Funding This research was supported by the ESRC Research Centre on Micro-Social Change (MISOC), ESRC References Number RES-535-25-0090.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.
Statistics from Altmetric.com
- Employment status
- mental health
- longitudinal studies
- employed CG
- longitud surveys SI
- mental health DI
- psycholog distress
Participation in the labour force persists for a majority of a person's lifespan. There has been much research dedicated to understanding the benefits of labour force participation on the worker; some of these benefits include increases in self-esteem1 and life satisfaction or well-being.1–3 While continuous employment may be ideal, some people experience spells of non-employment. These spells may be due to voluntary (ie, education or training, taking care of family) or involuntary causes (ie, illness, redundancy). Just as employment can have positive effects on psychological health, unemployment has been shown to have negative effects on psychological health.1 ,2 ,4–12 There have been some indications that the UK labour market is changing and that job flexibility is increasing.13 For example, the proportion of people employed with the same firm for 5–10 years decreased from 1996 to 2001, while the number of people employed between 2 and 5 years increased over the same spell.13 The increase in flexibility may lead to increases in the number of unemployment spells. It is therefore important to understand the effects of multiple unemployment spells on psychological well-being.
A person's prior employment status may have an impact on the relationship between unemployment and well-being. In the UK, economic inactivity usually refers to those persons who are on maternity leave, are full-time home carers, in full-time education or training, with long-term illness and the retired.14 The features of economically inactive persons vary by individual characteristics including gender and age.14 There are more economically inactive women than men, with a majority of economically inactive women aged 25–49 years old staying home to take care of the family, while most economically inactive men are long-term ill or disabled.14 A majority of persons aged 16–17 years are economically inactive, mostly due to participation in full-time education.14 While some persons who are economically inactive expect to enter the labour market, for example, full-time students, others may not have such an expectation, for example, retirees or those who are disabled. Therefore, the well-being of a person in the labour market, for example, employed, who then reports a spell of unemployment may be different than for a person who is out of the labour market and then becomes unemployed (ie, seeking work). Multiple spells of unemployment further complicate this picture as some people may conceivably move from different labour market states (eg, employed, economically inactive) at each unemployment spell.
Adaptation and sensitisation theories may help to explain the effects of multiple unemployment spells. Adaptation theory posits that exposure to a stimulus or stressor will result in a reaction, which will lessen with time or with repeated exposure to that stimulus.15 ,16 The process through which adaptation occurs is not fully understood despite several mechanisms having been explored.17–20 In the case of unemployment, a person who experiences less deterioration of their well-being at subsequent spells compared with the first spell is believed to adapt.
Sensitisation, based on theories of diathesis–stress models of depression, posits that repeated exposure to a stimulus will result in an increase in response.21–23 In effect, sensitisation is the opposite of adaptation; a person will experience greater deterioration of their well-being with subsequent unemployment spells than they did at their first.
This paper builds on previous research by exploring the effects of multiple events and examining employment status prior to each unemployment spell. In addition to the main effects of unemployment on well-being, we looked for potential effect modifiers of this relationship. Research has shown that household income,9 ,11 ,24 marital status,11 age,9–11 ,24 ,25 regional unemployment rates,5 ,7 pre-study unemployment24 and gender9–11 24–26 have differential effects on the well-being of unemployed persons. While most of these are individual effects, regional unemployment rate is not. However, regional unemployment does appear to have differential effects on the employed and the unemployed, with the difference in well-being between the unemployed and employed being lower in regions with higher regional unemployment rates than in regions with lower regional unemployment rates.7
The study examines whether people who have experienced multiple unemployment spells adapt or sensitise. We aim to answer the following questions in this analysis:
What are the effects of multiple unemployment spells on well-being?
Is adaptation or sensitisation more likely if one moves into unemployment from employment or from economic inactivity?
Are there factors that moderate the effect of unemployment on well-being?
The data used for this study came from the British Household Panel Survey (BHPS). The BHPS is a nationally representative longitudinal population survey, which began in 1991 with more than 10 000 individuals in about 5500 households.27 All persons in the household aged 16 years and older were surveyed annually through face-to-face interviews on a variety of topics, including employment status, psychological well-being and income. A more detailed description of the sampling procedure and survey methods is provided by Taylor et al.27 Seventeen years of data, 1991–2008, were included in this study. Due to the sampling scheme of the BHPS and the addition of extension samples, the number of households has increased to more than 10 000. Only participants who reported at least one spell of unemployment were included in the analyses, reducing the analysis sample to 1642 persons with 20 999 person-year observations.
