Youth Violence and Video Games: A Second Data Analysis

In September 2018, I published the world’s largest study on the relationship between video games and violent behaviour. Using data from 203,121 young people, I found that the relationship between video games and physical violence was miniscule, with factors such as risk-taking behaviours and home life explaining more of violent behaviour than gaming. I argued that to keep young people safe, more resources should be dedicated to conflict resolution and coping skills, better anti-bullying campaigns and more.

Fifteen months after this study, it is a disheartening time to be a data scientist. Further research has been conducted on the topic of video games and violence, including the finding that 27 hours of violent video games per day would be required to influence violent behaviour (Ferguson & Wang, 2019). Despite evidence to the contrary, divisive rhetoric has seen video games blamed for further atrocities (Farokhmanesh, 2019), causing hashtags such as #VideoGamesAreToBlame to trend on social media.

While it is disheartening, researchers must continue to keep people educated and informed in the face of rhetoric. In the mission to keep people educated, a key scientific principle researchers follow is something known as ‘falsifiability’. To keep things simple, falsifiability means that we should try to disprove and discredit our theories and findings wherever possible. If a researcher keeps trying to disprove themselves and keeps finding support for their theory, it is a good indication that our findings are strong and valid.

I refuse to let rhetoric prevail over science. I am returning once again to conduct a second data analysis on video games and violence. Following the principle of falsifiability, I have set out to discredit my original findings. The current study will follow a very similar procedure to my previous study, but will use a brand new large dataset of participants. Was my original finding a fluke, or am I able to replicate it with brand new participants? Let’s take a look.

As usual, there will be a summary at the end if you do not wish to read everything. Thank you and please enjoy!

Contents

1. The Data
2. Data Analysis Strategy and Checking
3. Results
4. Discussion
5. Critique
6. Summary
7. References

The Data

In the previous study, I used data from the 2014 wave of the Health Behaviour in School-Aged Children (HBSC) study. For this data analysis, I will be using the 2010 wave of HBSC data.

The HBSC is a large-scale endeavour to improve young peoples’ wellbeing by researching their health behaviours and social context. Data is gathered for this study via an anonymous questionnaire. The 2010 wave of the study includes a total of 213,595 11, 13 and 15 year old participants from 43 countries.

This dataset is open-access and is open for all researchers to analyse, providing that they abide by the License Agreement. In accordance with the License Agreement, I must copy and paste the acknowledgement text below. Feel free to move on to the next section.

HBSC is an international study carried out in collaboration with WHO/EURO. The International Coordinator of the 2009/10 survey was Prof. Candace Currie and the Data Bank Manager was Prof. Oddrun Samdal. The 2009/10 survey was conducted by Principal Investigator Candace Currie in 43 countries. For details, see http://www.hbsc.org.

Data Analysis Strategy and Checking

Data will be analysed using what is known as a regression analysis. This analysis is the same as my previous study, which was influenced by research by DeCamp and Ferguson (2017).

Simply put, regression analyses are used to explain why something happens (known as ‘variance explained’). For example, when trying to explain number of ice creams sold on a day, significant predictors may include variables such as temperature, time of year etc.

As regression analyses produce a lot of statistics, I have decided to report on the variance explained by the model alongside β and t values for each significantly contributing variable. While β values can be thought of as how much the variable contributes to explaining a behaviour, t values demonstrate the strength of the relationship.

The HBSC study provides rich data on the health and socioeconomic contexts of young people. To help readers understand the variables that have been entered into the regression analysis, I have grouped variables together into easy-to-understand categories. Due to differences in the questionnaire between 2010 and 2014 data collection, some categories will differ from the previous analysis.

The dependent variable (what we’re looking to explain) is how often a participant has been involved in a physical fight in the past 12 months; answers range from ‘none’ to ‘four or more’. The categories that will be used to explore why young people get into fights include:

Individual factors: This category includes the age and gender of the participant.

Free time: This category involves how young people choose to spend their free time. Variables include: hours spent watching TV (weekdays and weekend), hours spent using a computer (weekdays and weekend), hours spent playing video games (weekdays and weekend), and number of hours of exercise per week.

Risk factors: The HBSC study identifies the following variables as risk factors for young people: smoking, drinking, getting drunk, cannabis use and underage sex. All variables apart from underage sex looked at how often these behaviours occurred in the 30 days prior to the study. For underage sex, young people were simply asked to report ‘Yes’ or ‘No’.

