Investigating online harassment and offline violence among young people in Thailand: Methodological approaches, lessons learned

Abstract

Violence in the physical (offline) world is a well-documented health and social issue among young people worldwide. In Southeast Asia, online harassment (defined as intentional behaviours to harm others through the Internet or through mobile devices) is less well documented. In this paper, we describe and critically discuss the mixed-methods data collection approach we used to build a contextualised understanding of offline violence and online harassment among 15-24 year-old students and out-of-school youth in Central Thailand. We mapped linkages between offline violence and online harassment, and with their possible correlates including gender, sexuality, and mobile media or Internet use. Data collection methods included in-depth interviews, focus group discussions and a custom-built, self-administered computerised survey. Using mixed methods enabled us to collect holistic qualitative/quantitative data from both students and out-of-school youth. In our discussion, we focus on gender, sexuality, class and ethnicity issues in recruiting out-of-school youth; definition and measurement issues; technical issues in using a computerised survey; ethical issues surrounding data collection from minors as well as privacy and confidentiality concerns in collecting data in both in-school and out-of-school settings; and the general implications of using mixed methods.

Introduction

Intentional behaviours to harm others through the Internet or through mobile devices have recently received much attention. Various terms have been used to refer to such behaviours, for example cyberbullying (Tokunaga, 2010), online bullying (Microsoft Corporation, 2012), electronic bullying (Raskauskas and Stoltz, 2007), internet harassment (Ybarra, Diener-West and Leaf, 2007) or online harassment (Wolak, Mitchell, and Finkelhor, 2007).

Various definitions exist for some of these terms, some of which draw on offline bullying theory and exclude behaviours that are not repeated or acted out in within a power relationship (Tokunaga 2010). To avoid implying these more restrictive criteria, we use the term online harassment (Wolak, Mitchell and Finkelhor 2007) as an umbrella term to refer to all types of intentional behaviours to harm others through the Internet or mobile devices. The negative consequences of online harassment can include fear for personal safety, distracting headaches, sleep problems and suicidal ideation (Hinduja and Patchin, 2010; Sourander et al., 2010). Online harassment is also associated with offline bullying (Sourander et al., 2010; Beran and Li, 2007).

Large-scale research programmes (e.g., Livingstone, Haddon, Görzig and Ólafsson, 2011; Microsoft Corporation, 2012) and interventions (e.g., National Crime Prevention Council, 2012) have been launched to map and contain the risks of online harassment. Most of this research has been conducted in North America and Europe. Yet, a 2012 survey of 25 countries (Microsoft Corporation) demonstrates that online harassment is also a problem in developing countries.

In Southeast Asia, online harassment is understudied. However, a recent 25-country survey found that Singapore had the second highest self-reported prevalence of online victimisation in the world (58% in the past year); China had the highest (70%) (Microsoft Corporation, 2012). Thailand was not surveyed. However, a 2010 regional youth media survey indicated that Thai young people were among the region's most avid mobile media users. They spoke on the phone twice the regional average and had more social network contacts than other young people in the region (personal communication, Ipsos, July 10, 2012). This intense media use may be a risk factor for online harassment perpetrated through mobile devices. One published article (Songsiri and Musikaphan, 2011) and other research literature (e.g., Rungsakorn, 2011) demonstrate the presence of online harassment among Thai young people. According to Songsiri and Musikaphan, 52.4 per cent of 1,200 Bangkok high school and vocational school students had experienced online harassment in the past year.

Offline youth violence has been studied more in Thailand than online harassment. Among a national sample of 1105 Thai youth, the self-reported 1-year prevalence of offline victimisation ranged from 4.3 per cent to 18.6 per cent depending on the type of violence (Pradubmook-Sherer et al., 2008). A comparison between these findings and published Thai research on online harassment (Songsiri and Musikaphan, 2011) suggests that a higher proportion of Thai youth might be victimised online than through offline violence.

Studying young people involves specific challenges that may account for the scarcity of previous research on online harassment. Young people have diverse subcultures; representing them accurately may be difficult. Rigorously sampling out-of-school youth is another challenge. Institutional review boards may be reluctant to permit studies that ask minors about sexuality or substance use.

