Students under lockdownComparisons of students’ social networks and mental health before and during the COVID-19 crisis in Switzerland

Timon Elmer, Kieran Mepham, Christoph Stadtfeld•View original

OverviewBalancedalloy voice

Picture a university that goes quiet almost overnight. Lecture halls empty. Cafeterias close. Students head back to their apartments and dorm rooms, and campus life compresses into video calls and text threads. That was Switzerland in early April 2020, about two weeks into nationwide social distancing and a full university lockdown. And in that sudden stillness, a simple, hard question came into focus: when your social world thins, what happens to your mind? Elmer, Mepham, and Stadtfeld set up a rare natural experiment to watch this unfold. They had been following Swiss undergraduates since 2018, so when the lockdown hit they could compare students to their own pre-pandemic selves and, just as important, to an earlier cohort that never experienced the crisis that spring. Two lenses, same shock. They looked at within-person change from September 2019 to April 2020, and a between-cohort contrast: April 2020 under lockdown versus April 2019 in normal times. The focus was tight, examining four kinds of ties among classmates—pleasant interaction, friendship, co-studying, and support, both informational and emotional. They also assessed four mental health outcomes using standard scales: depressive symptoms on a scale from zero to sixty, anxiety on a scale from zero to twenty-one, perceived stress on a scale from zero to forty, and loneliness on a scale from one to four. In April 2020, they also asked students to list important people in their personal networks, how often and how they communicated, and whether those people offered emotional support. They layered in COVID-specific worries, including concerns about health, family, money, and the future, along with a checklist of isolation experiences. If you're wondering how they turned all that into answers, here's the high-level map. For the social networks, they counted how many classmates each student named in each category, and what share named no one at all—an "out-isolate." For mental health, they compared each person's spring score to their own fall baseline. For the between-cohort comparisons, they matched the lockdown group to that earlier, unaffected year and tested mean differences. Because they had many potential predictors of mental health change, they used an exploratory regression with backward selection. This means they started with a full model and pruned it until the remaining variables cleared a liberal p-value threshold to identify the strongest associations. That's not a causal design; it's a way to make sense of a lot of moving pieces in a small-to-moderate sample. So what actually changed when the campus shut down? The short version is that everyday interaction and study partnerships shrank, while friendships and support were more resilient. Within the same students from fall to lockdown, the number of pleasant interaction partners dropped by about zero point eighty ties, and co-study partners fell by zero point forty-six. Those aren't dramatic numbers individually, but across a cohort they add up. The share of students with no study partner at all jumped from twenty-two point six percent to thirty-nine point two percent—nearly doubling the odds of being academically on your own. Friendship nominations held steady, and support ties ticked up slightly, suggesting that the core emotional scaffolding didn't collapse even as casual and collaborative ties thinned out. Looking across years told a similar story. Comparing April 2020 to April 2019, the lockdown cohort named fewer partners in interaction and co-studying, with small to moderate effect sizes, while the proportion with zero ties didn't differ consistently across networks. The texture of connection changed too. Face-to-face contact plunged, and students leaned hard on their phones and laptops. In April 2020, they still reported around five point six pleasant interaction ties, but most of that ran through digital channels—roughly three messaging ties and about three call ties—and the same people often appeared across modes. About three in four in-person ties also existed digitally. In other words, communication didn't stop; it rerouted. One way to see the difference between a casual nod and a lifeline is to watch which ties survived. The team tracked specific pairs from September to April and asked, is this tie still there? For general interaction, two-thirds survived. But that average hides a striking gradient: if a tie stood alone with no overlap with friendship, support, or co-studying, its survival was about forty-five percent. If it overlapped with two to four other dimensions, survival jumped to the mid-eighties. Co-studying was even more fragile. A co-study tie with no other connection limped along just six percent of the time; one embedded in all four dimensions survived fifty-seven percent. On average, fewer than half of co-study ties made it. Multiplex ties, the ones that wear several hats, were the ones that endured. As the social fabric stretched, mental health frayed. Within the same individuals, depressive symptoms rose by four point forty-four points on that zero to sixty scale. Stress rose by two point sixty-seven on zero to forty. Anxiety crept up by zero point sixty on zero to twenty-one. Loneliness ticked up by zero point thirteen on a one to four scale. Those are modest to moderate shifts depending on the dimension, but they're consistent and statistically solid. Between cohorts, spring 2020 and spring 2019 didn't differ significantly on those mental health averages, which tells you cohort composition and timing matter; it doesn't erase the within-person deterioration observed in the lockdown