Psychosocial well-being before, during and after the COVID-19 pandemic: a nationwide study of more than half a million Norwegian adolescents

Research indicates that the COVID-19 pandemic has had adverse effects on adolescent psychosocial well-being, but little is known as to whether these changes persisted after pandemic-related restrictions were lifted. Addressing this issue, we examined whether declines in well-being remained after the pandemic and whether adverse changes during and after the pandemic varied according to parental socioeconomic status and adolescents’ sex and age. We used data from more than 500,000 Norwegian adolescents, assessed annually between 2014 and 2022. Multilevel analyses showed that after Norway lifted all restrictions, adolescents reported lower satisfaction with peer relationships and less-optimistic future life expectations, but less loneliness compared with long-term trends. Moderation analyses revealed that girls, younger adolescents and adolescents from poor families and with low parental education suffered more adverse changes both during and after the pandemic across several domains. These findings indicate a need for policy interventions targeting social inequalities in adolescent well-being. The severe restrictions put in place to mitigate the spread of the coronavirus dramatically changed adolescents’ lives during the COVID-19 pandemic. Several large-scale research efforts have found that adolescents’ psychosocial well-being deteriorated during the pandemic 1–5 and that this decline was most marked among adolescents from families with a low socioeconomic status 1,2 . However, less is known about adolescents’ situations after restrictions had been lifted and life had seemingly returned to normal. In particular, there is little research on whether adolescents’ psychosocial well-being quickly returned to pre-pandemic levels after restrictions were lifted or whether negative changes persisted even after the pandemic had been declared overcome. In this study, we address this issue by using nationwide, large-scale, repeated cross-sectional data from more than 500,000 adolescents in Norway, who were assessed on key psychosocial aspects before the COVID-19 outbreak, during the pandemic and after all restrictions were lifted in the spring of 2022. Our main objective is to examine whether psychosocial changes related to the pandemic were reversed when restrictions were lifted. Moreover, to respond to reports showing an increase in social inequalities in adolescents’ well-being during the pandemic 1,6,7 , we aim to examine whether social gradients according to parental socioeconomic status, adolescents’ sex and age returned to pre-pandemic levels or whether exacerbated inequalities remained after the restrictions were lifted. Furthermore, we also examine the role of regional socioeconomic factors on changes in adolescents’ psychosocial well-being during the pandemic and in its aftermath

