physical harm and sexual assault when drinking alcohol than men, not to mention that heavy alcohol consumption in women is associated with a range of reproductive problems. Moreover, Li et al. [28] and Cohen et al. [29] suggest that females’ own perception of being more closely monitored than males regarding alcohol drinking behavior may contribute to inhibit addictive substance consumption. We address this gap through estimation and comparison of gender specific models in addition to a more general model composed of both males and females students.

Therefore, the analysis presented in this study may contribute to better defining key aspects for formulating public policies aimed at inhibiting alcoholism. Examples include defining the target population among adolescents for preventive measures against alcohol consumption and identifying more effective channels to reach a specific population group that may have a pattern of alcohol consumption that is extremely harmful to health.

Relevant data, such as the evolution of the total number of enrollments in primary schools and the differences observed in the states of the federation can help to better understand the context of Brazilian education. According to the School Census, in 2015, there were 48.8 million enrollments in the 186.4 thousand basic education schools in Brazil, including public and private institutions. The municipal public school’s system holds 46.8% of the basic education enrollments. In the states of Pará, Maranhão, Piauí, Alagoas, and Bahia, the municipal system has a participation in basic education enrollments of more than 60%. In the Southeast region, the greatest participation of municipal system is observed in Espírito Santo (55.8%) and the smallest in São Paulo (37.8%). Although the state system holds 16.5% of the schools, its participation in enrollments in basic education is 33.9%. In the Midwest region, considering the number of enrollments, the state system exceeds the other systems, with a participation of 44%. The private system corresponds to 18.6% in the total number of enrollments in basic education, while the federal system corresponds to less than 1%. Even though 34.7% of Brazilian schools are located in rural areas, they hold only 11.7% of the enrollments (5.7 million). It is noteworthy that the differences in urbanization between the regions of the country are reflected in the participation of the rural zone in enrollments. For example, while in Maranhão 35% of enrollments occur in rural areas, in São Paulo the percentage is less than 2% [30].

There were 12.4 million enrollments in the final years of elementary school, 14.5% of which were in private schools. The private system grew 14% in seven years. The state education system had 5.4 million enrollments, with a participation of 43.6% in enrollments, dividing the responsibility of providing public education with the municipalities, which had 5.2 million enrollments in the final years of elementary school (41.7%) in 2015. The states of São Paulo and Mato Grosso have the lowest rates of age/grade distortion, with 89% and 86% of the municipalities with rates lower than 20%, respectively [30].

This study is organized in six sections following this introduction. In section two, the literature review is presented. In three, the data and methodology are exposed. Sections four and five of this paper report and discuss the estimations results. The last section presents a brief overview and suggestions for future research.

2. Literature review

Given the relevance of studying alcohol consumption among adolescents, this section discusses the main factors that lead to alcohol consumption. Among Spanish adolescents, alcohol consumption is mainly due to the influence of friends, smoking, positive expectations from the use, and being male [31]. In assessing the effects of alcohol consumption on popularity and improved social interaction in the United States, Ali et al. [32] found that those who drink are more likely to be having friends and increments in their friend network. For Kumar et al. [33], higher income level and living in urban areas were positively associated with increased alcohol consumption.

However, these characteristics may influence alcohol consumption differently depending on the intensity of the consumption [31]. The predictive variables of individuals who consume alcohol sporadically differ from those who consume it frequently and have previous problems and risks with alcohol consumption.

Rehm et al. [34] point out to the increasing scientific interest in reporting and analyzing different drinking patterns such as volume, frequency, and quantity of alcohol consumption. Berggren and Sutton [35] suggest that there is a clear rationale for disaggregating these behaviors instead of analyzing total volume of alcohol intake as there exists considerable evidence relating different alcohol consumption patterns to specific problematic outcomes. Rehm et al. [36] present a systematic literature review on patterns of drinking and the burden of disease, for example, and indicate the existence of a clear linkage between the occurrence of more damaging consequences of alcohol consumption, from injuries and accidents to heavy drinking occasions with intoxication.