group. What pushed those changes? The COVID-specific stressors weren't abstract. Students who worried more about family and friends saw larger increases in depression and stress—roughly one point five points more depression and one more point of stress per step up in family or friend worry. Worrying about one's future career tracked with higher anxiety and higher stress too, with the stress impact being about one point thirty-six points per step. And the standout signal was a simple, human one: people who felt that problems they usually keep in check were bubbling up showed bigger jumps across the board—about four points more depression, one and a half more anxiety, and nearly one more point of stress. You can hear the mechanism in that phrasing: the external shock turned up the volume on the internal noise. Isolation in the networks mattered as well, but not always in the way you might expect. Being the kind of person who named no one in the general interaction network predicted higher anxiety, as did being isolated in the informational support network; those coefficients were on the order of two points of additional anxiety. Feeling cut out of the day-to-day and lacking someone to turn to for advice are both jangling states in a crisis. A curious nuance showed up with friendship isolates: students who named no classmates as friends reported slightly lower increases in depression but higher increases in loneliness. One way to reconcile that is to remember that the classmate network is only part of a student's social world; some may have leaned more on partners, family, or old friends outside the cohort when campus ties went dormant. Personal networks helped explain who struggled and who coped. Smaller personal networks in April predicted larger increases in loneliness. Living alone during lockdown aligned with higher increases in depression, on the order of three points. The average level of emotional support in one's personal network—do the people you talk to actually comfort you?—was linked to better outcomes: more support predicted lower increases in depression by a bit more than one point and lower increases in loneliness by a small but reliable amount. Not just contact, in other words, but caring contact. Gender cut across these patterns. After accounting for social integration and COVID-related stressors, female students showed larger increases in all four mental health measures—about three points more depression, roughly one point three more anxiety, two points more stress, and a zero point fifteen bump in loneliness. That held even when bivariate comparisons by gender alone were muted, suggesting that the disparity emerged in the mix of stressors and supports during lockdown rather than in baseline differences. Now, a word on how firm these links are. The change models explained a meaningful slice of variance—about forty-two percent for depression, forty-eight percent for anxiety, forty percent for stress, and twenty-nine percent for loneliness—but they were exploratory, not pre-registered hypothesis tests. The authors used backward selection to handle a high number of candidate predictors in a sample that, after listwise deletion, ranged between about one hundred forty-four and one hundred seventy-eight, depending on the model. They bootstrapped tie-survival estimates and used standard t-tests for within- and between-cohort comparisons. Still, this is observational work under crisis conditions. Multiple testing and cohort differences limit causal claims. Think of the patterns as well-supported contours rather than chiselled certainties. Step back, and a coherent picture comes into view. Under lockdown, the light, everyday scaffolding of student life—chats before class, problem sets worked side by side—thinned. More people studied alone. Digital communication stepped in, often redundantly across messaging and calls, and many students remained well integrated. But the average number of interaction and co-study ties dropped, and ties that lacked multiple kinds of connection snapped at far higher rates. Inside the same students, depression, stress, anxiety, and loneliness rose. Those elevations were steeper for people who worried about loved ones and their futures, for those with fewer or less supportive personal ties, and for those living solo. In this setting, women's trajectories worsened more than men's, even adjusting for the social and stressor context. There's one last, human detail that echoes through the data. Some stressors went down. Fear of missing out fell. Perceived competition eased. With the race paused, some pressures receded. But that quiet did not compensate for the loss of casual company and collaborative rhythm. The signal that stayed strongest was simple: multiplex ties—relationships that carry friendship, support, and shared work—held people in place. When those are present, they survive shocks. When they're absent, everything feels shakier. If you run a university or you're just trying to make remote life kinder, the implications are practical. Keep the informal channels alive. Office hours that feel like a doorstep chat. Digital study groups that meet for the work and stay for the banter. Peer mentoring that bridges programs and years. And make it easy to find real emotional support, not just more Zoom. The data point to the students who need extra care when campus goes quiet: those with few direct contacts, low perceived support, thin personal networks, or a door that shuts on an empty room. Help them build multiplex ties. As Elmer, Mepham, and Stadtfeld show, the story of lockdown on campus isn't just one of fewer ties or higher scores on a scale. It's a story about which connections keep us steady when the ground moves, and how quickly the everyday fibers of social life can unravel or be rewoven when we're forced apart.

