There is growing evidence of a decline in the mental health of the young in Western developed countries such that, today, the young tend to be less happy than their older counterparts. This column explores new research across Asia and the Middle East which suggests that the mode of a survey influences reported wellbeing: web-based surveys suggest young people are less happy, while face-to-face or telephone surveys suggest the opposite. The authors consider why this might be the case, and the implications for our understanding of trends in mental health.
There is growing evidence of rising mental ill-health among young people in many parts of the world, notably in developed Western countries. In a previous Vox column, in seeking to explain this trend, we pointed to the rise in social media, underpinned by new technologies, most notably the smartphone, and proposed the regulation of smartphone usage by young people as one possible policy response (Blanchflower and Bryson 2024a).
Since then, further studies have been undertaken to see whether the decline in young people’s mental health is universal and, if not, why not. Further, albeit tangential support, for the role of social media was provided by studies showing no uptick in young people’s poor mental health in countries with little or no easily accessible internet infrastructure, such as part parts of Africa (Blanchflower and Bryson 2024b).
A cautionary tale emerges from the most recent efforts to map mental health by age: the importance of survey mode effects. Survey mode effects arise when survey respondents’ answers to survey questions vary systematically with survey design. Traditionally, surveys were conducted face-to-face by interviewers recording responses to questions on a laptop (computer-assisted personal interviewing, or CAPI). But with advances in technology these relatively expensive CAPI surveys have been replaced by computer-assisted telephone interviewing (CATI) and computer-assisted web interviewing (CAWI). CAWI differs from CATI and CAPI because it relies on self-reports with no intermediary survey interviewer.
When examining trends in wellbeing and illbeing over time, one needs to be cognisant of mode effects because push-to-web has become more common among survey companies over time, in part to save money. Individuals responding face-to-face report better psychological functioning than those who respond over the web (Kocjan et al. 2023). Furthermore, people completing interviewer administered questionnaires are more likely to provide socially desirable responses than if they are self-administered (Rickwood and Coleman-Rose 2023). This is known as social desirability response bias which is the tendency to underreport socially undesirable attitudes and behaviours and to over report more desirable attributes.
In a new paper (Blanchflower and Bryson 2025) we find clear, unambiguous evidence across 16 countries in Asia and the Middle East that young people are both less happy and more unhappy than older people when the survey is conducted via a the web. However, when the survey is conducted by telephone or face-to-face a different picture emerges: the young are the happiest on positive affect metrics, and also the least unhappy on negative affect metrics.
These results are suggestive of biases associated with survey mode effects, but they are not conclusive. Survey mode is not randomly assigned across respondents; instead, people tend to get offered the interviewer-based surveys when they have not responded on the web, making it hard to infer that the survey mode is having a direct effect on responses. Furthermore, the propensity of individuals to respond to a survey, conditional on mode, varies. Younger people are more comfortable with the web, so are more likely to respond to a CAWI survey than older people. Conversely, younger generations are less comfortable with using the telephone to make or receive calls than the older generations. So, a second effect will come via non-response biases that might differ systematically with age. These can be addressed, to some degree, by reweighting the analysis back to population distributions. Finally, there are real mode effects – that is, conditional on responding to a survey question, do you do so differently depending on whether there is an interviewer around or not? Even here one needs to be careful because what is ostensibly the same question may have different prompts depending on whether it is asked face-to-face, by telephone, or online.
Since the age profile of poor mental health differs across survey modes, is it reasonable to infer that the young are now less happy and more unhappy than older people, as the web-based surveys suggest? There is quite a bit of evidence that this is the case in Asia and the Middle East. For example, in Japan the numbers of suicides among young people, especially girls through high school, was at an all-time high in 2024 (Benozaapan 2024). There is also evidence of increases in anxiety, depression, and self-harm mental health disorders in Asian countries since 2019, especially among young females (Zhang et al. 2025), as well as evidence of rising levels of psychiatric admissions of the young along, with rising anti-depressant prescription rates and rising rates of self-harm and suicide, in some countries such as Australia and Japan. Such evidence leads us to suspect that social desirability response bias operates in Asia and the Middle East leading young people to under-report socially undesirable affective states to interviewers.
Source : VOXeu