Previous research in Thailand has only examined online harassment among school and college students, not among out-of-school youth. There is qualitative evidence that among heterosexual young people, online harassment may be motivated by intimate relationship conflicts and jealousy (Surat, 2010), but no quantitative data are available on the connections between online harassment and sexuality. Furthermore, while forced sex has been studied among young men who have sex with men and young transgender women (Guadamuz et al., 2011), the impact of online harassment on lesbian, gay, bisexual and transgender (LGBT) youth in Thailand has not been previously investigated. This is worrisome as homophobic and sexist themes are common in online harassment (Shariff, 2008). Furthermore, no previous research has investigated the linkages between online harassment and offline violence among young people in Thailand.

To address these gaps, we conducted a mixed qualitative/quantitative study covering online harassment and offline violence, as well as selected demographics, gender, sexuality and online or mobile media use, among 15-24 year-old Thai students and out-of-school youth. The novel contributions of the study are (1) inclusion of out-of-school youth; (2) detailed coverage of online and mobile media use; (3) inclusion of both offline violence and online harassment; and (4) inclusion of gender and sexuality items, which permit investigating connections between these issues.

The purpose of this article is to describe the data collection methods we used, how we identified participants and gained access to them, the key challenges we faced, and to critically discuss the methodological lessons we learned to inform future research. Empirical findings from the research project are reported separately (e.g., Boonmongkon et al., 2013).

Methods

Definitions

Offline violence and online harassment

The World Health Organization (WHO, 1996) defines violence as

the intentional use of physical force or power, threatened or actual, against oneself, another person, or against a group or community, that either results in or has a high likelihood of resulting in injury, death, psychological harm, maldevelopment or deprivation.

This definition is in our view appropriate for offline violence.

In online harassment, physical force is not used, and the people involved do not always have clearly defined unequal power (Shariff, 2008; Hinduja and Patchin, 2007). Nevertheless, online harassment (Wolak, Mitchell and Finkelhor, 2007) shares the characteristics of violence (as defined above) that it is intentionally harmful and has a high likelihood of resulting in psychological harm; it may even lead to suicide (Hinduja and Patchin, 2010). Only its means differ from offline violence.

Youth

In this study, we used the United Nations definition of youth as implying the age period of 15-24 years (WHO/UNFPA/UNICEF Study Group on Programming for Adolescent Health, 1999).

Design and approach

We aimed at building a contextualised understanding of online harassment, to investigate how it links with offline violence, media use, gender and sexuality among Thai young people. We chose a mixed methods approach to first qualitatively contextualise the phenomena involved and then to quantitatively establish their prevalence and correlates.

Research team

The research team was diverse, representing various academic disciplines (medical anthropology, psychology, epidemiology, and economics); ages (from early 20s to late 50s); genders and sexual orientations. This diversity facilitated access to youth and institutional gatekeepers, particularly when approaching LGBT youth. All researchers in the team except for one were Thai, and all spoke the Thai language.

Study site

The study site was a district bordering Bangkok in Central Thailand, with three subdistricts in suburban and rural areas.

Populations

We tried to include all major groups of 15-24 year-old youth in the district: (1) university students, (2) upper secondary school students (general and vocational, formal and non-formal) and (3) out-of-school youth. With respect to inclusion criteria, participants had to be 15-24 years old; live, work, or study in the study district, and able to participate using Thai language. We did not recruit young people from juvenile detention centres or young people ordained as monks due to the restrictions on their lifestyles, especially on mobile phone and Internet use. Likewise, we did not recruit deaf or blind young people as their media use is likely to be very different to other young people.

Ethical considerations

The study was reviewed and approved by an institutional review board at Mahidol University. After briefing all participants and answering any questions or concerns, we obtained written informed consent from participants who were over 18 years old, and parental consent and assent from participants who were 15-17 years old. Participant information sheets containing information of the study, the research institutions involved and contact information of the principal investigator were given to all participants. Participants received cash incentives: 50 baht (1.66 USD) for completing the survey and 150 baht (5 USD) for participating in focus group discussions (FGD) or in-depth interviews (IDI). Participation in the study was voluntary; participants were informed that they could withdraw from the study at any point without adverse consequences, and that their data would be kept confidential. To protect confidentiality of our participants, participating academic institutions or the study site district will not be reported.