Picture a university that goes quiet almost overnight. Lecture halls empty. Cafeterias close.

Students head back to their apartments and dorm rooms, and campus life compresses into video calls and text threads. That was Switzerland in early April 2020, about two weeks into nationwide social distancing and a full university lockdown. And in that sudden stillness, a simple, hard question came into focus: when your social world thins, what happens to your mind?

Elmer, Mepham, and Stadtfeld set up a rare natural experiment to watch this unfold. They had been following Swiss undergraduates since 2018, so when the lockdown hit they could compare students to their own pre-pandemic selves and, just as important, to an earlier cohort that never experienced the crisis that spring. Two lenses, same shock.

They looked at within-person change from September 2019 to April 2020, and a between-cohort contrast: April 2020 under lockdown versus April 2019 in normal times. The focus was tight, examining four kinds of ties among classmates—pleasant interaction, friendship, co-studying, and support, both informational and emotional. They also assessed four mental health outcomes using standard scales: depressive symptoms on a scale from zero to sixty, anxiety on a scale from zero to twenty-one, perceived stress on a scale from zero to forty, and loneliness on a scale from one to four.

In April 2020, they also asked students to list important people in their personal networks, how often and how they communicated, and whether those people offered emotional support. They layered in COVID-specific worries, including concerns about health, family, money, and the future, along with a checklist of isolation experiences.

If you're wondering how they turned all that into answers, here's the high-level map. For the social networks, they counted how many classmates each student named in each category, and what share named no one at all—an "out-isolate." For mental health, they compared each person's spring score to their own fall baseline. For the between-cohort comparisons, they matched the lockdown group to that earlier, unaffected year and tested mean differences.

Because they had many potential predictors of mental health change, they used an exploratory regression with backward selection. This means they started with a full model and pruned it until the remaining variables cleared a liberal p-value threshold to identify the strongest associations. That's not a causal design; it's a way to make sense of a lot of moving pieces in a small-to-moderate sample.

So what actually changed when the campus shut down? The short version is that everyday interaction and study partnerships shrank, while friendships and support were more resilient. Within the same students from fall to lockdown, the number of pleasant interaction partners dropped by about zero point eighty ties, and co-study partners fell by zero point forty-six.

Those aren't dramatic numbers individually, but across a cohort they add up. The share of students with no study partner at all jumped from twenty-two point six percent to thirty-nine point two percent—nearly doubling the odds of being academically on your own. Friendship nominations held steady, and support ties ticked up slightly, suggesting that the core emotional scaffolding didn't collapse even as casual and collaborative ties thinned out.

Looking across years told a similar story. Comparing April 2020 to April 2019, the lockdown cohort named fewer partners in interaction and co-studying, with small to moderate effect sizes, while the proportion with zero ties didn't differ consistently across networks. The texture of connection changed too.

Face-to-face contact plunged, and students leaned hard on their phones and laptops. In April 2020, they still reported around five point six pleasant interaction ties, but most of that ran through digital channels—roughly three messaging ties and about three call ties—and the same people often appeared across modes. About three in four in-person ties also existed digitally. In other words, communication didn't stop; it rerouted.

One way to see the difference between a casual nod and a lifeline is to watch which ties survived. The team tracked specific pairs from September to April and asked, is this tie still there? For general interaction, two-thirds survived.

But that average hides a striking gradient: if a tie stood alone with no overlap with friendship, support, or co-studying, its survival was about forty-five percent. If it overlapped with two to four other dimensions, survival jumped to the mid-eighties. Co-studying was even more fragile.

A co-study tie with no other connection limped along just six percent of the time; one embedded in all four dimensions survived fifty-seven percent. On average, fewer than half of co-study ties made it. Multiplex ties, the ones that wear several hats, were the ones that endured.

As the social fabric stretched, mental health frayed. Within the same individuals, depressive symptoms rose by four point forty-four points on that zero to sixty scale. Stress rose by two point sixty-seven on zero to forty.

Anxiety crept up by zero point sixty on zero to twenty-one. Loneliness ticked up by zero point thirteen on a one to four scale. Those are modest to moderate shifts depending on the dimension, but they're consistent and statistically solid.

Between cohorts, spring 2020 and spring 2019 didn't differ significantly on those mental health averages, which tells you cohort composition and timing matter; it doesn't erase the within-person deterioration observed in the lockdown group.

