Background
The COVID-19 pandemic and associated virus-suppression measures have disrupted society worldwide [
1,
2]. UK restrictions that required staying at home (except for limited purposes such as shopping or essential work), and closure of certain businesses (e.g. retail and leisure) came into effect from 23 March 2020 [
3]. COVID-19 has especially disrupted economic activity, with many stopping work, losing income or reducing hours [
4‐
6]. Several high-income countries have implemented furlough schemes to replace lost income and maintain employment for workers temporarily unable to carry out their job. The UK introduced the Coronavirus Job Retention Scheme (CJRS), providing 80% of pay (capped at £2500 a month) for employees who were unable to work during the pandemic [
3]. Understanding the impact of such policies is crucial to inform the COVID-19 response internationally and inform international post-pandemic employment policy.
Smoking cigarettes and drinking alcohol damage health and contribute substantially to health inequalities [
7‐
9]. Both are often social behaviours and disruptions to social and economic activity during the COVID-19 pandemic have the potential for considerable impact on smoking and drinking, though the evidence so far is mixed and suggests differential effects [
10‐
14]. Employment-related disruptions may be important: while reductions in both income and social contact with other employees might be expected to limit smoking and drinking behaviour, these behaviours often increase after the loss of work [
15‐
17]. Mechanisms posited to increase smoking and drinking include stress, financial anxieties, increased leisure time, and removal of work-place barriers. It is unclear whether similar effects would be seen for stopping work as part of a furlough scheme, or during a pandemic. Furlough could still reduce social contact, increase leisure time and remove work-place barriers, but mechanisms related to stress, financial concerns, and income loss may be mitigated because furlough is temporary and a portion of income is maintained. Stress and financial concerns could also be more ubiquitous in the context of a pandemic. Some evidence already indicates that furlough is associated with increased alcohol intake [
14,
18].
Health impacts of the CJRS could differ, for example, by age, sex or socioeconomic position, so we need to understand who is most affected and to what extent this could modify existing inequalities [
2,
6,
19]. Younger workers, women and low earners tend to work in industries or sectors where people are more likely to lose jobs or be furloughed [
20]. Socioeconomic inequalities in smoking are well known, while evidence is more mixed for drinking [
21‐
24]. Both behaviours tend to peak in young adulthood with either cessation or decreases thereafter [
25,
26]. Additionally, there is evidence that younger generations drink and smoke less [
27,
28] and that sex differences (namely that men tend to smoke and drink more) have narrowed [
29,
30]. Vaping (i.e. use of electronic or e-cigarettes), a potentially less harmful alternative to smoking, has also become more common in recent years [
31,
32].
The UK National Core Studies Longitudinal Health and Wellbeing initiative combines data from many of the UK’s largest, established population-based longitudinal studies, using coordinated analysis to answer priority pandemic-related questions, with a wealth of pre-pandemic sociodemographic and health data. Conducting harmonised primary analyses within each study and pooling results via meta-analysis, means we can provide more robust and nuanced evidence regarding pandemic impacts on population health and behaviours than is possible with any one study alone, and thereby support mitigation efforts. Here, we examine how changes in employment status during the pandemic, especially participation in the furlough scheme, are associated with smoking, vaping, and drinking behaviours. We further assess whether associations vary by sex, age or education.
Discussion
Coordinated analyses in eight UK longitudinal population surveys found that, compared to stable employment, furlough was associated with a greater risk of smoking during the early stages of the COVID-19 pandemic (April–July 2020). The magnitude of this greater risk was similar to that for no longer being employed but smaller than for stable unemployment. Furlough was not associated with excess risk of vaping or drinking, compared to stable employment, while no longer being employed was associated with higher risk for both outcomes. All of these differences were largely explained when accounting for sociodemographic characteristics and pre-pandemic health (mental and physical) and behaviour, though this did differ by sex. Among women but not men, furlough and stable unemployment remained associated with risk of smoking (the magnitude of excess risk was greater for stable unemployment than for furlough), and no longer being employed remained associated with vaping, even after adjusting for pre-pandemic characteristics.