One important drinking pattern that specially affects adolescent health is binge drinking (i.e., the consumption of five or more alcoholic drinks for males and four or more alcoholic drinks for females on the same occasion). Courtney and Polich [37] highlight the lack of definitional precision regarding the term “binge drinking”. In this work, we adopt the US National Institute on Alcohol Abuse and Alcoholism’s definition of binge drinking as the consumption of five or more alcoholic drinks for males and four or more alcoholic drinks for females on the same occasion on at least one day in the past month [38]. Binge drinking is a specially concerning health risk factor for adolescent health for several reasons. Miller et al. [5] highlight that alcohol consumption is a major risk factor associated with leading causes of death among young people, such as unintentional injury, suicide, and homicide. Hingson and Zha [39] indicate that many adolescent health risk behaviors (like illicit drug use, tobacco-based product consumption, and violent behavior) are more frequently observed in adolescents who binge drink. Chung et al. [40] and Kuntsche and Gmel [41] indicate that greater chances of binge drinking are usually observed in this period of life. Moreover, La Fauci et al. [42] indicate that binge drinking among adolescents has gone through a sharp upsurge in recent times.

Given the importance of the subject, some studies have dedicated themselves to analyzing the characteristics that distinguish binge drinkers from other types of drinkers. Examples include Hill et al. [43] and Tucker et al. [44], who investigate the developmental course of alcohol use from adolescence to early adulthood; Lac and Donaldson [45], who carry out a discriminant analysis to categorize drinker types between non-drinkers, moderate, and binge drinkers; and Fone et al. [46], who analyze the socioeconomic patterning of excess alcohol consumption and binge drinking in the UK.

Commenting on studies that used discriminant analysis, Lac and Donaldson [45] indicate that previous efforts to categorize drinker types have been limited in several ways. These limitations include the exclusion of nondrinkers as an investigated category and the absence of models that evaluate in a unique way the different responses between males and females.

Based on the work of Harris and Zhao [47] about zero-inflated outcomes on ordered models, the exclusion of nondrinkers in an empirical framework implies the exclusion of two distinct types of drinkers. The first type includes those individuals who abstain from alcohol consumption due to a disinterest in alcohol. The second type includes individuals who want alcoholic beverages but did not consume those goods during the study period for external reasons, like elevated prices or difficult access. Excluding these two groups of drinkers from empirical analyses may potentially bias the estimation results due to missing data in the selection process. We address this limitation through the analysis of an ordered scale of drinking patterns composed of four different alcoholic consumption patterns: students who never consumed alcohol in their lifetimes, who have not consumed alcohol for a brief period of time, who consumed a moderate amount of alcohol, and binge drinkers.

3. Methods

3.4 Methodology

We used a partially constrained continuation ratio logit model [51] to analyze the socioeconomic determinants of students’ alcohol consumption. It is also known as sequential response model [52].

In this paper, we have chosen the continuation ratio forward model which compares a certain category with higher categories, and thus estimates the chances of a student being in a certain category versus being above that category. For example, the ordinal response variable (never drank in life, no alcohol consumption in the reference period, moderate drinkers, binge drinkers) has four categories, the continuation ratio forward model compares category 1 with categories 2, 3, and 4; category 2 with categories 3 and 4; and category 3 with category 4 [53]. In this sense, a subject must pass through each antecedent category before reaching a latter outcome. Examples include level of educational attainment [54] and self-rated health [55]. Continuation ratio model was already previously used to study alcohol consumption by Fone et al. [46] who analyzed associations between level of neighborhood deprivation and daily alcohol consumption across the UK Department of Health guideline categories of alcohol consumption. Considering that the forward continuation ratio model estimates the chances of a student being in a certain category j versus being above that category, here is how it can be expressed: where is the conditional probability of being in the category j (never drank in my life, no alcohol consumption in the reference period, moderate drinkers, binge drinkers) conditional on being in or above that category given a set of predictors, i.e., are the cut-off points, and β₁,β₂,…,β_p are the logit coefficients. The continuation ratio model assumes that the logit coefficients for each predictor are the same across all ordinal categories, so this model is also called the restricted continuation ratio model [53].