The severe restrictions put in place to mitigate the spread of the coronavirus dramatically changed adolescents' lives during the COVID-19 pandemic. Several large-scale research efforts have found that adolescents' psychosocial well-being deteriorated during the pandemic [1][2][3][4][5] and that this decline was most marked among adolescents from families with a low socioeconomic status 1,2 . However, less is known about adolescents' situations after restrictions had been lifted and life had seemingly returned to normal. In particular, there is little research on whether adolescents' psychosocial well-being quickly returned to pre-pandemic levels after restrictions were lifted or whether negative changes persisted even after the pandemic had been declared overcome. In this study, we address this issue by using nationwide, large-scale, repeated cross-sectional data from more than 500,000 adolescents in Norway, who were assessed on key psychosocial aspects before the COVID-19 outbreak, during the pandemic and after all restrictions were lifted in the spring of 2022. Our main objective is to examine whether psychosocial changes related to the pandemic were reversed when restrictions were lifted. Moreover, to respond to reports showing an increase in social inequalities in adolescents' well-being during the pandemic 1,6,7 , we aim to examine whether social gradients according to parental socioeconomic status, adolescents' sex and age returned to pre-pandemic levels or whether exacerbated inequalities remained after the restrictions were lifted. Furthermore, we also examine the role of regional socioeconomic factors on changes in adolescents' psychosocial well-being during the pandemic and in its aftermath.
Economic research shows that economic crises have persistent negative effects on young people across a range of future life outcomes, including earnings, physical health, life satisfaction, drug use and Article https://doi.org/10.1038/s44220-023-00088-y Moreover, the severe pandemic-related restrictions, including school closures and the suspension of organized activities, led to interruptions in adolescents' academic and social lives. Disadvantaged families may have found these interruptions the most difficult to compensate for 22 , and a negative influence on adolescents from these backgrounds may have persisted after the restrictions were lifted.
The pandemic may also have had differential effects according to regional socioeconomic characteristics. Those in marginalized areas with higher levels of social inequality may have suffered more from consequences of the pandemic. This notion is supported by analyses showing that residents in deprived areas in the United Kingdom were more economically vulnerable than others to the effects of the pandemic 23 . We follow this line of research by examining whether adolescents from lower socioeconomic backgrounds experienced more negative changes during and after the pandemic than other adolescents. We also examine whether living in municipalities characterized by limited socioeconomic resources and high levels of inequality was associated with increased adverse changes, both during and after the pandemic.
In Norway, restrictions aimed at mitigating the COVID-19 pandemic were similar to those in other European countries, with a major lockdown introduced on 12 March 2020, including the closure of schools, the suspension of most organized leisure activities and the implementation of strict physical distancing measures. Junior and senior high schools reopened on 11 May 2020, but with restrictions in place such as smaller class sizes, physical distancing and partial digital schooling. Other regulatory measures were maintained and continuously adapted to account for the changing levels of infection risk. Most restrictions were removed on 1 February 2022, with the remaining restrictions lifted on 12 February, including all physical distancing measures, face-mask requirements and regulatory procedures for the isolation and quarantine of those with COVID-19 infections. On 12 February 2022, the Norwegian prime minister announced that "the COVID-19 pandemic no longer is a great threat to most of us" and that "we can return to normal everyday life" 24 .
Cross-national research examining adolescent well-being shows that Norwegian adolescents exhibit low levels of psychosomatic complaints and high levels of life satisfaction compared with adolescents in other Western countries 25 . Likewise, in the domain of substance use, Norway has the lowest reported rates of current and lifetime use of cigarettes and alcohol among adolescents compared with other European countries 26 . However, in the domain of sedentary behavior, research reports slightly higher rates of insufficient physical activity among teenagers in Norway compared with the average in high-income Western countries 27 .
To summarize, research has shown that adolescents reported declines in psychosocial well-being across several domains during the COVID-19 pandemic. Furthermore, disadvantaged adolescents exhibited greater adverse changes during the pandemic than other adolescents. However, we know little as to whether such adverse changes and exacerbated social inequalities remained beyond the period of the pandemic. In line with economic research on scarring effects, we hypothesized that adverse changes in psychosocial well-being during the pandemic may have lasted even after all pandemic-related restrictions were lifted. Likewise, we expected that increased social inequalities according to parental socioeconomic status, sex and adolescents' age would also remain beyond the pandemic period.
To address these issues, we used nationwide Norwegian survey data from 2014 to 2022 (N = 576,082; ages 13-18 yr; 51% girls) to examine psychosocial outcomes in adolescents before the pandemic, during the pandemic in 2021 and after all restrictions in Norway were lifted in the spring of 2022. We applied multilevel societal growth models to examine changes in levels of psychosocial well-being across different life domains relative to long-term trends. Moreover, we examined whether social inequalities in psychosocial well-being were reduced in the spring of 2022, after they had increased considerably during the first year of the pandemic 1 . We further investigated whether more criminal behavior [8][9][10][11] . These prolonged negative effects, which are difficult to explain by short-lived labor-market shocks, have been termed 'scarring effects'. They refer to situations in which those exposed to a recession at a young age experience a lasting disadvantage in later life, often years after the recession has ended. Although economic research on scarring effects has focused mainly on young adults entering the labor market, scarring mechanisms may also apply to younger age groups during the COVID-19 pandemic due to widespread school closures and negative effects on family finances 12 . Econometric modeling indicates that COVID-19 school closures may indeed have had a scarring effect on welfare and lifetime earnings for young people 13 . While a proper evaluation of scarring effects will be feasible only several years after the pandemic, the observed negative changes in adolescents' well-being immediately after the lifting of restrictions can serve as an early indication of such long-term effects. However, to date, there are no large-scale studies available that compare adolescents' psychosocial well-being before, during and after the pandemic.
In this Article, we examine early indications of COVID-19 scarring effects in the domains of social relationships, mental health, substance use, physical activity, screen time and future life expectations. Concerning social relationships, theoretical accounts have posited that physical distancing and school closures may have longstanding negative effects on family life and social relationships 14 . However, a recent Norwegian study from our research group did not find any substantial declines in satisfaction with relationships with parents and peers during the pandemic 1 , whereas a Dutch study showed a slight increase in peer relationship problems at the beginning of the pandemic but no subsequent changes in later pandemic phases 15 . Concerning mental health, several studies have found increasing rates of depressive symptoms during the pandemic [1][2][3]5 . Findings are mixed when examining related constructs, such as loneliness; one study found a significant increase in adolescent loneliness during the pandemic 16 , whereas another did not 1 .
A recent review, based on 49 studies, revealed a general decline in the prevalence of adolescent substance use during the pandemic, with reductions in the use of alcohol, tobacco and cannabis 17 . This decline has been explained by the fact that restrictions during the pandemic led to fewer opportunities to use substances 17 . Concerning physical activity and sedentary behavior, a systematic review has shown reduced physical activity and increased screen time among children and adolescents 18 . Moreover, a Dutch study found that physical activity decreased and screen time increased among children also after school lockdowns 19 . These findings may indicate that child and adolescent activity patterns underwent prolonged changes even after the most severe restrictions were lifted.
Finally, the pandemic may have had lasting effects on adolescents' expectations of future opportunities in life. Results from our research group indicated less-optimistic future life expectations during the pandemic than before 1 . These expectations may also remain low afterward, given that the pandemic and its consequences may have disrupted adolescents' beliefs in a safe and controllable world. In summary, studies have indicated adverse changes across a variety of psychosocial factors among adolescents, besides the domain of substance use; however, little is known as to whether such changes remained after pandemic restrictions were lifted and adolescents' lives seemingly returned to normal.
Recent research has revealed that the pandemic affected disadvantaged adolescents disproportionately across a variety of life domains and thereby widened social disparities during the time of the pandemic 1 . Research on scarring effects has shown that even years after a period of downturn, economic recessions have had a disproportionately negative effect on socioeconomically vulnerable groups across a variety of life domains, including income, health, life satisfaction, obesity and substance use 8,10,20 . It seems plausible to assume that the scarring effects of the pandemic are also unevenly distributed. For example, the pandemic was accompanied by an economic downturn that disproportionately affected families with low socioeconomic resources 21 .
Article https://doi.org/10.1038/s44220-023-00088-y adverse changes during and after the pandemic were observed in areas in Norway with greater socioeconomic challenges.

