Young Adults’ Relationship Happiness in England During COVID-19

Research examining how major disasters (e.g. natural disasters, terrorism) and external stress and adversities (e.g. job loss, recessions) affect romantic relationships provides a lens through which we can understand the potential implications of the COVID-19 pandemic for romantic relationships. Evidence from natural disasters, such as tornadoes, floods, and hurricanes, suggests that the effects of such calamities on partnership dynamics are observed in the short-term rather than the long-term (Deryugina et al., 2018). Following Hurricane Hugo, divorce, marriage, and birth rates increased before returning to pre-disaster rates. In contrast, divorce rates declined after the 9/11 terrorist attacks before returning to previous levels (Cohan et al., 2009). One explanation offered for the divergent findings is that a longer rebuilding and recovery process following Hurricane Hugo could have created chronic stress which provided a catalyst for divorce (Cohan et al., 2009). In contrast, terrorist attacks can leave communities and individuals feeling uncertain about the world, worried about future attacks, and, therefore, their mortality. If couples are seeking comfort during times of vulnerability and uncertainty, then they may be less likely to divorce (Mikulincer & Shaver, 2016).

Separate from disaster research is the family stress perspective which articulates that economic upheaval, such as a loss of income, attendant financial strain, or difficulty in paying bills, can weaken relationship happiness and stability (Conger et al., 2010). Studies indicate that financial pressure and low income are associated with relationship conflicts and undermine relationship quality (Halliday Hardie & Lucas, 2010; Hill et al., 2018). Further, theoretical and empirical evidence from psychology and behavioral economics suggests that uncertainty and anticipatory anxiety from declining macroeconomic conditions could have consequences on behavior (Baumeister et al., 2007) and decision making (Caplin & Leahy, 2001). Together, this literature suggests that uncertainty and strain resulting from adverse economic conditions associated with the pandemic could affect romantic relationships, either through one’s own employment, one’s partner, or both.

Our study is also informed by recent scholarship in relationship science that adapts the vulnerability-stress-adaptation model specifically to couples’ relationship quality during the pandemic (Karney & Bradbury, 1995; Pietromonaco & Overall, 2021a). Researchers argue acute and chronic stress resulting from the pandemic could affect the quality of couples’ interactions due to withdrawal, less responsive support, and hostility. Further, they suggest vulnerabilities, such as socioeconomic background, life stage, racial/ethnic minority status, and age, could exacerbate the effects of pandemic related stressors on couples. Relevant to our study, Pietromonaco and Overall (2021a) emphasize that the degree to which the pandemic strengthens or threatens relationship quality will depend on the context of couples’ relationships and the extent to which partners can reciprocate support and communication.

The conceptual and theoretical perspectives reviewed here suggest that the consequences of the pandemic on couples could be compared to ecological disasters and terrorist attacks due to health and economic consequences. However, the pandemic experience was complex and there are distinct differences from disasters. Unlike the specificity of time and place related to a natural disaster, the pandemic was geographically and temporally widespread. Similar to past recessions, a macroeconomic upheaval resulted from the pandemic, but, in addition, COVID-19 came with risk to public health and personal health circumstances. Thus, couples were facing an unknown situation with a potentially long recovery process alongside fear of their own mortality.

Though not explicitly articulated in theoretical models (Pietromonaco & Overall, 2021a), residential status of couples could be considered a modifying factor. Although economic strain, job loss, and health uncertainties may interfere with emotional resources and relationship quality for all couples, couples who are not residential have the additional stressor of living apart. This could mean effortful communication which could hamper relationship satisfaction. Alternatively, it could be argued that couples not living together could make extra effort to increase capacity to both request and provide partner support. Thus, including couples who do not live together provides an opportunity to examine if relationship satisfaction differs relative to couples who are living together.

In the early stages of the pandemic, evidence suggested relationship satisfaction did not erode as individuals became more forgiving and were engaged in fewer blaming attributions of their partners (Williamson, 2020). Further studies suggested a more complex picture with heterogenous effects on family life. Using a population-level study, Schmid et al. (2021) found 40% of respondents in a German sample reported a decrease in relationship satisfaction, which was more than double the share reporting an increase in relationship satisfaction. In contrast, analysis of UK data showed that although some experienced deterioration in relationship quality, even more reported improvements (Perelli‐Harris et al., 2023). Current evidence suggests that both worsening and improving relationship satisfaction were reported (Vignoli et al., 2025). A study reviewing over 40 papers found increases in relationship satisfaction facilitated by more shared time together, but equally found relationship satisfaction to deteriorate during the first few months of the pandemic (Estlein et al., 2022). Together this literature suggests that whereas the pandemic’s adverse effects may have eclipsed some partners’ emotional resources, hindering their ability to respond to relational needs such as comfort, security, and support, other couples exhibited resilience and growth instead of dysfunction during times of COVID-19 (Pietromonaco & Overall, 2021b).