Self-reported employment status was obtained from annual surveys. The data allowed for in-depth investigation of different employment/non-employment states. This study explores differences between people who were in employment and those who were economically inactive prior to unemployment. Participants who reported that they were employed or self-employed were categorised as employed; all others (ie, retired, maternity leave, family care, long-term illness, full-time education and other) were categorised as economically inactive.
The 12-item General Health Questionnaire (GHQ-12) was used as the outcome measure of psychological well-being. The GHQ-12 is an instrument that has been validated to screen for minor psychiatric morbidity, specifically distress and anxiety.28 ,29 Participants were asked 12 questions that were scored on a 4-point scale. The 12 questions were scored according to GHQ-Likert scoring rules to create a continuous scale ranging from 0 to 36, where higher scores indicate poorer well-being. Use of Likert scoring for the GHQ-12 allows for a approximately normal distribution compared with the caseness scoring; the recommended scores for identifying good health are 0–11, with 12+ identifying those as at risk for illness.29
Self-reported average annual household income was included to examine whether general standard of living modified the relationship between unemployment and well-being. Income from all sources in the 12 months prior to the interview was summed to create an annual household income variable. Annual household income was adjusted for inflation, averaged for each individual across the length of their participation in the study and then log-transformed and grand-mean centred.
Regional unemployment rate and percent of unemployed adult household members were examined as potential effect modifiers, while, pre-study unemployment was included as a control variable and an effect modifier. Regional unemployment rate was based on Labour Force Survey data.30 Regional unemployment was measured according to the International Labour Organisation labour force status definitions for each region and year. The BHPS reflects two methods of determining regions in the UK that have been in use since its inception. Standard Statistical Regions were used until 1994 when Government Office Regions were introduced. The percentage of household unemployed was calculated for each individual at each wave. Pre-study unemployment was a dichotomous variable that indicated whether individuals had reported any bouts of unemployment prior to beginning their participation in the survey. No pre-study unemployment was the reference group for this variable.
Age and gender were included as both covariates and potential effect modifiers; partnership status was explored as an effect modifier. Marital status was dichotomised as ‘partnered’ (ie, married or living as couple) or ‘non-partnered’ (ie, widowed, divorced, separated or never married) to create a partnership status variable, with non-partnered as the reference group. Age was centred on the grand-mean. Men were designated as the reference group for the gender variable.
Methods used to model changes in GHQ-12 scores were adapted from approaches used in previous studies, which examined various life events and life satisfaction.5 ,7 ,9 ,10 ,24 ,25 ,31 ,32 A reaction spell is measured from the year prior to reporting unemployment through all subsequent years of unemployment. The year prior to unemployment was included in the reaction spell because research has shown that people anticipate a negative event leading to a lower level of well-being during the initial pre-unemployment spell.9 ,10 ,25 ,32 For analysis, three or more spells were combined into one category, giving a three-category variable: 1, 2 or 3+ reaction spells. How these spells were coded and examples of reaction spells for different numbers and durations of unemployment are provided in appendix 1. The intercept for the model reflects the average individual GHQ-12 score when employed or economically inactive during enrolment in the study.
Multilevel regression models, with annual measurements (level 1) nested within individuals (level 2), were used to analyse the data with SAS/STAT software V.9.1.33 A benefit of using multilevel regression is the ability to model data that is unbalanced and has unequally spaced measurement occasions as is common in panel studies. Two models were constructed, the first, model 1, examined within-person changes:where U1ij to U3ij represents the dummy-coded variables for unemployment 1−unemployment 3+ spells, εij is the random effect for individual i at measurement occasion j, and μi is the random intercept for individual i around the mean intercept β0.
Model 2, examined both between- and within-person variables as potential effect modifiers of within-person effects of unemployment:where ΣγXi represents the collection of moderators included in the study. Two versions of each model were run, one which did not distinguish the employment status prior to the reaction spells, that is, the overall estimates, and the other which did, that is, effects of unemployment following employment and economic inactivity. Models are given below:(model 1)(model 2)where UE1 to UE3 represents spells 1 to 3+ of unemployment following employment and UI1 to UI3 indicates spells 1 to 3+ of unemployment following economic inactivity.