In this study, we choose to adopt a scale of increasingly undesirable outcomes. In doing so, we consider the absence of alcohol consumption over a student’s lifetime as the lowest (most preferable) outcome of this scale, followed by not consuming alcohol over the reference period as the second most desirable outcome, and the group of moderate drinkers as the third most desirable option. At the end of the scale lies binge drinking, which assumes the position of least preferable outcome of our ordered scale. Three models were estimated: one for each sex (male and female) and a complete model composed of both boys and girls using sex as a covariate. All estimates were made considering the sample weight assigned to each student as a result of the stratified complex sampling process of the research.

Following Cole and Ananth [51], we test for the existence of proportionality of odds. This hypothesis refers to the existence of equivalent (proportional) effect of a known covariate in the transition between different categories that makes up an ordered scale. Under this hypothesis, a covariate is expected to present the same statistical effect on each transition of state (for example, from never-drinkers to non-drinkers, moderate drinkers, and binge drinkers, and so on) in the estimated models. We use Brant [56] test to identify the presence of proportionality of odds. A 5% significance on the test statistic was used to determine the presence of symmetric effects for each of the variables described in Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Description of the sample by alcohol consumption categories.

https://doi.org/10.1371/journal.pone.0281065.t001

The results of the estimated Brant [56] tests are presented in the (S3 Table) and suggests that, in general model, a students sex (being a female), being between 14 and 17 years old, being asian or multiracial, emotional state, attending to a public school, living with both parents, number of people who live in the same household, time spent on physical activity, tobacco and illicit drugs consumption and all variables describing the number of friends who drink alcoholic beverages and geographic location have violated the proportionality of chances assumption. An equality restriction was applied to the maximum likelihood process for the remaining variables who did not violated the null hypothesis of proportionality of odds in the estimated model. The same process of diagnostic and estimation was applied to gender segmented models.

4. Results

4.2 Continuation ratio models

Tables 2 and 3 present the results of the estimated continuation ratio models. Results of both tables are displayed in odds ratios so that coefficients above unity indicate a higher chance of progressing to higher levels of alcohol consumption and coefficients below unity indicate a decreasing chance of moving on to higher categories (higher chance of a participant being in the current category). Grouped results indicate variables that presented proportionality of chances according to Brant [56] test. These are the cases of students having a professional occupation besides studying and living with a single parent (Table 2) and studying in a public school or performing unpaid activity for both models of Table 3.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Results of the sequential logit model (in odds ratios).

https://doi.org/10.1371/journal.pone.0281065.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Results of the sequential logit model (in odds ratios).

https://doi.org/10.1371/journal.pone.0281065.t003

Table 2 displays the continuation ratio model estimation results of a combined sample of male and female students. Regarding demographic trends, results indicate that female students were less likely to begin drinking. In cases where the drinking had already begun, female students also presented an elevated chance to declare they had consumed alcohol in the reference period. It should be noted that these chances were increasing in a sense that girls were more likely to binge drink than boys.

Older students presented an elevated chance of consuming alcoholic beverages. The risk of consuming alcoholic beverages varied in students aged fourteen to seventeen across different categories and it was constant in students aged eighteen and over. The consistency of the odds ratio observed in the last age group may relate to the fact that 18 years is the threshold legal age for a person to be allowed to buy alcoholic beverages in Brazil.

Students who were engaged in a professional activity displayed a higher chance of beginning to drink and consuming more alcohol. It should be noted that being engaged in a paid occupation indicated a higher chance of consuming alcohol than not being paid to work.

The consumption of tobacco-based products and illicit drugs and the number of friends who also consumed alcohol were important predictors of student’s alcohol consumption. Table 2 indicates that consumption of tobacco-based products and illicit drugs presented a decreasing positive effect. This result suggests that consumption of drugs is more common when the initiation into alcohol consumption has occurred since greater effect is observed in the first stage.