Results
In total, 576,082 adolescents from 248 municipalities in Norway participated in the study. The number of participants was the lowest in 2020 (Table 1) because data collection was discontinued when schools were closed on 12 March. At the time, many schools had not yet started data collection. As a result, data from 2020 were collected before pandemicrelated restrictions were introduced. Descriptive statistics in Table 1 show generally high satisfaction with peer and parental relationships, with a mean percentage of a maximum possible (POMP) score over 80 across all years, indicating an average satisfaction above 80% of the maximum possible scores. Adolescents scored toward the lower end of the POMP scale for depressive symptoms, loneliness and substance use. There were relatively high scores for physical activity, screen time and expecting a happy future, with averages of 72%, 66% and 71%, respectively, across all years. On average, 18% of participants reported that neither parent had completed at least some form of tertiary education (hereafter referred to as low parental educational attainment), and 5% perceived their family to be poor. Table 2 presents intercorrelations of all study variables.

Changes in well-being during and after the pandemic
In a first step, we estimated societal growth curves for all nine indicators of psychosocial well-being from 2014 to 2022, with the additional inclusion of dummy variables to model the effect of COVID-19 during the pandemic (2021) and after the pandemic when all restrictions were lifted (2022). Regression estimates of the dummy in 2021 provide information about how psychosocial well-being during the pandemic deviated from the expected societal growth-curve trajectory. Regression estimates of the dummy in 2022 provide information about deviation from expected trends after the pandemic. Table 3 and Fig. 1 show the main results of these analyses (see Supplementary Tables 1-5 for details).
Satisfaction with relationships with peers decreased somewhat across time (as shown by a significant negative linear time trend, β 1 ), whereas satisfaction with parental relationships increased significantly across time (as shown by a significantly positive linear trend). In line with our earlier work 1 , neither satisfaction with relationships with peers nor satisfaction with relationships with parents deviated significantly from the trend line during the pandemic in 2021 (β 3 was not significant). Thus, the results did not indicate adverse changes in social relationships during the pandemic in 2021. However, satisfaction with peer relationships was significantly lower than expected in 2022, after all restrictions were lifted (as indicated by a significant regression estimate for the post-pandemic dummy, β 4 ), whereas no post-pandemic change was observed for parental relationships.
Examination of mental health revealed higher-than-expected scores on depressive symptoms in 2021, but not in 2022. By contrast, no significant changes were observed for loneliness in 2021, whereas loneliness scores were significantly lower than predicted by growth curves in the aftermath of the pandemic in 2022.
Concerning substance use, both smoking and alcohol use declined significantly during the pandemic in 2021, but these reductions did not persist in the post-pandemic year of 2022. Physical activity did not deviate from the trend line during or after the pandemic, whereas screen time increased during the pandemic, but there was no postpandemic change. Finally, adolescents were less optimistic regarding their prospects for a happy future compared with the trend line in the periods both during and after the pandemic.