Unemployment following employment or economic inactive spells was included in the same model. This allowed for estimation of parameters for individuals with varied employment histories. For example, if a participant went from employment to unemployment in the first spell and from economically inactive to unemployment in the second, both spells are included in the same model.
In order to assess whether the effects of unemployment differed by gender or for individuals with multiple bouts of unemployment (versus only one spell), interaction terms including gender or multiple spells were included in preliminary models. No significant interactions were found; thus, the results shown in this paper combine estimates for men and women as well as individuals with single and multiple unemployment spells. Tests for linear trend in average GHQ-12 scores across unemployment spells were also conducted. Trend tests for adaptation assumed that the increase in GHQ-12 score were highest for the first unemployment period and decreased in subsequent spells. Sensitisation trend tests assumed that GHQ-12 scores were lowest for the first unemployment spell and increased in spells 2 and 3. Pre-study unemployment, gender and age were included in all models as control variables.
Over the 17 waves of the BHPS, 12% of respondents reported at least one unemployment spell (the analysis sample). Participants with unemployment spells were younger, Mdiff=9.44 years (p<0.0001), and had slightly higher average GHQ-12 scores, Mdiff=0.35 points (p=0.008), than those who never reported being unemployed (excluded sample) (table 1).
Among the analysis sample, 6% reported ever being unemployed prior to enrolment in BHPS. In contrast, 2% of the excluded sample reported having experienced unemployment prior to their enrolment in BHPS. Eighty-two per cent of those who experienced unemployment reported one unemployment spell, 15% reported two and 3% had three or more spells during the study.
Well-being during spells of unemployment
Table 2 shows the overall parameter estimates and standard errors for model 1. Psychological well-being was poorer during all three employment spells, p<0.01 for all spells, compared with average well-being when not unemployed. No significant differences in average increase in GHQ-12 score were observed among any of the spells.
The effects of unemployment did differ based on prior employment status (table 3). Previously employed individuals had significantly poorer psychological well-being for both the first (b=1.66, t=13.40, p<0.0001) and second (b=1.47, t=5.18, p<0.0001) unemployment spells. GHQ-12 scores were not worse than usual at the third unemployment spell (p>0.05).
Previously economically inactive individuals had significantly higher GHQ-12 scores at all three unemployment spells (p<0.05). GHQ-12 scores were significantly higher at the third unemployment spell than the first (b=1.58, t=2.28, p<0.05) and second (b=1.65, t=2.18, p<0.05) spells for previously economically inactive individuals.
Previously employed individuals had significantly poorer psychological well-being at the first (b=0.75, t=3.66, p=0.0003) and second (b=0.79, t=2.04, p<0.05) unemployment spells than previously economically inactive individuals. This was reversed for the third spell where previously economically inactive individuals experienced poorer psychological well-being than previously employed individuals (b=1.94, t=1.96, p=0.05).
Adaptation or sensitisation
Without consideration of previous employment status, the point estimates show no clear pattern of either adaptation or sensitisation (table 2). Examination of trends by previous employment status gave a much clearer picture of adaptation and sensitisation. Individuals employed prior to all unemployment spells showed adaptation (F=47.01, ptrend<0.0001). Conversely, previously economically inactive individuals had a significant linear trend for sensitisation (F=12.76, ptrend=0.0004).
While looking at the trends for only previously employed or economically inactive unemployment spells together provided one part of the adaptation or sensitisation picture, we were interested in individuals with varied employment histories. The findings showed mixed patterns of adaptation and sensitisation (data not shown). Generally, regardless of the number of unemployment spells, persons with a history of economic inactivity preceding unemployment experienced worse psychological well-being (ie, sensitisation).
The results for model 2, assessed potential effect modification, are provided in table 4. Only two variables were found to moderate the relationship between unemployment and psychological well-being. Previously employed individuals who reported having been unemployed prior to enrolment in BHPS had better psychological well-being at the first unemployment spell than those who did not report prior unemployment (b=−1.08, t=−2.05, p<0.05).