Bullying: As bullying may involve physical violence, both being bullied and being a bully in the two months prior to the study were included in the analysis.

Home life: As home life variables were significant predictors of youth violence in DeCamp and Ferguson’s study, all variables regarding home and family life from the HBSC were included in the analysis. These variables include: presence of mother in the home, presence of father in the home, and ease of communication with parents.

Social life: As friendship circles can be an influential factor in bullying and violence (O’Connell et al., 1999; Pepler et al., 2010), all friendship variables from the HBSC were included in the analysis. These include: ease of communication with best friend, ease of communication with friends of the same sex, ease of communication with friends of the opposite sex, number of close male friends, number of close female friends, how often time is spent with friends after school, and how often time is spent with friends in the evening.

Health: As poor physical and mental health can be an influential factor in youth violence (Resnick et al., 2004; Patel et al., 2007), all mental and physical health variables from the HBSC were included in the analysis. These include: general self-reported health, how often the young person feels low, how often they have temper issues, how often they feel nervous, and their general life satisfaction.

School life: As academic performance and school problems are linked with school violence (Resnick et al.), all school life variables from the HBSC were included in the analysis. These include: whether the young person likes school, their academic achievement, whether students like to be together, whether students are helpful, whether students accept the young person, and whether the young person feels pressured by school.

Socioeconomic status (SES): As SES was a significant contributor to youth violence in DeCamp and Ferguson’s research, several SES variables were included in the analysis. These variables include: number of cars in the household, how often the family go on holiday together, number of computers, whether the young person has their own bedroom, and whether they consider their family to be wealthy.

Before conducting the regression analysis, I wanted to ensure that the regression results would be accurate by checking a number of data assumptions. Please feel free to ignore anything written in red, this is all statistical talk.

The multicollinearity of variables was assessed using both Tolerance and Variance Inflation Factor (VIF) measures. No VIF value exceeded 5 and each Tolerance value exceeded 0.2, suggesting that variables are sufficiently independent for the analysis. The influence of undue cases was assessed using Mahalanobis Distance, Cook’s Distance and Centered Leverage Values. 416 participants had a Centered Leverage Value that exceeded three times the mean; these participants were removed from the analysis. Multiple participants had a Mahalanobis Distance value which exceeded 15. However, Stevens (2002) recommends that the Cook’s Distance value should be consulted before removing participants with a Mahalanobis Distance value exceeding 15. Due to the very small size of Cook’s Distance values in the analysis (Maximum: <0.001) which is to be expected from having a very large sample size, no further participants were removed as part of this check.

Results

After removing participants who did not answer the fighting question and those with a high Centered Leverage Value, the final participant pool for data analysis was 191,803. This pool included 97,782 females and 94,021 males.

The variables included in the analysis explained 26.4% of why young people get into fights (Adjusted R-Squared = .264). 31 variables emerged as significant predictors in this model. After my previous study was published, I saw some requests for a table of results rather than a list of results. Not one to ignore good feedback, the findings of this study are detailed in the table below.

Variables will be organised in accordance with β and t value sizes. This means that the first variable listed explains the most variance in youth violence, while the final variable explains the least. Study results are as follows:

Predictor	β Value	t Value
Being male**	0.24	57.12
Being younger**	0.17	44.11
Engaging in underage sex**	0.16	44.43
Being a bully**	0.16	43.37
Underage cannabis use**	0.06	17.24
Having temper problems**	0.06	15.12
Underage alcohol use**	0.06	13.00
Disliking school**	0.05	12.82
Being a victim of bullying**	0.05	12.55
Exercising often**	0.04	11.29
Spending evenings with friends**	0.04	9.82
Lower academic achievement**	0.03	8.82
Not having own bedroom**	0.03	8.56
Going on family holidays**	0.03	8.29
Often feeling nervous**	0.03	7.38
Having more close female friends**	0.03	7.11
Underage smoking**	0.03	6.80
Spending time after school with friends**	0.03	6.35
Having difficulties talking to mother**	0.02	5.51
Feeling pressured by schoolwork**	0.02	4.38
Watching TV on weekends**	0.02	4.26
Not viewing peers as helpful**	0.02	3.97
Finding opposite sex easy to talk to**	0.02	3.85
Playing video games on weekdays**	0.02	3.65
Playing video games on weekends**	0.02	3.62
Feeling healthy*	0.02	2.67
Having few/no computers at home*	0.01	3.27
Having low life satisfaction*	0.01	3.04
Having difficulties talking to father*	0.01	2.32
Enjoy spending time with peers*	0.01	2.22
Watching TV on weekdays*	0.01	2.07
**p < .001, *p < .05