Social inequality during and after the pandemic
Next, we examined whether disproportionately greater changes were observed among adolescents from disadvantaged backgrounds during and after the pandemic. To do so, we examined whether the estimated pandemic and post-pandemic effects were moderated by low parental educational attainment and perceived poverty while controlling for age and sex. Figure 2 shows the results of analyses with parental education (details of all moderation analyses in this section can be found in Supplementary Tables 6-10). During the pandemic in 2021, estimated satisfaction with peer relationships was on average 1.24 POMP scores lower than expected according to the overall trend line among adolescents with low parental education. Adolescents with higher parental educational attainment showed less deviation from the overall trend line (−0.61). The orange bar between the estimates indicates the difference between adolescents with high and low parental education to be statistically significant. Furthermore, Fig. 2 shows that adolescents with parents of low education experienced significantly more adverse changes during the pandemic regarding parental relationships, depressive symptoms and physical activity. We also found significant differences for screen time; however, screen time increased significantly less among adolescents with low parental education than among other adolescents. When examining parental education as a moderator in 2022, we found significantly more adverse changes for depressive symptoms and loneliness among adolescents whose parents had low educational attainment, whereas satisfaction with peer and parental relationships did not differ significantly according to parental education, even though trends were comparable to those observed in 2021. Moreover, screen time decreased significantly more among adolescents with low parental education.
Results show even clearer patterns of social disadvantage when including perceived family poverty as a moderator. As displayed in Fig. 2, in 2021, all included domains except smoking behavior, alcohol use and screen time showed significantly greater adverse changes for adolescents who perceived their family as poor compared with other adolescents. The same pattern of results was obtained in 2022.
We also examined whether changes during and after the pandemic differed according to participants' sex and age (Fig. 3). Girls seemed to experience significantly more adverse changes during the pandemic in 2021 across a variety of domains, including peer and parental relationships, depressive symptoms, smoking, alcohol use, physical activity and screen time. By contrast, loneliness increased significantly more among boys than among girls in 2021. Similar results were observed in 2022, where most domains showed more adverse changes for girls, but loneliness was significantly more reduced among girls than among boys. Figure 3 shows more adverse changes in 2021 for younger than for older adolescents across all domains but alcohol use. Also in 2022, several domains showed a greater estimated effect of the pandemic for younger adolescents; however, the moderation effect of age disappeared for parental relationships, smoking and screen time. For alcohol use, the pattern was even reversed, with younger adolescents experiencing a greater reduction in alcohol use than older adolescents.
Adolescents' future life expectations was the only domain with statistically significant deteriorations both during and after the pandemic. Therefore, we conducted a series of post hoc moderation analyses to investigate the role of future expectations for pandemic and post-pandemic changes in psychosocial well-being. We estimated whether pandemic and post-pandemic effects were moderated by expectations of a happy future while controlling for age, sex, parental education and family poverty (Supplementary Fig. 1 and Supplementary Table 11). The results showed that adolescents with optimistic expectations reported a lesser decrease in satisfaction with peer relations, a lower increase in depressive symptoms and less of a reduction in physical activity than their less-optimistic peers during the pandemic. Conversely, optimistic adolescents reported a lesser reduction in smoking over the same period. In the aftermath of the pandemic, optimistic adolescents indicated a smaller decrease in Article https://doi.org/10.1038/s44220-023-00088-y Data from 2020 were collected before the national lockdown implemented following the COVID-19 outbreak (before 12 March 2020). Data from 2022 were collected after all pandemic-related restrictions were lifted (after 28 February 2022). Continuous measures (except for age) are scaled as POMP. POMP scores can be interpreted as the percentage of the maximum possible scores achievable on the scale, with a minimum score of 0 and a maximum score of 100. M, mean; s.d., standard deviation.
Article https://doi.org/10.1038/s44220-023-00088-y satisfaction with peer relations and a lower increase in depressive symptoms. However, after the pandemic, future life optimism was related to a lower reduction in reported feelings of loneliness, smoking, alcohol consumption and screen time. Finally, we examined whether the estimated pandemic and postpandemic effects differed according to the socioeconomic context of the municipality (for details, see Supplementary Tables 12-16). In 2021, during the pandemic, adolescents from municipalities with a larger proportion of low-income households experienced a significantly lower reduction in satisfaction with peer relations, smoking and expectations of a happy future. In addition, adolescents from municipalities with fewer university graduates showed a significantly lower increase in depressive symptoms. No significant moderation effects in 2021 were found for the Gini coefficient and the share of migrant population. The results thus indicate that, with the exception of smoking, the well-being of adolescents living in socioeconomically disadvantaged municipalities was affected by the pandemic to a lesser extent compared with that of adolescents living in other municipalities.
The results for 2022 also did not provide unambiguous support for the notion that adolescents in socioeconomically disadvantaged municipalities experienced more adverse post-pandemic changes: parental relationships deteriorated to a greater extent in municipalities  Article https://doi.org/10.1038/s44220-023-00088-y with a high share of migrants after the pandemic, but alcohol use also declined more strongly in such municipalities. No other municipalitylevel moderation effects were significant in 2022.

Discussion
Results from this nationwide study among more than 500,000 Norwegian adolescents indicated that in 2022, satisfaction with peer relationships and future life expectations diverged negatively from long-term trends, thereby pointing to prolonged adverse changes in both these areas of adolescents' lives in the aftermath of the COVID-19 pandemic. By contrast, adolescents reported feeling less lonely during the post-pandemic year, 2022, compared with expectations derived from long-term loneliness trends. Taken together, our results do not indicate a general worsening of psychosocial well-being among Norwegian adolescents across a wide range of domains in the period following the pandemic. Article https://doi.org/10.1038/s44220-023-00088-y However, moderation analyses showed that adolescents from poor families, as well as those with low parental educational attainment, experienced more adverse changes during and after the pandemic across a variety of domains. Thus, despite the lack of general negative changes at the population level, there is concerning evidence that exacerbated socioeconomic inequalities in well-being during the pandemic have remained in its aftermath. The estimated adverse effects both during and after the pandemic had notable sex and age gradients, with disproportionately harsher changes among girls and younger adolescents. Evidence from post hoc moderation analyses further indicated that adolescents with an optimistic outlook on their future evinced fewer changes in psychosocial well-being during and after the pandemic. No coherent pattern of change was found regarding the role of the municipality socioeconomic context in moderating adverse pandemic and post-pandemic changes in psychosocial wellbeing among Norwegian adolescents.