Experiences and findings

Preparation

Community mapping

Mapping educational institutions to recruit students was relatively straightforward. Recruiting out-of-school youth in a way that would represent their diversity required more intensive mapping. We first identified areas with concentrated populations using satellite images from Google Maps. We then visited possible data collection sites at various times and days of the week to determine if there were enough out-of-school young people to justify data collection in that particular context, if data collection was logistically feasible there, and if gatekeepers needed to be contacted. The final study sites included small vendor stalls/shops, Internet cafés, snooker halls, factories, food courts, local markets, housing developments, private homes, sport venues, district council election sites, a temple fair, and a golf course.

Population estimation

We estimated the total student population in the study site by summing the number of students obtained from each relevant educational institution in the district. With respect to out of school youth, the number of all 15-24 year-olds residing in the district according to the Department of Provincial Administration (Population Data Service System) was roughly one tenth of the number of students in the district (1,902 out of 18,000). We considered this figure an underestimate, and extrapolated the total number of out-of-school youth in the district using the total number of students and national enrolment ratios for each level of education.

Sample size and recruitment

We aimed at a large enough sample to draw population inferences separately for university students, secondary students and out-of-school youth at the precision level 0.05, and used a simplified Yamane (1967) formula for calculating these sample sizes. The secondary student sample was a stratified probability sample. The university and out-of-school samples were essentially convenience samples, although we did attempt to represent the diversity of both populations in our sampling.

We aimed to recruit one focus group of male and another focus group of female students from each educational institution (from various grade levels), to capture the diversity of the student population. We aimed to recruit one additional focus group of gay and lesbian students each at one large educational institution. We also sampled male and female out-of-school young people for FGDs and IDIs. We recruited participants from each FGD for an additional IDI whenever the FGD had participants who seemed to have more information than they could provide in the FGD, and were willing to participate in an IDI. We recruited up to two participants for IDIs from each FGD.

Data collection: Measures, instruments and field guides

Computerised survey

Informed by our qualitative findings, we developed a youth friendly computerised survey¹ to collect quantitative data on (1) demographic characteristics, (2) experiences of online harassment and offline violence, (3) online and mobile media use and (4) sexual life and partnerships. We included several items to quantitatively test phenomena described by young people in the FGDs and IDIs.

We adapted some items on offline violence from Pradubmook-Sherer et al. (2008) and others on online harassment from Rungsakorn (2011). We retained previously used item wordings when appropriate to increase the comparability of the findings, and edited wordings when necessary.

Given the broad scope of the survey, the initial paper-and-pencil version took at least 25 minutes to complete. To minimise the risk that participants would find the survey too time-consuming or boring to complete, we computerised the survey. We included colourful animations portraying young people in situations reflecting the study's topic areas, which further helped to make the survey more engaging. Computerising the survey also reduced the time burden on participants through skip patterns and minimised missing or unreliable data through programme checks. We reviewed multiple versions of the computerised survey and animations. The editing, programming and testing of the computerised survey took roughly two months.

Based on pilot testing with students and out-of-school youth, we changed the order of the questions and reworded or deleted some items to shorten the survey. We moved questions on sexual behaviour toward the end of the survey because many pilot testers considered these questions sensitive, but also entertaining and relevant to youth lifestyles. We added a progress bar to give survey takers an idea of how much more was left. We asked our pilot testers if they would respond honestly to substance use questions; none said they would, so we did not include such items.

We operated the final version of the survey programme offline. After the programme stored individual participants' responses on each computer's hard drive, the data were exported onto an online password-protected server for storage, and finally uploaded onto a central password-protected computer to collate all responses and to build, manage and prepare the database for analysis.

Computers

We used netbook computers; they were affordable, lightweight, and had a long battery life, so we could use them for sustained amount of time, even outdoors. We considered netbooks more affordable, easier to use, more compatible with the web browser based survey programme, and more suitable for other uses after the completion of the survey, than tablets and other mobile devices.