What pushed those changes? The COVID-specific stressors weren't abstract. Students who worried more about family and friends saw larger increases in depression and stress—roughly one point five points more depression and one more point of stress per step up in family or friend worry.

Worrying about one's future career tracked with higher anxiety and higher stress too, with the stress impact being about one point thirty-six points per step. And the standout signal was a simple, human one: people who felt that problems they usually keep in check were bubbling up showed bigger jumps across the board—about four points more depression, one and a half more anxiety, and nearly one more point of stress. You can hear the mechanism in that phrasing: the external shock turned up the volume on the internal noise.

Isolation in the networks mattered as well, but not always in the way you might expect. Being the kind of person who named no one in the general interaction network predicted higher anxiety, as did being isolated in the informational support network; those coefficients were on the order of two points of additional anxiety. Feeling cut out of the day-to-day and lacking someone to turn to for advice are both jangling states in a crisis.

A curious nuance showed up with friendship isolates: students who named no classmates as friends reported slightly lower increases in depression but higher increases in loneliness. One way to reconcile that is to remember that the classmate network is only part of a student's social world; some may have leaned more on partners, family, or old friends outside the cohort when campus ties went dormant.

Personal networks helped explain who struggled and who coped. Smaller personal networks in April predicted larger increases in loneliness. Living alone during lockdown aligned with higher increases in depression, on the order of three points.

The average level of emotional support in one's personal network—do the people you talk to actually comfort you?—was linked to better outcomes: more support predicted lower increases in depression by a bit more than one point and lower increases in loneliness by a small but reliable amount. Not just contact, in other words, but caring contact.

Gender cut across these patterns. After accounting for social integration and COVID-related stressors, female students showed larger increases in all four mental health measures—about three points more depression, roughly one point three more anxiety, two points more stress, and a zero point fifteen bump in loneliness. That held even when bivariate comparisons by gender alone were muted, suggesting that the disparity emerged in the mix of stressors and supports during lockdown rather than in baseline differences.

Now, a word on how firm these links are. The change models explained a meaningful slice of variance—about forty-two percent for depression, forty-eight percent for anxiety, forty percent for stress, and twenty-nine percent for loneliness—but they were exploratory, not pre-registered hypothesis tests. The authors used backward selection to handle a high number of candidate predictors in a sample that, after listwise deletion, ranged between about one hundred forty-four and one hundred seventy-eight, depending on the model.

They bootstrapped tie-survival estimates and used standard t-tests for within- and between-cohort comparisons. Still, this is observational work under crisis conditions. Multiple testing and cohort differences limit causal claims.

Think of the patterns as well-supported contours rather than chiselled certainties.

Step back, and a coherent picture comes into view. Under lockdown, the light, everyday scaffolding of student life—chats before class, problem sets worked side by side—thinned. More people studied alone.

Digital communication stepped in, often redundantly across messaging and calls, and many students remained well integrated. But the average number of interaction and co-study ties dropped, and ties that lacked multiple kinds of connection snapped at far higher rates. Inside the same students, depression, stress, anxiety, and loneliness rose.

Those elevations were steeper for people who worried about loved ones and their futures, for those with fewer or less supportive personal ties, and for those living solo. In this setting, women's trajectories worsened more than men's, even adjusting for the social and stressor context.

There's one last, human detail that echoes through the data. Some stressors went down. Fear of missing out fell.

Perceived competition eased. With the race paused, some pressures receded. But that quiet did not compensate for the loss of casual company and collaborative rhythm.

The signal that stayed strongest was simple: multiplex ties—relationships that carry friendship, support, and shared work—held people in place. When those are present, they survive shocks. When they're absent, everything feels shakier.

If you run a university or you're just trying to make remote life kinder, the implications are practical. Keep the informal channels alive. Office hours that feel like a doorstep chat.

Digital study groups that meet for the work and stay for the banter. Peer mentoring that bridges programs and years. And make it easy to find real emotional support, not just more Zoom.

The data point to the students who need extra care when campus goes quiet: those with few direct contacts, low perceived support, thin personal networks, or a door that shuts on an empty room. Help them build multiplex ties.

As Elmer, Mepham, and Stadtfeld show, the story of lockdown on campus isn't just one of fewer ties or higher scores on a scale. It's a story about which connections keep us steady when the ground moves, and how quickly the everyday fibers of social life can unravel or be rewoven when we're forced apart.

Related lectures