Changes in well-being during and after the pandemic
This study does not support the notion that the pandemic has had farreaching scarring effects spanning across diverse domains of adolescent life. In line with previous research 1,2 , our study found an increase in depressive symptoms and screen time during the pandemic. However, these adverse changes, documented during the pandemic year, did not persist consistently in the period shortly after the pandemic. A notable exception is that of adolescents' expectations of a happy future, which dropped significantly below the long-term trend both during and after the pandemic. Research is scarce on how major crises such as the COVID-19 pandemic influence optimism, and findings on negative life events as a source of reduced optimism are mixed 28,29 . Our study provides indications that global crises, such as the recent pandemic, may result in a less-positive evaluation of adolescents' future life prospects. Such findings are concerning since research has consistently   Orange bars indicate statistically significant differences between groups as provided by interaction analyses (P < 0.05, two-sided, unadjusted for multiple tests). The red dots represent estimated conditional effects for adolescents with low parental education (a) and for adolescents who perceived their family to be poor (b). The blue dots represent estimates for adolescents with higher parental education (a) and for adolescents who did not perceive their family to be poor (b Article https://doi.org/10.1038/s44220-023-00088-y shown that optimistic future life expectations are associated with positive outcomes across a variety of different domains, including mental and somatic health, coping behaviors and social relationships 30,31 .
In the domain of substance use, studies showed a reduction in the use of alcohol and tobacco during the pandemic 1,2,15 . Nevertheless, our present results do not indicate that these reductions continued after all pandemic restrictions had been lifted. The findings indicate that the pandemic may not have resulted in lasting changes in patterns of substance use. Instead, after the lifting of physical distancing measures and the resumption of social events, adolescents may have resumed their pre-pandemic patterns of substance use, taking advantage of opportunities to engage in social contact with peers outside of parental supervision, where substance use is common.
In the domain of social life, our findings are somewhat inconclusive. On the one hand, we found that adolescents reported significantly less loneliness after the pandemic, which may indicate that the 'return to normal' may have alleviated social isolation. For some adolescents, the pandemic may even have provided new perspectives on the importance of social relationships, making them more appreciative of regained social contacts 32 and resulting in less post-pandemic loneliness. On the other hand, however, adolescents reported lesssatisfying peer relationships after the pandemic, which may be an indication that upholding social relationships in the same way as before the pandemic may have been challenging. Further research is needed to examine the long-term effects of the pandemic on social aspects of adolescents' lives, in particular, when data from several years following the pandemic become available. In conclusion, our results on a variety of domains of life do not provide clear indications of a pandemic-related deterioration of psychosocial well-being among the whole population of Norwegian adolescents after restrictions were lifted.

Social inequality during and after the pandemic
The results of this study indicate a coherent pattern of intensification of social inequalities across domains during the pandemic, which   Even more distinctive social inequalities were observed when examining perceived family poverty as a marker of socioeconomic status. Social gradients both during and after the pandemic showed disadvantages for adolescents who perceived their families as poor in all domains except screen time, physical activity and alcohol use. Our results are in line with similar studies investigating economic recessions, which have repeatedly demonstrated that recessions have a disproportionately adverse effect on the most vulnerable social groups across several life domains, including health, life satisfaction and substance use 8,10,20 . According to our findings, even after the pandemic restrictions were lifted, the widening social inequality observed during the pandemic was not alleviated. This finding is of specific importance as it indicates that increasing social inequalities in psychosocial wellbeing during the pandemic may not be attributed completely to the short-term effects of suddenly implemented restrictions. Instead, persisting inequalities may be caused by continuing, disproportionately large socioeconomic consequences of the pandemic (for example, losses in income and work intensity, which particularly affected disadvantaged families) 21 . Moreover, it may have been more difficult for vulnerable families to compensate for interruptions to adolescents' academic and social lives 22 , and these families may have had fewer resources to support a return to 'normal life'.
Overall, the estimated pandemic effects were stronger for girls than for boys during the pandemic, and this pattern remained stable in 2022. These disproportionately negative changes among girls are consistent with other studies examining pandemic effects on mental health and substance use 2,7 . However, our study provides evidence that the disproportionately negative changes also seem to extend beyond the pandemic period. As previously suggested 1 , one potential explanation for this sex difference is girls' greater general reactivity to stress exposure. However, this mechanism cannot account for sex differences that persist beyond the pandemic period. More research is needed to examine sex-specific, long-term impacts of the pandemic.
Our study also shows that younger adolescents were more adversely affected than older adolescents, both during and to some extent after the pandemic. It is possible that the daily life of younger adolescents may depend to a larger degree on external structures related to school and organized recreational activities. Their well-being may thus be more susceptible to interruptions in the normal functioning of those structures. Older adolescents may have greater personal autonomy, which allows them to adapt more flexibly to changes to the organized aspects of their lives. Still, future research is needed to provide robust information on the mechanisms that may explain the age differences in pandemic susceptibility.
Our post hoc moderation analyses indicated that an optimistic outlook on the future buffers adverse changes in mental health and sedentary behaviors during the pandemic and in mental health and social relationships in its aftermath. However, optimism was also associated with smaller positive changes in domains in which adolescents generally experienced beneficial developments, such as a reduction in smoking during the pandemic and reduced loneliness, substance use and screen time in the period following the removal of pandemic restrictions. This finding indicates that optimistic future life expectations may strengthen adolescent resilience toward both positive and negative changes in the face of crises. According to cognitive theories, optimistic people believe that negative events are of a rather temporary nature and are unlikely to have a lasting or global impact on personal lives 33 . Thus, the pandemic may influence optimistic adolescents to a lesser degree because they tend to perceive the pandemic as being less impactful for their life situation.
Our study finds no evidence for disproportionately negative changes during or after the pandemic in municipalities characterized by low socioeconomic resources, which stands in contrast to our findings of interaction effects at the individual level. This finding may be explained by the relatively large degree of homogeneity among Norwegian municipalities.
There are several implications of our findings for public health policy and future research. First, the persistent drop in expectations about the future is concerning. Our study also shows that optimism seems to be an important stabilizing factor against changes during and after the pandemic. While strengthening of adolescents' optimism for the future might be a promising way of enhancing their resilience in times of crises, more research is needed to uncover the mechanisms that render optimistic adolescents more resistant to positive developments in the aftermath of the crisis. Second, we identified subgroups of adolescents who were particularly vulnerable to adverse changes during and after the pandemic. Social policy measures should, therefore, focus on supporting adolescents from low socioeconomic backgrounds and their families. Moreover, particular emphasis should be placed on psychosocial interventions to alleviate adverse changes in well-being among girls, who seem to have suffered more during and after the pandemic.