Field guide for FGDs and IDIs

We devised a semi-structured field guide to aid data collection during IDIs and FGDs. The field guide comprises 30 open-ended questions covering the following issues: (1) online and mobile media use, (2) patterns of online and mobile media use, (3) social, cultural and economic factors affecting mobile and online media use, (4) effects of mobile and online media use on gender and sexuality, and (5) offline violence and online harassment. Questions on individual behaviours were prepared for IDIs, whereas group norm questions were prepared for FGDs.

Gaining access to school-based and out-of-school youth and engagement with community leaders

School-based young people

We first telephoned all relevant universities and upper secondary schools in the study area to ask for their participation, sent them a formal letter, and made a follow-up visit if necessary to speak to the school's gatekeepers. Only one secondary school chose not to participate, first stating that the school had no violence, but later explaining it was not convenient for the school to participate.

Teachers and administrative staff in participating institutions helped to recruit the requested number of students. These staff briefed the identified students about the study and gave them the informed consent form to decide whether to participate.

In secondary schools, where most students were under 18 years of age, only students contacted in advance participated, since informed parental consent was required from their guardians, as per IRB regulations. At universities, we sometimes recruited participants directly from common areas within various academic faculties, since university students could give informed consent themselves.

Out-of-school youth

To recruit out-of-school youth, we contacted community leaders (four village headmen and one community health worker) after obtaining their telephone numbers from the district office, but they only played a minor role in recruitment. A community health worker knew the parents of youth in the community and helped to recommend venues where young people congregated. The village headmen, on the other hand, had more limited involvement with young people and therefore not helpful to the study. Overall, we found most participants through observation in community locales where young people congregated, like snooker halls, small food stalls, or community stores.

However, we noticed that in these places there were many men, but relatively few women. We therefore had to find other ways of accessing women. One way was to recruit women from a local golf course where many women worked as caddies, as well as in nearby factories where most of the workers were women. When recruiting participants in private enterprises, gatekeepers needed to be contacted. In smaller enterprises, we usually reached an informal agreement to collect data straight away from customers and staff inside or just outside their premises. In larger enterprises (the golf course and one factory) we had to send a formal request letter in advance and make follow-up telephone calls to gain permission. Some enterprises stated they did not have staff in our age bracket of interest.

Our community observation reminded us that out-of-school young people often work during the day. Thus, they are only accessible to researchers after working hours. This prompted us to shift most of our out-of-school recruitment efforts to evenings.

Snowball sampling technique, whereby participants helped to recruit their friends as participants, also worked well. Other community members met during fieldwork also volunteered to help recruit participants, even if they themselves could not participate due to the participant inclusion/exclusion criteria, for example if they were over 24 years old.

Data collection

Data collection was sequential in that FGDs and IDIs were first conducted in order to obtain qualitative data. Then, informed by the qualitative findings, the survey instrument was developed and pilot tested with young people and later used in the quantitative phase that was conducted three to eight months later in each site. The study took about two years to complete, spanning 2011 and 2012.

Qualitative data

We recruited out-of-school participants for FGDs and IDIs directly by ourselves or through existing community contacts; participants in educational institutions were mostly recruited by teachers.

We convened 22 FGDs (out of 26 targeted; 18 with students, 4 with out-of-school youth) with overall 110 participants (52 female, 58 male). We arranged all student FGDs inside educational institutions. Teachers typically had recruited a group of students and introduced them to us, then departed. FGDs with out-of-school youth were conducted in community locations where they spent their free time.

In most cases, at least two research team members facilitated each FGD, with one facilitating the FGD and the other taking notes, observing non-verbal communications, drawing a figure of the seating arrangement and digitally audio recording the discussion. In the beginning of each FGD, the facilitator briefly explained the aims and process of the study, obtained informed consent, and provided snacks to the participants. The facilitator (and sometimes, the note taker) then asked prompt questions loosely based on the IDI/FGD field guide. At the end of each discussion, participants were asked if they had anything to add, or to ask the facilitator or note-taker. If not, we thanked them for their time and gave them their incentives.

