[Practice] Connection between escapism and risk behaviour
Sharin Nikita
30 March 2022
Project base info
Topic: connection between escapism and risk behavior among adolescents of Saint-Petersburg.
Research question: are escapism and risk behavior among adolescents of Saint-Petersburg associated with each other? If so, how exactly?
Literature review
The first thing to note: connection between escapism and risk behavior was already questioned in research of Jouhki (2021). In result, an association was found between escapism and the main types of risk behavior – smoking, alcohol and drug use (at the same time, gambling turned out to be an exception). It’s important to note that the sample of research was limited to the population of one country – Finland.
Jouhki (2021) have not studied whether there are any mediators in connection between escapism and risk behavior. Although the study of these mediators seems important: it can expose potential impact points, knowing about which will allow to produce more thoughtful approaches to reduce/prevent risk behavior – so I will do exactly it. There are two candidates to be these mediators: peer support and life satisfaction. Let’s consider each of them in detail.
As for peer support, Hakimzadeh et al. (2016) on a sample of Iranian students found 2 next associations:
- Between peer support and life satisfaction;
- Between peer support and student engagement in academic activities.
The second potentially can mean the presence of connection between peer support and escapism. Reason here is next: student engagement in academic activities implies active participation in school environment. This school environment, in its turn, provides the setting to escape from (= change academic activities on something else) for an escapist student. Therefore, there is reason to suppose that low engagement in academic activities is somehow connected with escapism.
Next thing is life satisfaction. Here picture is next: association between life satisfaction and most of risk behaviours has been proved many times. I will provide 2 examples, while there are many others. The first example is the association between life satisfaction and illicit drug use, tobacco, alcohol. This association has been found on the sample of American students by Topolski et al. (2001). According to the second example, for both Romanian and Italian youth alcohol consumption results in lower life satisfaction (Esposito et al., 2020).
Moving forward, concept of escapism is crucial for topic under consideration – so, let’s take a look at paper devoted completely to it. In their work Calleja (2010) on the difficult case of computer games discusses the nature of escapism. But the most important thing about this paper is the next: it also provides a comprehensive discussion of what acts can be considered as escapism and what cannot. That discussion, in its turn, allows to produce the conceptual definition of escapism that will be used later in this project.
Last thing to say: none of studies mentioned above included a sample of Saint-Petersburg students – so, this project will not only investigate escapism–risk behavior relationships and its mediators but also whether the already established connections are generalisable to the students of Saint-Petersburg.
Relationships to check
Based on literature review and project’s RQ next relationships are expected to be proved:
- Escapism – Peer support, negative association;
- Escapism – Risk behavior, positive association;
- Peer support – Life satisfaction, positive association;
- Life satisfaction – Risk behavior, negative association.
Concepts: definitions and operationalisations
Based on the literature review and the RQ posed, concepts for SEM analysis were identified:
- Person avoidance
- Person friendship
- Peer friendship
- Escapism
- Peer support
- Life satisfaction
- Risk behavior
Let’s consider each of these concepts in details:
Person avoidance
Conceptual definition: the number of people in relation to whom an individual experiences a regular desire to avoid (or actually avoids) any interaction or co-existence.
Operationalisation: out-degree centrality in avoidance net. Net is produced by asking students to indicate those classmates whom they avoid and with whom they try not to communicate.
Person friendship
Conceptual definition: the number of people in relation to whom an individual experiences a continuous state of mutual affection without romantic feelings.
Operationalisation: out-degree centrality in friendship net. Net is produced by asking students to indicate those classmates with whom they communicate the most.
Peer friendship
Conceptual definition: the number of people who experiences a continuous state of mutual affection without romantic feelings in relation to an individual.
Operationalisation: in-degree centrality in friendship net. Net is produced by asking students to indicate those classmates with whom they communicate the most.
Escapism
Conceptual definition (Calleja, 2010): is an act of departure from situation or environment to one that seems to be more desirable. It always implies return to the departure point. It is performed with the hope of improving the initial situation upon return, or temporarily alleviating the current burden.