Strengths and limitations
The main strength of this study is its large, nationwide sample with yearly assessments across almost a decade, providing a unique possibility to estimate changes in adolescent psychosocial well-being across a variety of domains before, during and after the pandemic while accounting for time trends. However, our results should be interpreted in light of several limitations. First, the pandemic-related restrictions were lifted two weeks before the first participants responded to the survey in the 2022 data collection wave. Our study, therefore, provides only indications concerning sustained adverse changes relatively shortly after all restrictions were lifted. Moreover, some of the measures of psychosocial well-being that we used in previous research 1 could not be used to examine post-pandemic changes because they asked about behaviors within a time frame that partly included periods in which restrictions had not yet been lifted. Second, our approach for estimating changes during and after the pandemic, by accounting for time trends, implies that our measures would have continued to show the same pattern of development in 2021 and 2022 as in previous years, given the counterfactual condition of not experiencing a pandemic. Such assumptions may not necessarily be valid, particularly for behaviors in domains where considerable changes had been identified before the pandemic, such as screen time. Third, several of our measures were based on single items, such as relationships with parents and peers and future life expectations. A more comprehensive assessment of these constructs would have been valuable. Fourth, adolescent psychosocial well-being in 2022 may have been influenced by sudden unexpected events other than the lifting of all pandemic restrictions. For example, the geopolitical threat of the Russian invasion of Ukraine, which began in February 2022, may have negatively influenced adolescents' future life expectations as Norway directly borders Russia. Finally, the results have to be interpreted in the context of the Norwegian welfare state with its extensive social security provisions, which may have mitigated the negative economic consequences of the pandemic. It remains to be seen how the pandemic affected adolescents in other societal contexts.

Conclusion
In summary, this study finds that Norwegian adolescents, on average, reported few adverse changes in psychosocial well-being after all national pandemic-related restrictions were lifted. However, there are some concerning findings indicating that the social inequality that widened during the pandemic seemed to have remained after restrictions were lifted. This observation indicates that the pandemic is associated with widening social inequality across a variety of different domains of adolescents' psychosocial well-being, which endures beyond the pandemic period. Addressing the issue of widening social inequality and the provision of support for vulnerable families hit hardest during the pandemic may be a task of utmost importance for policymakers in the period after pandemic-related restrictions are lifted.