We conducted 26 IDIs (out of targeted 52). Similar to the FGDs, IDIs took place in the premises of participating educational institutions, or in public spaces like coffee shops, based on each participant's preference. Since most participants had already been in FGDs, they already knew the aims of the study. To start the conversation, the interviewer reminded the participant what had been discussed in the participant's FGD, then asked open-ended probing questions. IDIs were digitally recorded by the interviewer with the participant's permission.

Quantitative data

We collected quantitative data between September 2011 and March 2012. In secondary schools, the informed consent process had been conducted in advance by teachers using the forms we had provided. In universities and among out-of-school youth, we usually obtained informed consent immediately before data collection. We also verbally briefed our participants about the aims of the study.

Research team members remained seated nearby while participants took the survey, in case participants requested assistance. Participants in educational institutions usually spent 15-30 minutes to complete the survey, whereas out-of-school participants typically spent 30-60 minutes. This reflected differences in their literacy and computer skills.

In secondary schools, we collected quantitative data in classrooms and other spaces set aside by the institution. In universities, we collected quantitative data in classrooms, study areas or open communal spaces. In out-of-school settings, we had to collect quantitative data wherever we found participants and had space for the participant and the netbook to take the survey. These spaces included the back of the research team's pick-up truck, motorcycle seats, local shops or cafés, market stalls, participants' homes, benches next to canals, pool/snooker halls and so on. Thus, in educational institutions, privacy was better than in community locations, where lack of space sometimes meant participants sat very close to each other, or that other youth in the community looked over their responses to the computerised survey.

Altogether 1234 participants completed the computerised survey – by self-identification, they comprised 595 men (48.2%), 554 women (44.9%), and 84 participants who identified with homosexual, bisexual or transgender identity categories (e.g., gay, kathoei, tom or dee; 6.9% of the total sample). Thai ethnicity was stated by 1165 (94.4%); 29 (2.4%) indicated Chinese and 32 (2.6%) stated mixed ethnicity. Most (1193, 96.7%) stated they were Buddhist, while 15 (1.2%) stated they were Muslim and 19 (1.5%) that they were Christian.

Data preparation

Qualitative data

Digital recordings of all IDIs and FGDs were transcribed by a research team member or an outsourced transcriber. The accuracy of transcription was checked by at least one research team member who had been present at data collection.

Quantitative data

Following each round of data collection, we exported raw, unidentifiable data files from each computer on which participants took the self-administered surveys to a password-protected online data storage account and then added them onto our database on a main computer.

Key challenges and lessons learnt

Gender and sexuality issues in recruiting out-of-school youth participants

Young people identifying as “men” were easier to recruit among out-of-school youth than young people identifying with other gender/sexuality categories, including “women”, so men seem over-represented in the out-of-school sample (65.0% self-identified as men and 31.5% as women). Firstly, this may reflect the reverse gender gap in Thai tertiary education – the gross enrollment rate in tertiary education was 55 per cent among women and only 44 per cent among men in 2007 (World Bank, 2009, 35). Secondly, the difficulty in recruiting out-of-school women corroborates the common perception that many Thai women do not spend their free time in public spaces as much as Thai men. Thus, recruiting out-of-school youth only in public places where young people gather during their free time may be insufficient for obtaining a gender-balanced out-of-school sample. Thirdly, most of the research team members were male; this may have contributed to the over-recruitment of male out-of-school youth. Had the team included young female researchers familiar with young women's lifestyles in the district, it might have facilitated data collection among young out-of-school women. To an extent, we countered this obstacle by recruiting more out-of-school young women through enterprises that mostly had female employees.

Sexual/gender minority youth (e.g., youth identifying as gay, tom, dee or kathoei) also seemed underrepresented among out-of-school youth. Only 3.6% of out-of-school survey participants self-identified with sexual/gender minority categories, in contrast to 8.4% of participants in educational institutions.

A study conducted in Northern Thailand among 17-20 year old youth (Tangmunkongvorakul, Banwell, Carmichael, Utomo and Sleigh, 2010) likewise found a larger proportion of male Thai youth attending educational institutions self-identifying with minority categories, or stating they were unsure of their gender or sexuality (9.8%, vocational schools; 21.4%, universities and general secondary schools) than young men who were currently not attending educational institutions (6.7%). However, the Northern Thai study demonstrated a gender difference as a higher proportion of out-of-school young women (19.3%) and vocational school students (22.7%) identified with sexual/gender minority categories (or were unsure) than female university or general secondary school students (10.7%). Overall, a lower proportion of our participants self-identified with sexual/gender minority categories than participants of the Northern Thai study.