Operationalisation: latent variable composed from out-degree centrality in avoidance net, out-degree centrality and in-degree centrality in friendship net.
Operationalisation reasoning (concept is quite complex, so it seems necessary here): in case of students the place to depart from can be presented by school environment (which is mostly presented by classmates of these students). Therefore escapism degree can be operationalised as a number of social nominations, the receiving of which implies expressed participation in school environment. Such nominations in case of our data are presented by the next ones: person avoidance (out-degree centrality in avoidance net), person friendship (out-degree centrality in friendship net) and peer friendship (in-degree centrality in friendship net).
Peer support
Conceptual definition (Ioannou et al., 2019; Hakimzadeh et al, 2016): the degree to which a person assesses his/her peers as a source of emotional and cognitive support in times when there is need in it.
Operationalisation: peer support scale made up from 3 questions. Questions are next: does a person have a friend with whom he/she can talk about his/her problems? does a person have a friend to whom he/she can turn in a difficult situation? all person’s friends do not care about him/her?
Life satisfaction
Conceptual definition (Veenhoven, 2009; Diener, 2009): individual perception of quality of his/her life as a whole.
Operationalisation: zest for life scale made up from 3 questions. Questions are next: is person usually felling tired? does a person usually have a lot of energy? can a person call himself cheerful?
Risk behaviour
Conceptual definition (Tariq et al., 2021): behaviour engaging in which increases the risks of disease or injury. Afterwards such behaviour can lead to social problems, disabilities and death.
Operationalisation: the regularity of alcohol consumption in 4 or more servings.
Preparation to analysis
Base preparation
What’s done:
- Libraries are connected;
- Dataset is loaded.
# Connecting necessary libraries
library(lavaan)
library(dplyr)
library(foreign)
# Importing the data
work_data = read.spss("students_data.sav", to.data.frame = T)Data variables preparation
What’s done:
- All necessary variables are formed;
- Network variables are protected from group size possible influence;
- Dataset is prepared for work.
# Creating life_satisfaction variable
work_data = work_data %>%
mutate(
zest_1 = case_when(
q220_3w_zest == "Согласен" ~ 4,
q220_3w_zest == "Скорее согласен" ~ 3,
q220_3w_zest == "Скорее не согласен" ~ 2,
q220_3w_zest == "Не согласен" ~ 1),
zest_2 = case_when(
q344_3w_zest == "Согласен" ~ 1,
q344_3w_zest == "Скорее согласен" ~ 2,
q344_3w_zest == "Скорее не согласен" ~ 3,
q344_3w_zest == "Не согласен" ~ 4),
zest_3 = case_when(
q349_3w_zest == "Согласен" ~ 4,
q349_3w_zest == "Скорее согласен" ~ 3,
q349_3w_zest == "Скорее не согласен" ~ 2,
q349_3w_zest == "Не согласен" ~ 1)) %>%
filter(!is.na(zest_1), !is.na(zest_2), !is.na(zest_3)) %>%
mutate(Life_satisfaction = (zest_1 + zest_2 + zest_3)/3)
# Creating life_satisfaction variable
work_data = work_data %>%
mutate(
peer_1 = case_when(
q373_3w_fr_support == "Согласен" ~ 4,
q373_3w_fr_support == "Скорее согласен" ~ 3,
q373_3w_fr_support == "Скорее не согласен" ~ 2,
q373_3w_fr_support == "Не согласен" ~ 1),
peer_2 = case_when(
q269_3w_fr_support == "Согласен" ~ 1,
q269_3w_fr_support == "Скорее согласен" ~ 2,
q269_3w_fr_support == "Скорее не согласен" ~ 3,