A similar pattern was observed regarding the number of friends who also consumed alcohol. Results suggest that an increasing number of friends who also consume alcohol is associated with an increasing chance of a student consuming alcoholic beverages. It is worth noting that students who had a few or some friends who consumed alcohol were not vulnerable to becoming binge drinkers, although it increased the chance of being initiated into alcohol and its consumption over the reference period. In that regard, all or most friends being alcohol drinkers was associated with higher chance of a student being a binge drinker.

The results of estimations divided by sex (male and female) are presented in Table 3. These models suggest the existence of some differences related to the effect a covariate may play on a student’s alcohol consumption. A more visible example of these differences is the contrast between the amount of time boys and girls spend on physical activities and between living with just one parent.

The results in Table 3 indicate that amount of time spent on physical activities increased male students’ chance of consuming alcohol. The presented results also indicate that time spent on physical activities did not influence female student’s alcohol consumption. A similar pattern of results was observed for students living with only one parent in the sense that it decreased male students’ probability of consuming alcohol and did not influence female students’ chance of consuming alcohol.

Apart from Indigenous male students, non-Caucasian ethnicities were associated with a decreased chance of consuming alcohol in the reference period. Other than that, being a non-Caucasian male student did not influence other aspects of alcohol consumption behavior, such as the probability of initiation or binge drinking. Black female students presented a positive chance of progressing along the alcohol outcome measure. This result indicates that black female students exhibit a greater chance of initiating alcohol consumption, consuming alcohol during the reference period, and being binge drinkers altogether.

Strong predictors of alcohol consumption include number of friends who consume alcohol and consumption of tobacco-based products and illicit drugs, and their importance in determining alcohol consumption persisted in both male and female individually estimated models.

5. Discussion

The present study found that different levels of consumption are associated with different characteristics in Brazilian students in terms of magnitude and direction, which may vary between boys and girls. As this association had not been studied in Brazil, this is an important finding since it reveals that implementing strategies focused on preventing a particular behavior, like adolescent casual day-to-day drinking, may have a positive side effect on other related habits related to alcohol consumption such as adolescent binge drinking or alcohol initiation. Following on that, the very knowledge that the characteristics associated with these different alcohol consumption profiles do not differ may be a gain by itself. Knowing that these characteristics remain the same could possibly simplify the process of public policy formulation for alcohol consumption prevention agenda considering that implementing an action focusing on a certain alcohol consumption profile will not present negative side effects for others.

Specifically, based on the sole analysis of the methodology used to analyze the alcohol consumption profiles studied in this paper, one could expect the existence of three patterns or groups of results concerning the influence of the studied characteristics on the classification of the analyzed students in the groups of never-drinkers, non-drinkers, moderate drinkers, and binge drinkers.

The first group corresponds to some characteristic having a tenacious effect on possibly influencing students to present an unhealthier pattern alcohol consumption. Variables that exemplify this logic are household economic status and the consumption of tobacco or illicit drugs, which remained significant across all estimations and only presented a positive effect in the sense of elevating the chances of a student presenting more unhealthy level of alcohol consumption. The second group is represented by the case of characteristics of parental supervision. The model that combines both male and female students indicated that more attentive parents reduced the chances of a student being a binge drinker and experimenting with alcohol but did not interfere with the chances of the transition from nondrinker to moderate or binge drinker. The third pattern of results would be the case of a variable having a conflicting effect on the alcohol consumption progression scale in a way that it could influence the student being in an unhealthier category (such as moderate drinker) and discourage the student from being in an even more unhealthy category (such as binge drinker). Across all estimations, only two examples of this pattern were observed. These cases occur in the model that combines both male and female students, relating to the context of a student being a female or living in the southern region of Brazil.

Of course, some exceptions are acknowledged. As pointed out before, our results indicate an apparently conflicting behavior in relation to the student’s sex. The model estimated with boys and girls combined suggests that female students present a lower chance of experimenting with alcoholic beverages but a higher chance of exhibiting unhealthier profiles of alcohol consumption, such as moderate or binge drinking.

