Page 160 - GHES-3-1

Global Health Economics and
Sustainability
Affect heuristics in substance use

Table 2. Description of selected variables: NSHS, Brazil 2019
Pr ( y = 1, y = 1, y = ) 1
3 i
1 i
2 i
Variable Description Pr ò ≤ β = ' x ,ò ≤ β ' x ,ò ≤ β ' x )
Alcohol Have you ever consumed any alcoholic ( i 1 1 i 1 2 i 2 i 2 3 i 3 i 3
'
'
x
x
'
x ò
beverages? Pr ò 3 ≤ β = 3 i 3 |ò 2 i < β 2 i 2, 1 i < β 1 i 1 )
( i
Drugs Have you ever consumed any illicit drugs? Pr ò ≤ β × ' x |ò < β ' x ) Pr× ò < β ' x )
Cigarette Have you ever smoked a cigarette? ( i 2 3 i 2 1 i 1 i 1 ( i 1 1 i 1 (8)
Friend_alcohol Do any of your friends drink alcohol? The probability expressed in (8) involves conditioning
Friend_drugs Do any of your friends use illegal drugs? on unobserved variables that are correlated with each
Friend_smoke Do any of your friends smoke cigarettes? other. The GHK simulator performs an approximation
of these conditional distributions (Cappellari &
Parent_alcohol Does either of your parents or any guardian Jenkins, 2003).
drink alcohol?
Parent_smoke Does either of your parents or any guardian Börsch-Supan & Hajivassiliou (1993) stated that the
smoke cigarettes? desirable properties for simulated maximum likelihood
SLI Standard of Living Index (values closer to 1 methods are generating unbiased simulated choice
indicate a worse standard of living) probabilities that are bounded within the range of 0 and 1
RBI Risk Behavior Index (values closer to 1 and that represent continuous and differentiable functions
denote risker behavior) of the model parameters. According to Cappellari &
Work After high school, you intend only to work Jenkins (2003), the simulation bias is reduced to tiny levels
(base variable: no plans) in the GHK simulator as the number of random drawings
Study After high school, you intend only to study increases with the sample size. They recommended that the
(base variable: no plans) number of random drawings in sufficiently large samples
Study_work After high school, you intend to study and should approximate the closest integer to the square root
work (base variable: no plans) of the number of respondents in the sample. In the present
Another_plan After high school, you have another plan study, the analyzed sample comprised 55,883 respondents.
(base variable: no plans) Therefore, the number of 243 random draws was defined
Male Male sex for the model estimation.
Age Age 3. Results and discussion
School_mom Mother’s education level
n_people Number of people living in the household First, a descriptive analysis was conducted on the data
White White ethnicity on substance use among high school students in Brazil.
Figure 1 presents the relative frequency of use for each
Priv Attending private school substance (alcohol, illicit drugs, and cigarettes).
Urb Households located in an urban area
Source: Author’s elaboration based on NSHS 2019 data. It was observed that alcoholic beverages were the most
commonly used substance among adolescents, probably
Where K = 2y −1, for each i, m = 1,2,3, The matrix Ω because of their greater social acceptance and easier access
im
im
has the following elements: notwithstanding the prohibition of alcohol sales to minors
in Brazil. A significant proportion of adolescents also used
Ωmm = 1,m = 1,2,3 (7) cigarettes and illicit drugs even though this percentage was
Ω = Ω = K K ρ 21 much lower than the proportion of alcohol users, which
i1
12
i2
21
Ω = Ω = K K ρ probably reflected greater difficulty of access and lower social
31 13 i3 i1 31 acceptance, particularly in the case of illicit drugs. Figure 2
Ω = Ω = K K ρ 32 illustrates the relative frequency of substance use by sex.
23
32
i3
i3
According to Cappellari & Jenkins (2003), the Notably, the examination of the relative frequency of
GHK simulator takes advantage of the fact that the substance use by gender revealed that alcohol consumption
multivariate normal distribution can be defined as the was higher among female adolescents, while the use of
product of sequentially conditioned univariate normal cigarettes and illicit drugs was more prevalent among male
distributions. Eight joint probabilities corresponding adolescents. Figure 3 displays the relative frequency of
to eight combinations of y = 1 e y = 0 exist for a substance use by age.
im
im
trivariate case. Given the probability of all outcomes The evaluation of substance use trends by age disclosed
being y = 1, that the youngest high school students (11 and 12 years
im
Volume 3 Issue 1 (2025) 152 https://doi.org/10.36922/ghes.3829