q269_3w_fr_support == "Не согласен" ~ 4),
peer_3 = case_when(
q340_3w_fr_support == "Согласен" ~ 1,
q340_3w_fr_support == "Скорее согласен" ~ 2,
q340_3w_fr_support == "Скорее не согласен" ~ 3,
q340_3w_fr_support == "Не согласен" ~ 4)) %>%
filter(!is.na(peer_1), !is.na(peer_2), !is.na(peer_3)) %>%
mutate(Peer_support = (peer_1 + peer_2 + peer_3)/3)
# Creating risk_alcohol_consumption variable
work_data = work_data %>%
mutate(
Risk_alcohol_consumption = case_when(
q257_3w == "Каждый день или почти каждый день" ~ 7,
q257_3w == "3-5 раз в неделю" ~ 6,
q257_3w == "1-2 раза в неделю" ~ 5,
q257_3w == "2-3 раза в месяц" ~ 4,
q257_3w == "1 раз в месяц или реже" ~ 3,
q257_3w == "1-2 раза за 6 месяцев" ~ 2,
q257_3w == "никогда или почти никогда" ~ 1))
# Getting rid of group size possible influence on net variables
work_data = work_data %>%
mutate(
Friends_outdegree = Fr_outdegree/(N_BREAK - 1),
Friends_indegree = Fr_indegree/(N_BREAK - 1),
Dislike_outdegree = Dislike_outdegree/(N_BREAK - 1),
Dislike_indegree = Dislike_indegree/(N_BREAK - 1))
# Selection only necessary variables for creating a SEM
prepared_data = work_data %>%
select(
Life_satisfaction,
Peer_support,
Risk_alcohol_consumption,
Friends_outdegree,
Friends_indegree,
Dislike_outdegree,
Dislike_indegree)
# Some mess occurred with variables descriptions – so, they were deleted
attr(prepared_data, "variable.labels") = rep(c(""), times = 7)Model
Method: SEM, as escapism is a concept requiring latent variable to be analysed.
Base of links choice: literature review and RQ of the project.
SEM scheme:
Note about links that is not covered by literature: residual variances correlations of all net variable are accounted in SEM as it is supposed that all net variables have something in common (due to their joint net nature) that is not captured by the escapism variable.
Note about escapism operationalisation: in a SEM the variable Escapism absence will be used instead of Escapism for better readability and interpreatation of the model. The reason: as the variables that make up escapism increase, Escapism itself is expected to decrease.
SEM summary:
model = '
# measurement model
Escapism_abscence =~ Friends_outdegree + Friends_indegree + Dislike_outdegree
# regressions
Escapism_abscence ~ Risk_alcohol_consumption
Escapism_abscence ~ Peer_support
Peer_support ~ Life_satisfaction
Life_satisfaction ~ Risk_alcohol_consumption
# residual correlations
Friends_outdegree ~~ Friends_indegree
Friends_indegree ~~ Dislike_outdegree
Dislike_outdegree ~~ Friends_outdegree'
# Missing = "ML" helps to conduct case-wise maximum likelihood estimation
fit = sem(model, data = prepared_data, missing = "ML")
summary(fit, standardized = T, fit.measures = T)lavaan 0.6-10 ended normally after 166 iterations
Estimator ML
Optimization method NLMINB
Number of model parameters 20
Used Total
Number of observations 345 346
Number of missing patterns 1
Model Test User Model:
Test statistic 16.364
Degrees of freedom 5
P-value (Chi-square) 0.006
Model Test Baseline Model:
Test statistic 152.427
Degrees of freedom 15
P-value 0.000
User Model versus Baseline Model:
Comparative Fit Index (CFI) 0.917
Tucker-Lewis Index (TLI) 0.752
Loglikelihood and Information Criteria:
Loglikelihood user model (H0) 471.547
Loglikelihood unrestricted model (H1) 479.729
Akaike (AIC) -903.095
Bayesian (BIC) -826.224
Sample-size adjusted Bayesian (BIC) -889.669
Root Mean Square Error of Approximation:
RMSEA 0.081
90 Percent confidence interval - lower 0.040