This result may be related to behavioral traits and social stigma linked to womanhood. Nolen-Hoeksema [27] indicates that women are less likely to show personality traits linked to alcohol consumption, such aggressiveness and antisociality. Nolen-Hoeksema [27] also indicates that women are more prone to adverse consequences linked to alcohol consumption, like requiring less exposition to develop alcohol-induced diseases and greater decline of cognitive and motor skills.

Likewise, Li et al. [28] state that girls usually report feeling more monitored by parents than boys and that the perceived parental monitoring tends to decrease with age. McArdle et al. [57] and Guo et al. [58] point out that an affective bond between parents and children contributes to limiting alcohol consumption. The combination of this evidence may explain the conflicting result, considering that a negative effect was observed on the transition that reflects alcohol initiation, which may take place at younger ages, and that being a girl was associated with a higher chance of moderate and binge drinking, which may be more commonly observed in more advanced ages of adolescence.

Another important result related to a possible gender bias was observed on the effect of time dedicated to physical activity. The model estimated only with female students indicated that physical activities did not influence girls’ alcohol consumption whereas the boys’ model suggests that more time performing physical activities might increase the odds of unhealthy alcohol consumption. Kwan et al. [59] systematically reviewed longitudinal studies analyzing sport participation and alcohol and illicit drug use in adolescents, and found that practicing sports is positively associated with alcohol consumption. Hoffman [60] suggests that being involved in sports may influence alcohol consumption due to socialization and cultural behaviors connected to sport activities. Wichstrøm and Wichstrøm [61] point out the existence of differences according to the type of sport. According to Wichstrøm and Wichstrøm [61], while endurance and individual sports reduce consumption, technical and team sports have been shown to increase the proportion of substance use in adolescence. Unfortunately, PeNSE 2015 does not consider the type of physical activity a student performs. Considering the importance that this detail may play in influencing adolescent alcohol consumption in Brazil, further research should focus on investigating the influence of different modalities of physical activity on teenage alcohol consumption.

Furthermore, looking at the results that relate to both masculine and feminine students, an important result concerns the influence of characteristics that may facilitate a student’s access to alcoholic beverages, in particular, the household economic status and whether a student takes up a paid occupation.

Regarding household economic status, Martins-Oliveira et al. [62] and Hanson and Chen [63] indicate the absence of consensus in the literature. Luthar and Becker [64] and Luthar and Latendresse [65] indicate that adolescents whose families possess greater economic conditions may be at greater risk of participating in risky health behaviors due to a combination of greater availability of disposable income, higher achievement pressures, and less adult supervision.

Our results appear to be in sync with those of Luthar and Becker [64] and Luthar and Latendresse [65] since they show that living in a household with greater economic conditions was positively associated with unhealthier profiles of alcohol consumption. A possible explanation for that result might be linked to Kumar et al. [33] remark that the combination of greater economic conditions and a more permissive environment may be what encourages adolescents with higher economic conditions to drink alcoholic beverages. Ibitoye et al. [66] studied the concentration of outdoor alcohol advertising and liquor stores near places where teenagers get together and posit that the absence of underage drinking laws enforcement might encourage teenage alcohol consumption.

Another result that deserves attention is the variable of employment in a paid occupation, like a job or internship. Our results show that students in the paid employment category presented a higher chance of being in unhealthier alcohol consumption categories. According to Kuntsche et al. [67], once students have greater financial resources, complementary to the socioeconomic status analysis, and in an environment characterized by greater interaction with older persons, their alcohol consumption increases. The very fact of drinking also affects the labor market. A review of some studies relative to the adult labor market outcomes and alcohol outcomes suggests the absence of a consensus between wages, employment rates, and alcohol consumption. Dave and Kaestner [68] results on US adults suggest that alcohol consumption does not negatively affect labor market outcomes. Böckerman et al. [69] twin data longitudinal analysis of Finnish men and women indicates that former drinkers and heavy drinkers have approximately 20% lower earnings when compared to moderate drinkers. Teckin [70] longitudinal analysis of the Russian population indicates a small linear effect on wage rates. Further research focused on adolescent alcohol consumption and labor market outcomes might provide new important evidence regarding the relation between them. In this context, data from the 2019 PNS showed that alcohol consumption tends to change according to per capita income, where the higher the income, the higher the proportion of people who consume alcoholic beverages. The rate of people who consume alcoholic beverages is 18.6% among those with no income, 27.8% among those earning between 1 to 2 salaries, and 49% among those earning more than 5 salaries [18].