Previously economically inactive individuals with higher average household incomes had significantly poorer psychological well-being at the second unemployment spell compared with those with lower average household income (b=9.26, t=2.47, p<0.05).
Generally, unemployment is associated with a worsening of psychological well-being; however, the magnitude of the deterioration differs by preceding employment status. Without consideration of previous employment status, psychological well-being worsened for all unemployment spells and tests for adaptation (ie, reacting less negatively with each unemployment spell) and sensitisation (ie, reacting more negatively) were inconclusive.
The exploration by previous employment status helped to create a clearer picture of the effect of unemployment on psychological well-being. Employed individuals experienced significantly worse psychological well-being at the first and second unemployment spells compared with those who were previously economically inactive. This was reversed for the third unemployment spell, where transition from economic inactivity resulted in higher GHQ-12 scores than for those who transitioned from employment. Trend tests for adaptation and sensitisation appear to be dependent on preceding employment status, where previously employed individuals adapted with each spell, while economically inactive individuals sensitised. One possible explanation for the difference between employed and economically inactive individuals is that employed individuals appear to remain economically active; therefore, with each unemployment spell they experience they may be less anxious about their chances of finding new employment and thus adapt. Economically inactive individuals, however, appear to have difficulties entering or re-entering the labour force. Hence, when they become unemployed, their anxiety might rise with each spell resulting in sensitisation.
Few moderators were found in the analyses conducted. Employed individuals who had experienced pre-study unemployment had better psychological well-being than those who did not. This finding is to be expected since employed persons appear to adapt to each unemployment spell. Having experienced a previous unemployment spell would result in better reaction to the next spell for employed individuals. Pre-study unemployment was found to be a significant moderator in previous studies as well.8 ,10 ,24 Average annual household income was a moderator in that economically inactive individuals with higher household incomes experienced worse psychological well-being when making an unsuccessful attempt to enter employment. These people may be especially affected by the loss other household member's income. It could also be that people who take early retirement find themselves to be unhappy and may become depressed and try to find a job in order to combat their depression.
Other studies found that gender10 ,24 and age10 were moderators; however, these findings were not observed here. This finding of few moderators of the effects of unemployment indicates that regardless of previous employment status, personal, regional or household characteristics, individuals have similar reactions to unemployment.
The results from previous studies are varied with regard to whether adaptation or sensitisation occurs with multiple unemployment spells. Clark et al8 found that men with previous unemployment experienced adaptation; however, Luhmann and Eid10 found an overall sensitisation pattern with multiple unemployment spells. There are three distinct differences between this study and those. First, those studies used life satisfaction as their measure of well-being. Life satisfaction is one of the three components of well-being34 and is often used as a proxy measure of happiness. We used the GHQ-12, which measures the negative affective component of psychological well-being, more specifically, anxiety and depression. Research has shown that life satisfaction and negative affect are distinct constructs that measure different aspects of psychological well-being.35 The use of the GHQ-12 in this study may therefore result in different findings with respect to well-being and unemployment as there may be different cognitive processes that underlie the appraisal of this measure.
Second, this study was able to examine the employment status of participants prior to their unemployment spells. This additional stratification was necessary as economically inactive participants may seek work for different reasons, such as leaving full-time education, while employed participants may have been made redundant. As the findings showed, adaptation and sensitisation was different for previously employed than for previously economically inactive participants.
Finally, previous studies were based on the German Socio-Economic Panel Study, a complimentary household panel study to the BHPS. This study covers the many historic events in Germany that did not occur in the UK, and the reunification of Germany may have had differential effects on well-being that are not seen in a UK panel.
There are several limitations to this study. This study used self-reported employment status given at the time of the annual field interview only. It is possible that some participants could have been unemployed for the majority of the year, but at the time of the interview, they were employed or economically inactive. This could influence their GHQ-12 score for that year resulting in an underestimation of the effect of unemployment on well-being.
We could not include a number of potential modifiers that may affect the relationship between unemployment and well-being. These include personality factors, coping strategies, social support, individual and household savings and potential future earnings. Some of these are not found in the BHPS, and others would create models that were not parsimonious. Future studies should address the impact of these on this relationship.