90 Percent confidence interval - upper 0.126
P-value RMSEA <= 0.05 0.100
Standardized Root Mean Square Residual:
SRMR 0.045
Parameter Estimates:
Standard errors Standard
Information Observed
Observed information based on Hessian
Latent Variables:
Estimate Std.Err z-value P(>|z|) Std.lv Std.all
Escapism_abscence =~
Friends_outdgr 1.000 0.126 1.420
Friends_indegr 0.621 0.701 0.885 0.376 0.078 0.897
Dislike_outdgr -0.967 0.281 -3.448 0.001 -0.122 -2.104
Regressions:
Estimate Std.Err z-value P(>|z|) Std.lv Std.all
Escapism_abscence ~
Rsk_lchl_cnsmp -0.003 0.002 -1.251 0.211 -0.020 -0.025
Peer_support 0.007 0.004 1.976 0.048 0.058 0.042
Peer_support ~
Life_satisfctn 0.309 0.046 6.691 0.000 0.309 0.339
Life_satisfaction ~
Rsk_lchl_cnsmp 0.023 0.035 0.653 0.514 0.023 0.035
Covariances:
Estimate Std.Err z-value P(>|z|) Std.lv Std.all
.Friends_outdegree ~~
.Friends_indegr -0.006 NA -0.006 -1.852
.Friends_indegree ~~
.Dislike_outdgr 0.010 NA 0.010 2.314
.Friends_outdegree ~~
.Dislike_outdgr 0.016 NA 0.016 1.685
Intercepts:
Estimate Std.Err z-value P(>|z|) Std.lv Std.all
.Friends_outdgr 0.131 0.014 9.386 0.000 0.131 1.483
.Friends_indegr 0.129 0.013 9.696 0.000 0.129 1.480
.Dislike_outdgr 0.061 0.013 4.818 0.000 0.061 1.056
.Peer_support 2.527 0.133 19.021 0.000 2.527 3.485
.Life_satisfctn 2.696 0.113 23.945 0.000 2.696 3.391
.Escapism_bscnc 0.000 0.000 0.000
Variances:
Estimate Std.Err z-value P(>|z|) Std.lv Std.all
.Friends_outdgr -0.008 NA -0.008 -1.016
.Friends_indegr 0.001 NA 0.001 0.196
.Dislike_outdgr -0.011 NA -0.011 -3.426
.Peer_support 0.465 0.035 13.134 0.000 0.465 0.885
.Life_satisfctn 0.632 0.048 13.134 0.000 0.632 0.999
.Escapism_bscnc 0.016 NA 0.998 0.998
Analysis
Note: estimates are standardized for better interpretation.
General measures:
- Degrees of freedom: 5, which is bigger than 0 – it means that model is over-identified and we have enough cases to explain all necessary links of the model, as well as to assess model fit;
- P-value of Chi-square: 0.006, which is lower than 0.5 – so, we reject hypothesis that model fits the data: the discrepancy between model-implied covariance matrix and observed covariance matrix is greater than expected because of sampling variability. At the same, according to Brown (2015), Chi-square test is oversensitive to sample size (which in our case includes ~350 observations, which is not a little) – so we can ignore this measure;
- RMSEA (absolute measure of fit): 0.081, which is according to Cline (2016) can be treated as reasonable approximate fit of the model;
- CFI (comparative measure of fit): 0.917, which is according to Cline (2016) can be treated as a reasonably good fit of the model.
Regressions:
Escapism_abscence ~ Risk_alcohol_consumption: negative (- sign), insignificant (P-value > 0.05), decrease in Escapism absence by 0.003 sd would be (if it was significant) associated with Risk alcohol consumption that is 1.0 sd higher;Escapism_abscence ~ Peer_support: positive (+ sign), significant (P-value < 0.05), increase in Escapism absence by 0.007 sd is associated with Peer support that is 1.0 sd higher;Peer_support ~ Life_satisfaction: positive (+ sign), insignificant (P-value > 0.05), increase in Peer support by 0.309 sd would be (if it was significant) associated with Life satisfaction that is 1.0 sd higher;Life_satisfaction ~ Risk_alcohol_consumption: positive (+ sign), significant (P-value < 0.05), increase in Life satisfaction by 0.023 sd is associated with Risk alcohol consumption that is 1.0 sd higher.