In our study, out-of-school sexual/gender minority young people might have been more concerned about privacy and confidentiality than their counterparts in educational institutions. The social context of educational institutions might facilitate the adoption and disclosure of sexual/gender minority identities, for example because there are more role models. Or, some groups of sexual/gender minority young people might actually have higher participation rates in some educational contexts. In contrast, some sexual/gender minority youth may not feel that the physical spaces in the communities where we collected out-of-school data were welcoming, safe, or interesting places to spend their free time. If this is the case, public spaces in these communities may mostly offer recreational opportunities for gender-normative, heterosexual male young people. More research is needed to explore these possibilities. In any case, the findings from both our study and the Northern Thai study (Tangmunkongvorakul et al., 2010) indicate that when the aim is to find out the proportion of sexual/gender minority youth in a given population (e.g., as part of a needs assessment), this proportion depends heavily on the educational (or out-of-school) context of the youth sampled.

Inclusion of migrant youth

Our study site has visible cross-border migrant populations, especially from Cambodia, Burma, Laos and Vietnam. Like Thai youth, young people from these groups use the Internet and mobile technologies, such as Facebook, YouTube, or sending/receiving video clips through mobile phones and making mobile calls. Although many migrant workers speak good Thai, none of the migrant workers we approached could read and write Thai well enough to complete the survey independently. We therefore did not recruit migrant young people as participants in our study.

If migrant young people are to be recruited, self-administered surveys in future studies will need to be translated into each major migrant language. Their appropriateness specifically for migrant youth will need to be checked. However, some migrant workers may not be literate in their own language or the official language of their country of origin, so audio-computer assisted self-interview (ACASI) may a more appropriate alternative. Similar procedures might also be necessary if any non-migrant linguistic minorities present in a given study site would be recruited as participants.

Several young migrant workers we approached seemed afraid of the possible negative consequences of participating in the study, perhaps due to their illegal migrant status and/or previous negative experiences with immigration or other law enforcement officials. Field researchers with matched ethnicity and language skills might facilitate community mapping and participant recruitment.

Appropriateness of computerised surveys for school-based and out-of-school youth

Computerisation made our survey more engaging and interesting to participants because it fit with youth lifestyles that already incorporate computers, the Internet and mobile technologies. Some participants said they felt more comfortable completing the survey on a computer, as there could be no suspicion of their handwriting being identified. Minimal modification would be needed to use our survey program in other Thai contexts or in a nationwide online survey.

However, some participants had difficulty taking the computerised survey, mostly out-of-school youth. Some even declined to participate because they did not think they had the necessary computer skills to take the survey. And so, similar to data collection among migrant populations, ACASI may be more appropriate for data collection outside educational institutions.

Privacy and confidentiality of data collection

During the administration of the computerised survey in educational institutions, we sometimes needed to change seating arrangements or remind other participants and teachers of the importance of privacy, to safeguard participants' privacy and to ensure confidentiality of the data. With out-of-school participants, this was more challenging. Often, friends and family members were curious to see participants use the netbooks. Our requests to respect participants' privacy were sometimes ignored. We did not always feel able to insist on this point given the close relationship between each participant and their friends and/or family. However, many participants themselves invited their friends or relatives to be involved while they took the survey, which suggests they did not find lack of privacy a problem or that they did not think the information they gave needed to be kept confidential from their friends or family members. For example, it was not uncommon for friends of participants to remind or correct participants when they responded to the survey.

Participants' lack of literacy skills also became a privacy issue since some out-of-school participants were not literate enough to complete the survey independently. The research team was subsequently faced with the options of compromising privacy by helping to read some questions to participants and type in open-ended answers on their behalf, or excluding these participants from the survey. Our decision was to offer help if it was minimal (i.e., helping to spell or write out responses to open-ended questions), but if the help needed was beyond this (i.e., reading out the entire survey and filling out the survey for participants), we politely explained to participants that the survey inclusion criteria needed to be observed. Yet, among the research team, this was a contested issue since the criteria on when to request participants to stop completing the survey were not very clear.