Given the heightened sense of uncertainty, initial lockdown measures may have exacerbated any negative effects, suggesting a longer observation period is needed. Further, the conflation of intimate relationships and co-residence noted previously is of particular concern in the context of the pandemic because the policy focus on restricting in-person interactions to members of the same household unit. In England, the first and strictest lockdown lasting from mid-March to mid-June 2020, included stay-at-home orders and banned in-person contact with individuals from other households. With limited easing of lockdown and phased re-opening of schools as non-essential shops, ‘support bubbles’ were introduced from mid-June. These enabled single adult households to socialise in person with one other household as if they lived together. Couples living apart but with other adults did not qualify for such bubbles. Only in September 2020 was the ‘rule of six’ introduced where two households, or up to six individuals from more than two households, could have in-person contact outdoors. This was, however, followed by a second national lockdown at the beginning of November 2020, lasting four weeks nationally but in some regions remained in place through December. A third national lockdown began January 2021 and lasted two months. Schools re-opened in early March 2021, followed by the lifting of stay-at-home orders and re-introduction of ‘the rule of six’ outdoor contact at the end of March 2021.

Thus, the focus on married and cohabiting couples in studies (Estlein et al., 2022; Pauly et al., 2022; Vignoli et al., 2025) at the exclusion of romantically involved partners who do not live together does not clarify whether the pandemic’s policies may have posed unique challenges to couples not living together. Additionally, as younger adults are less likely to live with their romantic partner (Coulter & Hu, 2017; Statistics, 2024), the current literature has disproportionately reflected the experiences of middle-aged or older adults, and those in more established relationships. Further, research should continue to explore the heterogenous circumstances of couples during the pandemic given that current evidence on relationship satisfaction among couples does not suggest a single, uniform narrative.

As argued by the Second Demographic Transition theory and the Pattern of Disadvantage theory (Lesthaeghe, 2010; Perelli-Harris & Gerber, 2011), recent changes in values and lifestyles and increasing economic and housing precarity facing younger generations have resulted in patterns of delay, extension and increased complexity in the transition to adulthood (Billari & Liefbroer, 2010; Pelikh et al., 2022). For example, in the UK, evidence suggests the Millennial generation (born in the 1980s) face postponement in first union formation, higher rates of separation among cohabiting unions, and higher rates of re-partnering compared to older cohorts (Pelikh et al., 2022). In the UK, evidence also indicates potential increases in mental health difficulties among younger Millennials compared to cohorts born around 1960 and 1970 (Henderson et al., 2020; Pierce et al., 2020; Wielgoszewska et al., 2022). It is therefore particularly relevant to better understand how younger adults and their romantic partnerships fared during the pandemic, because the outbreak and associated policy response to protect public health occurred when they were at the stage of establishing themselves in the labor market, settling down in intimate relationships, or amid early family-formation. As media reports suggest, couples may have had to unexpectedly alter their living arrangements (Fielding, 2021; Ross; Singer, 2020; Southard Ospina, 2021), or individuals living apart from their romantic partner may have felt lonelier with the restrictions, in turn impacting relationship quality.

Understanding the circumstances that shape young adults’ relationship happiness can shed light on how individuals navigated and responded to the stressors during the pandemic. Despite an abundance of COVID-19 related social research, remarkably little scholarly research attention has been paid specifically to young adults and to their intimate relationships in this context. Many studies have focused on families with children, household division of labor (Hipp & Bünning, 2021; Yerkes et al., 2020), and employment and furlough (Wels et al., 2022a). In this study, we examine relationship happiness as a key measure of wellbeing among romantically involved young adults.