Lastly, we recommend some caution when interpreting the results concerning the variables related to the number of friends who also consume alcohol, the consumption of tobacco-based products and illicit drugs, and mental health conditions. Although the first two groups of variables presented elevated odds ratio, meaning that consumption of other drugs and peer effects were highly associated with unhealthier alcohol consumption categories, previous studies have already indicated that these characteristics might be correlated with other unobserved aspects of a student’s life.

Many studies have shown the existence of an association between consumption of alcohol, tobacco, and other drugs (examples include [71–73]). Kandel [74] and Kandel et al. [75] gateway drug theory suggest the existence of an ordered sequence of stages in which the consumption of soft drugs (alcohol and tobacco) leads to consumption of harder drugs. Comparing the results regarding the relation between alcohol and tobacco and alcohol and illicit drugs, the greater odd ratio associated with the consumption of tobacco-based products could be interpreted as a possible outcome of a gateway effect. Nonetheless, French and Popovici [76] indicate that the consumption of addictive substances may be linked to unobserved characteristics, like preferences and attitudes towards risk and drugs. In this regard, it is important to emphasize that our results also suggest that alcohol prevention strategies may present a positive side effect in reducing the consumption of other drugs.

Regarding peer effects, McCann et al. [77] suggest that adolescents tend to imitate their peers’ level of alcohol consumption. Ali et al. [32] indicate that this behavior may be related to social motivations such as wanting to be popular or to be accepted in a new group of friends. Nesi et al. [78] state that exposure to social media, where users often tend to compare their own lives to others, may predict adolescent alcohol consumption. In that regard, Fletcher [79] notes that it is difficult to distinguish the group effect on an individual from the effect an individual may have on a group. Fletcher [79] indicates that the simultaneity arising from trying to determine who influences who makes the analysis of peer effects a not-so-straightforward task. Considering that adolescent drinking is an ongoing problem, this result suggests that public policies aimed at adolescent alcohol consumption prevention should have better effects when focused on a group level.

6. Conclusions

In this paper, we analyzed how various socioeconomic determinants might influence the alcohol consumption profile of ninth grade students in Brazil. Our analysis draws upon the literature focused on describing adolescent alcohol consumption in Brazil to investigate the existence of possible differences related to adolescent drinking based on the number of doses an adolescent consumes on the occasions when he or she consumes alcohol. Our results show that for a major part, the characteristics associated with different drinking profiles remain in the same direction. An exception is found in the context of gender. We found that female students present a lesser chance of experimenting with alcohol, however, they also present a higher chance of falling into an unhealthier pattern of drinking. Based on these results, our study’s main contribution to the literature and to the formulation of public policies is the suggestion to adopt a “one size fits all” approach because the different groups of alcohol consumption studied presented a similar profile. The very knowledge that adolescents first experimenting with alcoholic beverages and binge drinking adolescents happen to have remarkably similar characteristics raises the possibility of simplifying public policies aimed at this problem.

For future research, we suggest two advancements. First, the analysis of longitudinal data composed of both adolescents enrolled and not enrolled in school. The analysis of such data might show important differences on the determinants of adolescent alcohol consumption, especially considering that being enrolled in the education system may be a protective factor in limiting adolescent alcohol initiation and binge drinking. Second, from an analytical point of view, a further integrated analysis combining more elements related to alcohol consumption could be carried out. This might include aspects of intensity (how much alcohol an individual takes with each dose) and other consumption habits (like how many times a month one drinks alcohol and in which environments an adolescent consumes alcohol) that may further contribute to the identification of unhealthier habits of alcohol consumption and add more layers of precision for the formulation of public policies aimed at preventing underage drinking (in all of its possible forms) and reducing the negative side effects of continued alcohol consumption and abuse.

Supporting information