Data and participants
We used data from Ungdata, a nationwide data collection scheme with regular survey assessments of adolescents at the municipal level in Norway. The Ungdata survey is intended to comprehensively cover key aspects of adolescents' lives, including psychosocial well-being and mental health. Since 2014, all junior and senior high school students (grades 8 to 13, students aged 13 to 18) have been routinely assessed; Ungdata surveys are conducted regularly in almost all municipalities, typically every third year. During a school period, all students were invited to complete an online questionnaire, which covered domains including social relationships, mental health, substance use, leisure activities and future life expectations.
Data were collected each spring. In 2020, data collection started in January but was discontinued when the lockdown was implemented on 12 March. As a result, in most of the participating municipalities, data collection was not completed, and therefore, a considerably lower number of adolescents participated in the 2020 wave. In 2021, 210 municipalities participated, whereas 124 municipalities participated in 2022. Data collection in 2022 took place between 17 January and 3 June. Because this study aimed to examine psychosocial well-being after the pandemic, we excluded all participants from the 2022 wave who responded before 1 March. Thus in 2022, we included only participants who had responded to the survey at least two weeks after all restrictions were lifted in Norway on 12 February. Of all participating municipalities, 67 were excluded because one or several study variables were not assessed. We also excluded data from five municipalities that had not conducted any Ungdata surveys before 2021. Another 12 were omitted from the analysis because the 2022 data collection took place before 1 March. Forty-five municipalities had conducted 1 wave of data collection before 2021, and 125, 75 and 3 municipalities had conducted 2, 3 and 4 previous waves, respectively. Following listwise deletion of individual observations with missing values, data from N = 576,082 adolescents from 248 municipalities were used. Informed consent was obtained from all participants, and participants and their parents received information about the study beforehand. Moreover, participants were informed that they could skip questions that they did not wish to answer and could withdraw from the study at any time. The participants were also informed about the possibility of contacting the school health services or the Red Cross helpline if they required support after completing the Ungdata questionnaire. The study was recommended by the internal research ethics committee of the University of Oslo, Department of Psychology (reference # 13710027).

Measures
Social relationships. Peer and parental relationships were assessed by one item each, with respondents being asked about satisfaction with peer friendships and parents, respectively. Both items were measured on a five-point scale, ranging from 'very unsatisfied' to 'very satisfied'. The two items were modeled according to domain-specific, subjective well-being instruments, such as the Personal Wellbeing Index 34 , which assesses satisfaction with particular areas of life. Domain-specific satisfaction measures have been shown to be particularly useful when evaluating specific policy interventions 35 .
Mental health. Depressive symptoms were measured by the six-item Depressive Mood Inventory 36 derived from the Hopkins Symptom Checklist 37 , which assesses depressive symptoms experienced within the preceding one-week period. The four-point response scale ranges from 'affected not at all' to 'affected extremely' (Cronbach's α = 0.89). Loneliness was assessed by one item, capturing feelings of loneliness over the past week, with the same response scale. Previous research has demonstrated that single items that assess feelings of loneliness have good face validity and predictive utility 38 .

Substance use.
Smoking was assessed by an item distinguishing among adolescents who did not smoke at all (1), those who smoked less than once a week (2), those who smoked every week but not daily (3) and those who smoked daily (4). Alcohol use was assessed by a question on the frequency of drinking alcohol, ranging from never (1) to a few times (2), less than once a month (3), one to three times a month (4) and every week (5). Existing research has demonstrated a high reliability of substance-use self-reports, especially when measured with open-ended questions similar to those used in the present study 39,40 .
Physical activity and screen time. Physical activity was assessed by a question on the frequency of respondents' engagement in physical activity that made them feel 'out of breath or sweaty'. The six-point scale ranged from 'never' (1) to 'at least five times a week' (6). Among young people, the use of physical activity measures, based on a single item, has been advocated as more reliable than assessments based on more-complex and comprehensive measures 41 .
Screen time was assessed by a question about the daily amount of time spent in activities 'that involve looking at a screen' outside of school, with six response options ranging from 'no time' (1) to 'more than three hours' (7). We opted for this simple and straightforward measure because multi-item instruments capturing the use of a variety of screen-based devices or behaviors can be problematic over longer periods with rapid technological development and new forms of media consumption 42 .

Future life expectations.
We assessed future life expectations by an item asking about respondents' expectations for 'a good, happy life'. Response options included 'yes', 'no' and 'don't know', and we contrasted the group of adolescents expecting to live a good and happy life (1 if yes) with all other adolescents (0 for no and don't know). Similar measures, based on single items, have been commonly used in longitudinal research on anticipated life satisfaction, and their rank-order stability has been shown to be adequate 43,44 .
Indicators of socioeconomic status and other demographics. Low parental educational attainment was measured by a binary variable, contrasting respondents with neither parent having completed at least some form of tertiary education (1) with the remainder of their peers (0). Responses to the question on family socioeconomic situation over the past two years were used to measure perceived family poverty. Five response options ranging from 'always good' to 'always bad' were recoded to contrast those who perceived their family's economic situation over the past two years as bad (1 for the responses 'mostly bad' and 'always bad') with those who did not (0). Sex was also assessed. Due to anonymity concerns, age could be assessed only indirectly through a proxy variable, indicating the respondents' school grade. However, since grades within the Norwegian educational system are organized strictly according to birth cohorts, and grade repetition because of unsatisfactory academic performance is rather rare, school grade and age largely overlap. Article https://doi.org/10.1038/s44220-023-00088-y Municipality-level variables. We assessed socioeconomic characteristics of all municipalities included in the study by using municipalitylevel data from the Municipal Health Statistics Bank, provided by the Norwegian Institute of Public Health. Low municipality-level education was measured by the percentage of adults in the municipality who did not complete any form of tertiary education. Relative poverty levels were measured by the share of the municipality population living on incomes lower than 60% of the national median income, after tax and adjustment for household composition. Municipality socioeconomic inequality was measured by the Gini coefficient. Finally, we measured the share of the municipality population with a migrant background, defined as those whose parents and grandparents had been born outside Norway. For all municipality-level measures, values for each municipality were averaged across all available years within the 2014-2022 period.