Although we argue that it is necessary to scrutinize a wider range of relationship living arrangements to avoid conflating relationship status with co-residence, we are mindful that this is not without its own complexities. Younger couples who do not co-reside are likely a diverse group, in which some may be in a committed intentional ‘living apart together’ relationship, some may be better conceptualized as in an early ‘pre-cohabitation’ stage, and some may be in a casual dating relationship. By contrast, given the key role of relationship functioning as a mechanism of selection into cohabitation and marriage, young people who have moved in together may be a more homogenous group in terms of commitment and satisfaction. Nevertheless, and without making claims of causality, distinguishing partnerships based on living arrangements enables our study of relationship happiness to be more inclusive of young couples at different stages of their relationship (Duncan & Phillips, 2010).

This work contributes to the literature in several ways. First, we distinguish between residential (cohabiting or married) and non-residential romantic relationships, focusing the attention on those romantic partners who do not live together, and who can rarely be examined because of data limitations. To do so, we consider a cohort of young adults, an age group that has been largely neglected in the investigation of relationship happiness during the time of COVID-19. Also, we consider a longer window of observation in our study that allows us to investigate living arrangements during the first year of the pandemic and their association with relationship happiness a year after the beginning of the pandemic. Additional to the empirical contributions, we extend existing theoretical and conceptual frameworks to include couples who are not living together as well as young adults, features that are largely absent from recent theoretical developments in the literature (see more in Pietromonaco & Overall, 2021a).

The current study makes these contributions using high quality survey data from an existing longitudinal cohort study in England, Next Steps, which included three waves of data collection during the first year of the COVID-19 pandemic. This nationally representative sample allows us to take into account several potential factors associated with both living arrangements and relationship happiness (e.g. family background, socioeconomic status). In our analyses, we explore whether different types of living arrangements are associated with individual relationship happiness when the respondents are 30/31 years old, and investigate how this association changes when including available confounding variables.

We assess the relationship functioning of respondents using an ordered categorical measure of relationship happiness as our outcome variable. In Wave 3, relationship happiness is asked with the following question: “On a scale from 1 to 7, where 1 means that you are ‘very unhappy’ and 7 means that you are ‘very happy’, how happy is your relationship with your partner at the moment, all things considered?” This measurement is a one-item version of the Dyadic Adjustment Scale (Hunsley et al., 1995) and has been similarly used to investigate relationship happiness (Blom & Perelli-Harris, 2021; Brown et al., 2017). Single-item measures of different domains of subjective wellbeing have a satisfactory level of reliability (Diener et al., 2013). This question was asked to everyone who declared to be in a romantic relationship, irrespective of cohabitation or marital status.

Using information collected in all the three waves of the COVID-19 surveys we categorized living arrangements in three groups: ‘co-residing (living with partner before and during COVID-19), ‘moved in’ (started co-residing with partner during COVID-19), or ‘living apart’ (in a non-residential relationship both before and during COVID-19). All respondents were asked who they live with at the time of the survey in each of the three COVID-19 survey waves. Respondents who answered ‘Husband/Wife/Cohabiting Partner’ were categorized as ‘moved in’ if they answered yes to the subsequent question ‘Have you started living with your partner since the Coronavirus outbreak?’ or as ‘co-residing’ if they answered no. We categorised as ‘living apart’ those respondents who answered yes to the question ‘Are you in a relationship with someone at the moment?’ and did not indicate a ‘Husband/Wife/Cohabiting Partner’ as a household member.

For the aforementioned relationship status and living arrangement questions, we used Wave 3 questions in the first instance to construct the living arrangement variable as outlined above, to ensure the living arrangement related to the relationship for which the respondent rated their happiness in Wave 3. However, the question ‘Have you started living with your partner since the Coronavirus outbreak?’ was only asked the first time respondents took part in the COVID-19 online survey. For respondents who reported living with a partner in Wave 3 (using the question stated above) and had also responded to a previous wave (survey waves 1 and/or 2), we retrieved information on moving in with the partner during the pandemic from their earlier response. Of 3171 partnered individuals taking part in Wave 3, we derived the living arrangements variable directly from questions in Wave 3 for 2266 respondents, who are essentially first-time COVID-survey participants.

