Effects of Survey Mode and Age Cohort Differences on the Measurement and Perception of Violence: Findings from the Luxembourgish Victimization Survey

Violence is undoubtedly one of the most sensitive topics in modern society. For a long time, victims remained largely invisible in the public discourse on violence. In criminal law, police and judicial statistics, and conventional criminology, the (suspected) perpetrator often takes center stage. In recent decades, however, attention has increasingly focused on individuals as victims of crime. Victimization surveys have become a standard tool in social science research in many countries.

Face-to-face interviews, often considered the gold standard, especially when using specialized techniques, are preferred to collect data on such sensitive topics. However, these surveys are costly, time-consuming, and not always feasible. Traditional paper and pencil surveys are unsuitable due to the complexity of filtering. Telephone and web surveys also present challenges, especially in reaching specific populations (e.g., "mobile-only" respondents vs. the elderly or those with limited literacy skills).

In the Luxembourgish Victimization Survey (N = 5,695), we used a mixed-mode design (CAWI/CATI) to determine the extent to which the population is affected by violence (victimization). In addition, we asked whether certain behaviors (e.g., parental violence, sexual violence in marriage, etc.) are perceived as violent phenomena (awareness). With regard to social desirability effects, we find that the survey mode per se has no significant effect on the probability that respondents report or do not report experiences of violence (victimization), nor on whether or not certain phenomena of violence are perceived as such (awareness). However, there were significant interactions between survey mode and age cohort. In older cohorts, respondents were significantly more likely to report experiencing violence when they participated via the Internet rather than by telephone, even after controlling for sociodemographic factors. Conversely, younger respondents were significantly more likely to classify scenarios as violent incidents when interviewed by telephone compared to the Internet.