Inclusion of minors

We understood that asking minors about their sexual behaviour and violence or harassment experiences could be sensitive, and that these sensitivities could result in our IRB refusing to permit data collection. However, offline youth violence and online harassment are characteristically problems involving and affecting minors (Microsoft Corporation, 2012; Pradubmook-Sherer et al., 2008; Songsiri and Musikaphan, 2011). Many Thai youth also experience their sexual debut and engage in sexual risks at a relatively early age. For example, sexually active 18-24 year-olds studied in the 2006 National Survey on Risk Behavior and HIV/AIDS and ART Knowledge stated they had typically had their sexual debut between 10 to 14 years of age (Sabaiying, 2009). Choosing not to study the risks minors face through violence, harassment and sexual behaviours would mean denying them the protection they need, because in the absence of research evidence, interventions may not be launched at all or will be designed and implemented on the basis of misguided assumptions. With this rationale, we were able to convince the IRB for this study to permit data collection with minors.

Class differences in the appropriateness of the informed consent process

In educational institutions, the informed consent process was generally simple. However, in out-of-school settings, where most participants were working-class, several youth seemed intimidated by the amount of text and the formal, legalistic wordings used in the participant information sheet and informed consent form. These wordings were largely mandated by our IRB, so we were unable to change them.

Signing a consent form and providing further details (e.g., full postal address and citizen identification card number) was a difficult decision for many out-of-school youth. Besides our assurances that their confidentiality would be upheld, they had no way of knowing if their contact information would be used to link them with the sensitive information they disclosed in the survey. Perpetration data divulged by participants could in theory be used as evidence against them. In practice, the participant code numbers we gave them cannot be linked to any identifying information.

The wordings and the process of the informed consent procedure, largely derived from clinical research and intended to protect participants, might increase the confidence of an adult, middle-class, well-educated person to participate in a clinical trial. Yet, these procedures may be less appropriate for social research involving young people, especially out-of-school youth, for whom literacy, command of formal language, and fears about loss of confidentiality may pose issues. This could make such youth under-represented in the scientific literature.

Verbal informed consent might be more appropriate and less intimidating to the supposed beneficiaries of the informed consent process. Parental consent should perhaps be possible to waive in surveys where data cannot be linked back to individual participants, and obtaining parental consent may indeed be difficult for youth who do not spend much time with their parents, or live far away from them. In contexts where many young people migrate to work or study elsewhere, young people living far from their parents may constitute a significant proportion of all youth. Even when the parents are accessible, requesting young people to ask their parents to consent to their participation in a research study may actually increase their risks. For example, sexual minority youth asking their parents to participate in a “gay” study would implicitly disclose their sexual orientation identity to their parents and risk negative reactions. Similarly, young people abused by their parents could be put at risk if researchers asked them to obtain parental consent for participating in a study on child abuse. We are planning to investigate and report how Thai out-of-school youth experienced our written informed consent process.

Definition and measurement issues

Violence and harassment

In FGDs, we noticed that participants had a narrow perception of violence. For example, offline violence was perceived to only include physical violence. Verbal violence and most types of online harassment fell outside most participants' definition of violence. We circumvented these problems by framing the survey questions as descriptions of specific behaviours (e.g., “pressing others to have sex or to make a sexual performance over a webcam”) rather than as abstract categories (e.g. “online sexual harassment”).

Sexual behaviour

Participant interpretations of survey items affect their responses. For example, our question “have you been forced to have sex” may have been less problematic than asking if they had been raped (DeKeseredy and Schwartz, 2001). However, survey takers may still have had different perceptions of what constitutes having sex or being forced. Surveys in Thailand have found that few participants think that non-penetrative or non-vaginal sexual behaviours constitute having sex (Jackson, 1999). This may affect responses to sexual violence and items on sexual behaviour in general, implying under-reporting of non-vaginal, non-penetrative sex unless survey items are worded in terms of specific sexual behaviours.