Escapism (latent variable):
Friends_outdegree: base of the made up latent variable – increase in Friends outdegree by 1.0 sd is associated with Escapism absence that is 1.0 sd higherFriends_indegree: positive (+ sign), insignificant (P-value > 0.05), increase in Friends indegree by 1.0 sd would be associated (if it was significant) with Escapism absence that is 0.621 sd higher;Dislike_outdegree: negative (- sign), significant (P-value < 0.05), increase in Dislike outdegree by 1.0 sd is associated with Escapism absence that is 0.967 sd lower.
*It is not really possible to interpret Covariances as measure that is necessary for it (P(>|z|)) is absent. Reasons why it could happen were not found in official Lavaan site (https://lavaan.ugent.be/) and other topic-related resources.
Conclusion
According to both absolute and comparative measures, fit of the model is more or less good.
4 relationships between variables were checked. Association between Escapism (absence) and Risk alcohol consumption, as well as association between Peer support and Life satisfaction were found statistically insignificant. At the same time other 2 associations were found statistically significant. In case of Escapism absence and Peer support association is positive, as it was expected. But in case of Life satisfaction and Risk behavior (in face of alcohol consumption in large doses) association is not the one that expected – it is positive, although the negative one was expected.
Such unexpected results probably can be explained by cultural features of St. Petersburg students, as well as by problematic operationalisations that do not correctly reflect the conceptual definitions (due to limited data available). The example of problematic operationalisation is Escapism, which is actually measured either by 1 or 2 variables, while initially 3 variables measuring the same – is what was expected.
Answer to RQ
No, escapism and risk behavior among St. Petersburg students are not associated with each other.
References
- Jouhki, H., & Oksanen, A. (2021). To Get High or to Get Out? Examining the Link between Addictive Behaviors and Escapism. Substance use & misuse, 1-10.
- Hakimzadeh, R., Besharat, M. A., Khaleghinezhad, S. A., & Ghorban Jahromi, R. (2016). Peers’ perceived support, student engagement in academic activities and life satisfaction: A structural equation modeling approach. School psychology international, 37(3), 240-254.
- Topolski, T. D., Patrick, D. L., Edwards, T. C., Huebner, C. E., Connell, F. A., & Mount, K. K. (2001). Quality of life and health-risk behaviors among adolescents. Journal of adolescent health, 29(6), 426-435.
- Esposito, M., Ferrara, M., Panzaru, C., & De Vito, E. (2020). The Relationship between Life Satisfaction and Risk Behaviors: A Cross-Cultural Analysis of Youth. Advances in Applied Sociology, 10(9), 356-368.
- Calleja, G. (2010). Digital games and escapism. Games and Culture, 5(4), 335-353.
- Ioannou, M., Kassianos, A. P., & Symeou, M. (2019). Coping with depressive symptoms in young adults: perceived social support protects against depressive symptoms only under moderate levels of stress. Frontiers in psychology, 2780.
- Veenhoven, R. (2009). World database of happiness tool for dealing with the’Data-Deluge’. Psihologijske teme, 18(2), 221-246.
- Diener, E., & Diener, M. (2009). Cross-cultural correlates of life satisfaction and self-esteem. In Culture and well-being (pp. 71-91). Springer, Dordrecht.
- Tariq, N., & Gupta, V. (2021). High risk behaviors. StatPearls [Internet].
- Kline, R. B. (2016). Principles and Practice of Structural Equation Modeling (Fourth; TD Little, Ed.).
- Brown, T. A. (2015). Confirmatory factor analysis for applied research. Guilford publications.