This study did not examine how psychological well-being changes when re-employment or return to economic inactivity occurs. It is possible that well-being as measured by the GHQ-12 returns to pre-unemployment levels, which may explain some of the non-significant results. Set point theory indicates that levels of life satisfaction do not return to pre-event levels; however, whether this is true for other measures of well-being is not as clear.15 ,24 ,31 ,32 Future studies should examine differences between life satisfaction and other measures of well-being as well as examine the mechanisms through which these differences occur.
Types of employment (ie, full or part time, permanent, contract or seasonal) were not explored in these analyses. Bardasi and Francesconi36 found differences in well-being, physical health and life and job satisfaction by type of employment. It is possible that reactions to unemployment would differ by type as permanent workers may be expected to have different reactions to unemployment than contract workers.
Finally, the findings from this study may not be generalisable to populations other than the British population. As noted earlier, there are some differences and similarities with findings based on a German household panel.8 ,10 ,24
The findings from this study showed that spells of unemployment were associated with declines in psychological well-being. Previously employed and economically inactive individuals differed in their reactions to multiple unemployment spells. Specifically, individuals who experienced repeated unemployment spells following economic inactivity showed a pattern of sensitivity (ie, worsening psychological well-being with each unemployment spell). While many countries have invested in welfare-to-work programs in an attempt to decrease long-term unemployment and boost their economies, findings indicate that further investigation is needed to better understand how these programs may affect the psychological well-being of economically inactive individuals who are encouraged to seek employment. In the UK, the newly implemented UK fit note, the Health, Work and Well-being initiative,37 and the proposed changes to the current welfare scheme aim to reduce the number of economically inactive persons. However, measures should also be taken to ensure that good quality employment is achieved so that repeated spells of unemployment are less likely to occur.
What is already known on this subject
With increasing job flexibility, people are likely to have more experiences of unemployment over their working life.
Unemployment is known to have negative effects on psychological well-being, but the effect of repeated exposure to unemployment is unknown.
What this study adds
In a British longitudinal panel study, we have found that employed people who became unemployed coped better psychologically with each repeated period of unemployment only if they were able to regain employment in between.
For those who made several attempts to re-enter the labour market following economic inactivity, psychological well-being was more adversely affected at each attempt. The implications for welfare to work policies are clear.
The support of both the Economic and Social Research Council (ESRC) and the University of Essex is gratefully acknowledged. Data from the British Household Panel Survey were supplied by the ESRC Data Archive.
Coding of spells
Table A1 gives an example of how initial employment status and unemployment spells were coded for persons with different numbers and durations of unemployment spells.
The unemployment spells is made up of the employed or economically inactive year prior to the first unemployment year and all consecutive years of unemployment. For example, the first reaction spell for person 1 was made up of the employed year immediately before the first year of unemployment, year 6, and all consecutive years of unemployment, years 7 through 9. Unemployment spells classified by employment status were numbered, 1, 2 or 3+, according to their overall number. For example, person 2 experienced three unemployment spells although their second spell was also their first spell that occurred after being employed; it was coded as ‘following employment spell 2.’ Person 3 experienced an unemployment spell that was counted but not included in the analyses, that is, treated as missing, because it was preceded by a wave where no data were collected, either through temporary attrition or errors, thus we do not know what their employment status was.
We tested for potential multicollinearity among the moderator variables. The correlations between these variables were very small with none exceeding 0.18. Univariate General Linear Model analyses were conducted for all nominal variables, with other nominal and continuous variables. Individuals who had experienced pre-study unemployment lived in areas with significantly higher regional unemployment rates than those who had not (p<0.05). Partnered individuals were significantly older, had a higher percentage household unemployment rate lower average annual income, lower regional unemployment rates (p<0.05 for all) than their non-partnered counterparts. Compared with individuals who only reported one bout of unemployment, those with two or more were significantly younger and had significantly more pre-study unemployment, higher incomes, lived in regions with higher unemployment and households with higher unemployed adult members (p<0.05 for all). Women were significantly younger, experienced less pre-study unemployment, had higher incomes and lived in households with lower unemployed adults than males (p<0.05 for all). Individuals who reported having been unemployed prior to enrolment in the BHPS had significantly higher incomes and lived in regions with higher unemployment rates (p<0.05 for both), than those who did not report pre-study unemployment.
Funding This research was supported by the ESRC Research Centre on Micro-Social Change (MISOC), ESRC References Number RES-535-25-0090.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.