Analyses
We harmonized the scores of all psychosocial outcome variables before the analyses by converting them into POMP scores 45 with a minimum score of 0 and maximum score of 100. POMP scores can be read as percentages, which means that they combine the advantage of intuitive interpretation with the possibility of direct comparison across measurements recorded on different scales 45 .
All estimates were obtained from multilevel regression models, known as societal growth-curve models 46 . The societal growth-curve specification is an extension of multilevel modeling for analysis of comparative longitudinal survey data, where the same cross-sectional units are observed across time in multiple waves, but at each wave, a new representative sample of the cross section is drawn from the population 47,48 . In summary, the method estimates how an aggregated individual characteristic (for example, adolescent psychosocial wellbeing) develops over time within repeatedly sampled higher-level geographical units (for example, municipalities). It can be easily extended to estimate differences in development over time for different groups of individuals within the geographical units, as well as to compare developmental trajectories over time between the higher-level geographical units.
As a baseline model, we used a slightly adapted version of the growth-curve model we developed in a previous study 1 to model time trends at the municipality level for nine indicators of psychosocial well-being in the period from 2014 to 2022. All presented models were estimated as three-level multilevel linear regressions, where individuals at the lowest level are nested within municipality-years at the second-highest level and municipalities at the third-highest level.
The overall pattern of change over time was modeled as a curvilinear trend through inclusion of both linear and quadratic terms for the number of years that had passed since the first included survey wave in 2014. In addition to the overall trend, we included binary predictors for the 2021 and 2022 waves, estimating the effects of the pandemic and post-pandemic period beyond the curvilinear development. Both variables were coded such that respondents participating in the corresponding wave (1) were compared with respondents participating in all other waves (0). More formally, the societal growth model can be expressed by the following equation: Y itj = β 0 + β 1 time tj + β 2 time tj 2 + β 3 dummy2021 tj +β 4 dummy2022 tj + v 0j + u 0tj + e itj with e itj ∼ N(0, σ e 2 ) A given psychosocial characteristic Y itj for adolescent i at survey wave t in municipality j was modeled as a function of the grand intercept across municipalities β 0 , the linear slope β 1 and the quadratic slope of the societal growth curve β 2 , together with the pandemic dummy coefficient β 3 and post-pandemic dummy coefficient β 4 , indicating the deviation in the dependent variable beyond the general trend expressed by the growth curve in the pandemic year 2021 and postpandemic year 2022, respectively. The random intercepts v 0j for the municipality level and u 0tj for municipality-year level were included and were, together with the individual-level error term e itj , assumed to be distributed normally, with a mean of 0.
Under the POMP score scaling, estimated regression coefficients for pandemic and post-pandemic effects can be interpreted in terms of percentage-point changes of the maximum possible score achievable on the scale 45 . All models included controls for parental education, sex and age to adjust for individual-level compositional differences that may have affected the societal growth curves or the estimated effect of pandemic and post-pandemic years 49 .
To examine whether specific groups of adolescents within municipalities showed disproportionately adverse changes during and after the pandemic, the baseline societal growth-curve model was extended to include cross-level interaction terms of both the pandemic and post-pandemic effect variables with the individual-level predictors of parental educational background, perceived family poverty, sex and age. Point estimates of these conditional effects for specific values of moderator variables were computed 50 and used to display the interaction effects graphically.
To address potential differences in the change in psychosocial well-being during and after the pandemic between municipalities with different socioeconomic characteristics, we expanded the baseline societal growth-curve models to include cross-level interaction terms between both pandemic and post-pandemic variables with time-invariant municipality-level predictors of low education, relative poverty, socioeconomic inequality and share of the population with a migrant background 46 .
All predictors and controls were grand-mean centered to facilitate convergence of the models 51 . We used R version 4.0.3 for all analyses. Multilevel regressions were fitted using the lme4 package for R, version 1.1.26 52 . Statistical significance of all parameter estimates from multilevel models was evaluated by P values on the basis of the t statistics using the normal distribution. Estimates with P values lower than 0.05 were considered statistically significant. The interplot package for R, version 0.2.3 53 was used to estimate conditional effects of interaction analyses.

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability
The data that support the findings of this study can be obtained from Norwegian Social Research (NOVA), but restrictions apply to their availability. Parts of the data are publicly available at https://doi.org/ 10.18712/NSD-NSD3007-V3. The whole dataset used in this study is not publicly accessible and was used under license for the current study. However, it is available upon reasonable request to the corresponding author and with permission of the owner of the dataset, Norwegian Social Research (NOVA).

Code availability
The code for all analyses reported in the manuscript is available at https://doi.org/10.17605/OSF.IO/62FAH.