Other evidence indicates declines in smoking and vaping during the pandemic [
10], with increases in smoking cessation attempts and successful quitting [
12]. Meanwhile, the frequency of drinking and binge drinking have increased [
10,
13], but younger cohorts have reduced their drinking [
46]. There has been little evidence on how employment disruptions have contributed, though some small-scale and self-selected surveys show associations between furlough and increased drinking [
14,
18]. New unemployment has previously been associated with increases in smoking and drinking [
15‐
17], so furlough could potentially have adversely impacted on smoking and drinking behaviour, but the preservation of income and recognisably temporary nature of furlough may have softened usual impacts of ceasing work. After adjusting for pre-pandemic characteristics, there was little evidence of adverse impacts. We found only that furlough was associated with a higher risk of smoking among women, and with increased alcohol consumption among men and that only when specifically analysing the self-reported change in drinking.
Where we did find associations, these were weak enough to be plausibly explained by unmeasured confounding. Furthermore, there was no clear evidence that the association between furlough and smoking for women was actually stronger than that among men: the overall analyses did not show a clear relationship. The association between furlough and increased drinking for men was not consistently apparent across all analyses, as would be expected if it were a strong and robust causal relationship: the main analyses did not support this, either before or after adjustment for pre-pandemic characteristics (including pre-pandemic drinking). This discrepancy could be because: associations with change outcomes can include effects of employment on pre-pandemic drinking; or because analyses of change were sensitive to relatively minor changes in behaviour above or below the threshold used to identify heavy drinking in the main analyses (and such minor changes may have less overall relevance for health). Measurement may be especially important, as a previous study only found associations with furlough for heavy episodic drinking [
14], rather than for frequency or quantity measures. There was also an indication that furlough was associated with increased vaping, but some or all of this may represent moves towards smoking cessation. Thus, furlough does not appear to have had any clear, adverse impacts on smoking, vaping or drinking behaviour. Loss of work has previously been shown to be detrimental, being associated with increases in smoking and alcohol consumption [
16,
17]. The lack of such associations with furlough, at least in the early stages of the pandemic (when furlough was at its peak), suggests the policy, as an alternative to job loss, may have been protective.
Combining analyses from eight UK prospective studies (six of which were nationally representative of their target age range) makes a clear contribution to understanding the potential impact of furlough, but limitations should be taken into account. We were not able to achieve full harmonisation of measures across studies, for example, a number of studies had only asked questions on recent pre-pandemic smoking of those who were smoking during the pandemic, meaning smoking cessation during the pandemic was unobserved in those studies. This means the analyses of change in smoking for these studies focused only on reductions in cigarettes smoked, rather than outright cessation. Main analyses will be less affected, though there may be some potential residual confounding from participants who had smoked but given up before being surveyed during the pandemic. Maximising comparability of measures across studies limited our scope to explore varied definitions with respect to frequency, quantity or other aspects of use, such as binge drinking or concurrence of smoking and vaping behaviour. Our findings represent the early stages of the pandemic (April–July 2020) in the UK. Findings may not necessarily generalise to other countries and relationships could change with: the duration of lockdown or furlough; subsequent changes to either; or even with seasonal changes in weather. Further research could explore heterogeneity over the course of the pandemic (e.g. as seasons or restrictions change), within workers who retained employment or who were furloughed from differing occupational classes or industries, and between different outcome measurement definitions/thresholds.