Relationship happiness and living arrangements might be influenced by other confounding variables that should be taken into account in the analysis. We adjusted for gender and household living arrangements, which include whether there are non-relatives living in the household (e.g. friends or roommates), whether they live with relatives other than the partner and own children (e.g. parents, grandparents, aunts/uncles, siblings etc.), and whether there are dependent children living in the household. We adjusted for employment status during COVID-19 and assigned respondents to one of three categories depending on their answer in Wave 3: ‘In paid work (including furlough)’, ‘Not in paid work’, or ‘In education’. The question on employment status in Wave 3 was asked only to new respondents or to those who changed jobs. Hence, for respondents who were also interviewed in Wave 2 and/or in Wave 1 and had not change jobs, we again retrieved information on employment status from their most recent COVID-survey wave. Moreover, we linked the COVID surveys with the pre-pandemic Next Steps survey data in order to include several other control variables measured at the start of the study when the participants were 14 years old: ethnicity (White, South Asian (including Indian/ Pakistani/ Bangladeshi/ Other South Asian), Black (including Black Caribbean/ Black African/ Other Black), and Other), highest parental social class (based on the national statistics socio economic classification (NS-SEC)) and highest parental education (low = less than GCSE, General Certificate of Secondary Education, i.e. end of compulsory schooling age 16 exam; medium = GCSE or GCE, General Certificate of Education, i.e. age 18 exam/equivalent to US high school diploma/ European baccalaureate; high = at least some higher education). Finally, we also included the cohort members’ highest level of education reported at age 25 (low (GCSE or less), medium (GCE A levels, Advanced Levels), high (higher education)) and their level of life satisfaction at age 25 (scale from 1 = ‘very dissatisfied’ to 5 = ‘very satisfied’).

We first report descriptive statistics for relationship happiness, living arrangements and confounders used in the analysis. We then run ordered logistic regression models, given that our dependent variable has a natural ordering of categorical levels ranging from 1 to 7 (increasing levels of happiness). In our first baseline regression model we predict relationship happiness from living arrangements and only add gender and ethnicity as covariates (M0). The second specification adds the family background information, i.e. parental social class and parents’ educational attainment (M1). The third model includes all of the other confounders (i.e. household living arrangements and employment status during COVID-19, and level of education at age 25) (M2), except for life satisfaction at age 25 that is added to a fourth specification (M3). All analyses (both descriptive statistics and regression models) are conducted using survey weights which account for sampling design and non-response across the three waves (for details on these non-response weights, see Brown et al., 2021). Some differences remained between the COVID-19 survey sample and the main cohort study despite weighting (see Appendix Table 3).

To test the sensitivity of our main results to various analytic decisions, we conducted a series of specification checks. First, we ran additional regression models (starting from M3) including interactions between living arrangements and gender, and between living arrangements and employment status. The association between living arrangements and relationship happiness is not different between men and women and does not depend on employment status. Second, we also performed sensitivity analyses to consider different categorizations of relationship happiness. In particular, we created two dichotomous variables: the first one is equal to 1 if respondents answer ‘very happy (7)’ and 0 otherwise; the second one groups together those who report the top two levels (7 or 6) on the relationship happiness question. We then ran logistic regression models using these two binary variables with the same model specifications as indicated above for the ordered logistic regressions. Our findings are substantively similar for these different categorizations. Finally, to try to take into account the potential selection among those who have been co-residing since before the pandemic (i.e. happier couples co-reside, while less happy couples do not live together or move in together because of lockdown restrictions), we performed an additional analysis to ensure temporal ordering of our dependent and independent variables: we ran a lagged regression with relationship happiness in wave 3 as an outcome, living arrangements in wave 1 as the main exposure, including all the confounders, with the addition of relationship happiness in wave 1. Such models account for time-invariant characteristics by adjusting for unobserved variables associated with the lagged relationship happiness. Results from this analysis yielded substantively similar conclusions to our main specification.

We conducted supplementary analyses that introduced additional control variables to assess if our results are robust. We adjusted for household income at age 14 to account for economic disadvantage that may select individuals into certain living arrangements, above and beyond social class and education. We accounted for the respondent’s current household income (Wave 3 of the COVID-19 surveys), which may be correlated to levels of relationship happiness. We adjusted for children’s age (when present), to account for increased parenting time and emotional resources with younger children, and partner’s employment status, which may eclipse relationship quality and time. Additionally, we considered marital status at age 25, mental health at age 25 (using the ‘12-item General Health Questionnaire’ score), and number of people per room in Wave 1 of the COVID-19 surveys (as a proxy for overcrowding) as further markers of social and economic circumstances that may select individuals into particularly partnerships. We used information on the amount of conflict between partners to assess if the associations between living arrangements and relationship happiness can be partially explained by relationship conflicts. Specifically, in Wave 3, we used a question asking about changes in the amount partners argued since the beginning of COVID-19: “Since the Coronavirus outbreak began, has the amount you have argued with your partner changed?”. Respondents could answer ‘My partner and I have argued more often’, ‘No change—same as before’, or ‘My partner and I have argued less often’. The results are robust to these additional confounders (introduced separately in M3). These results are not reported here given the smaller sample size due to the increase in missing data introduced by including these additional variables.