Discussion

In my original study, I wished to examine the relationship between video games and violence using a more realistic and reliable measure of violence. Previous video game violence research has used measures of violence that include playing a loud noise to a computer and measuring a portion of hot sauce (Hollingdale & Greitemeyer, 2014; Hasan et al., 2013). When examining real-life violence, factors such as family and socioeconomic status were substantially more influential than video games (DeCamp et al.). I too found this in my data analysis, and set out to replicate this finding to see if my original finding was a fluke.

In my previous data analysis, I identified 20 factors that explained more of why young people get into fights than video games. In the current data analysis, this number increased to 23 factors. The largest factors in the previous analysis remained significant in the current analysis. These factors include: being a young male, being a bully, engaging in risk-taking behaviours such as underage drinking, and having an unsupportive home life.

Despite setting out to discredit myself, I found further support for the idea that focusing on video games to minimise violent behaviour would be a pitiful waste of resources. To repeat myself from my previous study, focusing on video games to reduce violence could be considered a social injustice for two reasons.

Firstly, the weak relationship between gaming and youth violence suggests that reducing gaming will not be massively effective in reducing violence. This would continue to put young people at risk of injuries such as concussions. Secondly, this analysis suggests that placing resources into other factors will not only be better at reducing youth violence, but will also be more beneficial for young people. A few data-driven examples from the current analysis include:

• Teaching young males healthier ways of dealing with conflict and emotions. This would be helpful as this population experiences many social, emotional and hormonal changes at this age (Currie et al., 2001).
• Providing more funding to anti-bullying initiatives that address factors such as physical bullying and gang violence.
• Further funding for initiatives that try to reduce risk-taking behaviours such as underage drinking and drug use.

In my previous analysis, one particularly ludicrous finding gained attention: family vacations explained more of why young people get into fights than video games. This finding was replicated in the current study as family holidays once again explained more of youth violence than video games. However, an even more ludicrous finding emerged in this study. It was found that number of close female friends was a larger predictor of youth violence than video games. This finding suggests that interventions to minimise number of female friends would be more effective at reducing youth violence than focusing on gaming. Please note that I don’t advocate for this in any way, but is an indication of the ludicrous measures that would be more effective at minimising violence than controlling video games.

Critique

All good research provides self-critique so that future researchers can learn from it. Before concluding, I’d like to list one positive this study has over the previous study, then offer a few self-critiques:

• During my time conducting data analysis on my website, I have drawn some ire for using data that does not systematically exclude those who play mobile games. In the 2010 HBSC questionnaire, young people were asked ‘About how many hours a day do you usually play games on a computer or games console (Playstation, Xbox, GameCube etc.) in your free time?’. This definition does not include mobile games, so these readers can rest easy.
• As you may have noticed from the question, gaming data includes those who may or may not play violent video games. In my original study, this was justified due to a violent act occurring at a gaming event for a non-violent game. This resulted in discussions on whether components of video games themselves (e.g. competitiveness) influence violent tendencies. I feel that this justification continues to be valid in 2019 discourse. #VideoGamesAreToBlame was a globally trending hashtag (rather than #ViolentVideoGamesAreToBlame), and media headlines frequently reference ‘video games’ rather than ‘violent video games’ (example headlines here). As parents may be concerned that their child’s favourite hobby could be a gateway to destructive and harmful behaviour, I wished to provide data to ease their concerns.
• HBSC 2010 data was collected from an impressive 43 countries. However, data was primarily collected from Europe and North America. It would be useful to have further global representation in this data, such as from Asian countries.
• As both being a bully and being bullied were related to youth violence, asking young people whether they have been in a fight may capture both perpetrators and victims. To minimise this in the future, I recommend using DeCamp and Ferguson’s phrasing: ‘hitting someone with the intent to hurt them’.