Internet and mobile phone use

We measured both Internet and mobile phone use with specific items, for example, how many minutes/hours participants used any given application, or spoke on the phone with certain types of contacts (family, partner, friend etc.) per day. The response choices were in 10-minute increments for the first hour, and in hours thereafter, up to “more than four hours” per program or type of person. This approach may provide meaningful estimates of specific program use or specific types of calls, but our experience suggests that it leads to inflated overall estimates of Internet use or mobile phone calls when the specific categories are summed up. In part, this may be due to the accumulation of inaccurate estimates. More importantly, adding up time estimates for each programme inherently assumes that only one programme is used at a time, while in fact many young people use several Internet or mobile functions simultaneously. Youtng people may have counted the time they were online with each program, rather than the time they actively used each program, so their estimates should not be taken to mean they are not engaged in other activities at the same time, online and/or offline. Thus, the total time spent daily on the Internet or mobile phones might need to be assessed separately from the time spent on specific programs or functions.

Technical issues in using a computerised survey

We initially considered marketing and conducting our survey entirely online. The benefit of an anonymous online survey could facilitate truthful responses to sensitive items, such as illicit substance use. However, unless careful precautions were taken, participants could face the risk of legal repercussions for their past illegal activities, since it is technologically possible, and currently permitted by Thai law, for state authorities to track individuals using their IP addresses.

For us, this option was not feasible mainly because we wanted geographically focused data from youth, and verifying the geographic location and age of participants would have been difficult. In a Thai-language nationwide survey with a broader age bracket, these problems could be avoided, as most Thai speakers live within Thailand. However, obtaining written informed consent (if necessary from parents) would be challenging with this kind of survey. A waiver of parental consent would thus need to be in place for such studies. We also felt that it was necessary to include young people who did not use the Internet to be able to make comparisons between Internet users and non-users. We therefore chose to operate our survey program offline. Doing so meant additional workload in transferring data from netbooks to our central database. However, it also helped to circumvent data loss due to network failure, which would have been likely had we relied on wireless Internet connections, especially when collecting data outdoors among out-of-school youth.

Implications of mixed methods

Schatz and Williams (2012) have argued that uniform, multi-country surveys would greatly benefit from the addition of a qualitative component to assess the validity of survey questions and to culturally contextualise quantitative findings. Ideally, the qualitative component in mixed methods studies should be implemented before, during and after survey data collection to develop and validate survey questions, contextualise the findings, and finally, help to interpret survey data.

Correspondingly, we benefited from conducting IDIs and FGDs before collecting survey data, as it helped us to develop and validate our survey items. We added culturally specific phenomena (e.g., specific types of violence or online harassment) as survey items and adjusted the wordings of items adopted from previous research to make them more culturally appropriate.

However, the qualitative data that helped us make sense of our survey findings were not only gathered formally through IDIs and FGDs. Also important were the notes we made during survey data collection about procedural issues, for example on how the youth we asked to participate in the study felt about our computerised survey or about the informed consent process. As explained above, these issues sometimes determined how likely each group of young people was to participate in the study, and consequently, how well these groups are represented in the resulting data.

Conclusions

Youth researchers have a responsibility to reflect diversity among young people in their work. School-only samples may be easy to obtain but only represent students. Similarly, studies that do not involve minority ethnic participants or questions about gender and sexuality fail to represent important minorities and bias the research literature on youth. Yet, capturing this diversity involves methodological challenges. We have presented one approach of studying offline violence, online harassment, gender/sexuality and media use among students and out-of-school youth. We feel the present study greatly benefited from the use of a mixed methods approach, the computerised survey that permitted inclusion of a large number of survey questions, coverage of both Internet users and non-users, a holistic understanding of youth violence, and the linkages between offline violence, online harassment, gender and sexuality. However, our approach was less successful in including youth with low levels of literacy, computer or Thai language skills. We discussed alternatives, such as ACASI or surveys in multiple languages. We also reflected on how an excessively legalistic informed consent process can intimidate its supposed beneficiaries and how the creation of overall media use estimates by summing up specific variables can lead to unrealistically inflated estimates. In conclusion, the experiences we have presented in this paper reflect the challenges involved in a mixed methods study but also the strengths of mixed methods in creating culturally relevant and statistically representative data on violence, gender/sexuality, and media use.