As with most observational studies, unobserved confounding could have affected our estimates. We did not adjust for occupational class (since it was unobservable for those who had not been employed), and there may have been differences between participants whose jobs were retained, versus those who experienced furlough or job loss. Our fully adjusted models account for differences in key pre-pandemic characteristics among employment groups or industry sectors (self-rated health, mental health, smoking, vaping and alcohol consumption), but associations with furlough could reflect other traits of these employment groups, such as how workers in different industries were responding to the pandemic, rather than being effects of furlough specifically. Nevertheless, since we largely did not find differences in behaviour to be associated with furlough, it seems implausible that confounding from occupational characteristics somehow obscured the presence of a genuine effect. Despite being embedded in long-standing cohorts, pandemic survey responses were selective, and while weighting was employed to correct for this, bias due to selective non-response cannot be excluded [
47]. Adjustment for pre-pandemic characteristics was also important, but may have introduced bias in estimates if there were unobserved determinants of both pre-pandemic characteristics and behaviour during the pandemic [
48].
Furlough was considerably more common in the early stages of the pandemic than being no longer employed or in stable unemployment. Therefore, while meta-analyses of furlough exhibited low heterogeneity between estimates (<50%), estimates for these latter groups were based on small numbers with more heterogeneity and considerable imprecision, especially in stratified analyses, or for rarer outcomes like vaping or change in smoking. Nevertheless, confidence intervals aside, the estimated magnitude of the associations for no longer being employed and stable unemployment were close to the null after adjusting for pre-pandemic characteristics (the association between no longer being employed and risk of vaping was the clearest exception to this pattern). Moving out of employment within the context of the pandemic may have different impacts compared to more typical labour market circumstances.
The lack of association that we find between furlough and either smoking or alcohol consumption may indicate that stress-related mechanisms are especially important for the increases in these behaviours normally associated with loss of work. A qualitative study on smoking during the pandemic identified stresses associated with confinement, removal of barriers/distractions, curtailment of social routines and feelings of boredom as mechanisms that could contribute to increased smoking [
49], and these are similar to the mechanisms posited for increases in smoking and drinking behaviour following loss of work [
15,
16]. Compared to the loss of work under normal circumstances, furlough will likely still have removed some work-place barriers and increased leisure time or boredom, but stress-related mechanisms would be mitigated by the maintained income and temporary nature of furlough. Furthermore, during a pandemic, even people in stable employment have had their life and working patterns disrupted, for example, by having to start working from home or make other adaptations, so stress is less-specifically associated with stopping work (by furlough or otherwise). The ubiquity of disruption may therefore help explain the lack of association observed overall, with clear differences only emerging in sub-groups or additional analyses of behaviour change.
With respect to vaping, the clearest association observed was for workers who were no longer employed during the pandemic. We interpret this cautiously, since numbers in this group were small and vaping was rare, but positively since most vapers were or had been smokers, and this group also decreased their cigarette use. While not what our study was designed to address, this is consistent with e-cigarettes being used as an aid to reducing or quitting smoking [
50,
51] among a disadvantaged group (socioeconomically disadvantaged groups have historically tended to be less successful with smoking cessation). There was also evidence indicating that furlough was associated with less vaping among those without degree-level education. Again, we should be cautious given the small numbers involved, and it may simply be that the finding is clearest in this group because the vaping prevalence is higher among those with less education (who are more likely to smoke). Decreased vaping associated with furlough is a potential concern but since there was not evidence of raised risk for smoking in this group, it is unlikely to represent a tendency for furloughed participants to choose smoking over vaping. It may be that furloughed respondents with less education find it more difficult to obtain e-cigarette devices and liquids when they are remaining at home more.
Acknowledgements
The contributing studies have been made possible because of the tireless dedication, commitment and enthusiasm of the many people who have taken part. We would like to thank the participants and the numerous team members involved in the studies including interviewers, technicians, researchers, administrators, managers, health professionals and volunteers. We are additionally grateful to our funders for their financial input and support in making this research happen.
GS: Drew Altschul, Chloe Fawns-Ritchie, Archie Campbell, Robin Flaig.
ALSPAC: Daniel J Smith, Nicholas J Timpson, Kate Northstone
Understanding Society: Michaela Benzeval
MCS, NS, BCS70, NCDS: Colleagues in survey, data and cohort maintenance teams
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.