Summary

• In my original data analysis, I found that the relationship between video games and violence was miniscule. 15 months later, video games continue to be blamed for tragic events. As rhetoric must be challenged using evidence and data, I am once again exploring the relationship between video games and violence to identify if my original finding was a fluke.
• My previous analysis was replicated using data from the 2010 Health Behaviour in School-Aged Children (HBSC) study. This dataset contains health and wellbeing data from 213,595 11-15 year olds across 43 countries.
• The following categories of variables were entered into a regression analysis to explain why young people get into fights: age and gender; free time; risk factors; bullying; home life; social life; health; school life and socioeconomic status. After removing undue cases and those who did not answer the fighting question, the final sample size was 191,803.
• The model explained 26.4% of why young people get into fights. 23 factors explained more of youth violence than video games, an increase of 3 from the previous study. Despite setting out to discredit my original finding, I found further evidence that the relationship between video games and violence is miniscule.
• In my original analysis, one particular finding gained attention due to its ludicrous nature: going on family vacations explained more of youth violence than video games. Not only was this found again in the current study, but an even more ludicrous finding emerged: number of close female friends explained more of youth violence than video games. These factors serve as indicators of just how weak the relationship between video games and real-life violence is.
• This analysis reinforces the idea that focusing on video games as a method of reducing youth violence would be cost-ineffective, would produce few benefits, and would continue to place young people at risk of physical injuries. To maximise social good and protect young people, interventions such as teaching healthier coping mechanisms, investing in anti-bullying schemes, and further investments in minimising risk-taking behaviours should be prioritised over trying to reduce video gaming.

Thank you all very much for reading! This hard work would not be possible without the support of my wonderful Patrons. I would particularly like to thank my Platinum Patrons: Albert S Calderon, Kyle T, redKheld, DigitalPsyche, Brent Halen, Dimelo ‘Derp’ Waterson, Hagbard Celine, Aprou, Austin Enright, SK120, NotGac, Shaemus, Edward Pang, Joey Rodriguez, Marcus Lo Re-Sant, DarrenIndeed, Morgan, Thomas Meszaros, Ciara Elizabeth, Jackson Jin, and Mulgar. Thank you!

 

References

Currie, C., Samdal, O., Boyce, W., & Smith, B. (2001). Health behaviour in school-aged children: A world health organization cross-national study, research protocol for the 2001/02 survey. Edinburgh: Child and Adolescent Health Research Unit

DeCamp, W., & Ferguson, C. J. (2017). The impact of degree of exposure to violent video games, family background, and other factors on youth violence. Journal of Youth and Adolescence, 46(2), 388-400.

Farokhmanesh, M. (2019). Trump and Republicans continue to blame video games for their failures on gun control. Retrieved December 3rd, 2019 from https://www.theverge.com/2019/8/5/20754793/trump-gun-control-video-games-violence-republicans-no-evidence-dayton-el-paso-texas-ohio

Ferguson, C. J., & Wang, J. C. (2019). Aggressive video games are not a risk factor for future aggression in youth: A longitudinal study. Journal of Youth and Adolescence, 48(8), 1439-1451.

Hasan, Y., Bègue, L., Scharkow, M., & Bushman, B. J. (2013). The more you play, the more aggressive you become: A long-term experimental study of cumulative violent video game effects on hostile expectations and aggressive behavior. Journal of Experimental Social Psychology, 49(2), 224-227.

Hollingdale, J. & Greitemeyer, T. (2014) The Effect of Online Violent Video Games on Levels of Aggression. PLOS ONE, 9(11): e111790.

O’Connell, P., Pepler, D., & Craig, W. (1999). Peer involvement in bullying: Insights and challenges for intervention. Journal of Adolescence, 22, 437–452. http://dx.doi.org/10.1006/jado.1999.0238.

Patel, V., Flisher, A. J., Hetrick, S., & McGorry, P. (2007). Mental health of young people: a global public-health challenge. The Lancet, 369(9569), 1302-1313.

Pepler, D., Craig, W., & O’Connell, P. (2010). Peer processes in bullying: Informing prevention and intervention strategies. In S. R. Jimerson, S. M. Swearer, & D. L. Espelage (Eds.), Handbook of bullying in schools: An international perspective (pp. 469–479). New York, NY: Routledge.

Resnick, M. D., Ireland, M., & Borowsky, I. (2004). Youth violence perpetration: what protects? What predicts? Findings from the National Longitudinal Study of Adolescent Health. Journal of Adolescent Health, 35(5), 424-e1.

Stevens, J. P. (2002). Applied multivariate statistics for the social sciences (4^th ed.). Hillsdale, NJ: Erlbaum