In Table 1 we report descriptive statistics on our analysis variables for the whole sample and by living arrangements. Nearly 80% of those in a romantic relationship in Wave 3 (February/March 2021) were already living with their partner before the pandemic outbreak. Of the remaining sample, 8.9% started living with their partner at some point after the outbreak and 12.9% were in a non-coresidential relationship. Relationship happiness is skewed towards high values, as 53.9% of the sample indicates being ‘very happy’ with their relationship all things considered.

In the sample, 56.5% are women. This is a reflection of more women than men being in a relationship at age 30/31. Half of the sample lives with at least one dependent child whereas nearly 20% live with some relatives (other than partner and children) and only 2.9% live with non-relatives (e.g. friends or roommates) in the household. For context, those who lived with dependent children had on average 1.8 children at home, who were mostly under age 5 (73.2%); only 36.9% have school age children (between ages 5 and 18). Finally, most respondents (87.0%) are in paid work (including furloughed), whereas 11.6% are not in paid work and 1.4% are still in education.

When we look at differences across categories of living arrangements, we notice that people in the ‘Living Apart’ group are less happy with their relationship compared to those who were co-residing even before the pandemic, and their life satisfaction at age 25 is also lower. This group has also a different ethnic composition compared to both other groups, with fewer white and more Black individuals. Moreover, they are more likely to live with relatives or with non-relatives and are less likely to have dependent children.

The association between relationship happiness and living arrangements from Table 1 is also confirmed looking at the unconditional mean of relationship happiness. In Fig. 1 we see that those who are not living with their romantic partner report slightly lower average happiness: 5.59 compared to 6.21 among those co-residing even before the pandemic and 6.28 among those who moved in together after the outbreak started.

Bild vergrößern

In Table 2, we show results from ordinal logistic regression models to predict relationship happiness from living arrangements, adjusting for control variables across the model specifications. In our initial regression (M0), adjusting only for gender and ethnicity, those not living with their partner report lower relationship happiness, on average, than those who lived with their partner since before the outbreak (Odds Ratio (OR) = 0.54, 95% CI 0.320–0.906). There is no difference in relationship happiness between those who moved in with their partner during the pandemic and those co-residing prior to the start of the pandemic. South Asian cohort members and other minority ethnic groups (excluding Black) report lower relationship happiness than white cohort members. Reported relationship happiness does not differ by gender in any of the models.

The association between living arrangements and relationship happiness remains robust when we include family background, in particular parents’ social class and educational attainment (M1), household living arrangements, employment status and education at age 25 (M2). Of the other variables included in model M2, living with dependent children in the household is associated with lower relationship happiness.

Finally, in M3 we also adjust for life satisfaction at age 25. We do not have any information on relationship happiness/satisfaction before the pandemic, but it is possible that those who are generally happier and more satisfied with their life also report a higher happiness with their partner. The inclusion of life satisfaction at age 25 does not change the results of our analysis (Living apart OR = 0.42, 95% CI 0.257–0.684).

This association can also be seen in Fig. 2, where we report the predicted probabilities of relationship happiness by living arrangements, derived from M3. Reports of being ‘very happy’ with the relationship are more likely among those who were co-residing before COVID-19 and those who moved in together compared to those who were not living with their partner. The three categories of living arrangements do not differ substantially for the other happiness ratings (lower than 7).

Bild vergrößern

Additionally, we replicate the models shown in Table 2 but changing the reference category of living arrangements to ‘Living Apart’. In this way it is possible to confirm that there is a difference also between those who moved in together during COVID-19 and those living apart throughout. The results show that both those who were already co-residing before the pandemic outbreak and those who moved in together report a higher level of relationship happiness (from M3: OR = 2.39–95% CI 1.463–3.896, and OR = 2.41–95% CI 1.315–4.425, respectively), relative to cohort members who were not living with their romantic partners. Hence, those who are not living with their partner can be said to be unhappier than the other two groups (see Table 4 in the Appendix).