Article Text

Original article
The effect of leaving employment on mental health: testing ‘adaptation’ versus ‘sensitisation’ in a cohort of working-age Australians
  1. A Milner1,
  2. M J Spittal2,
  3. A Page3,
  4. A D LaMontagne1
  1. 1The McCaughey Vichealth Centre for Community Wellbeing, Melbourne School of Population and Global Health, The University of Melbourne, Melbourne, Australia
  2. 2Programs and Economics, Centre for Health Policy, Melbourne School of Population and Global Health, The University of Melbourne, Melbourne, Australia
  3. 3School of Science and Health, University of Western Sydney, Sydney, Australia
  1. Correspondence to Dr Allison Milner, The McCaughey Vichealth Centre for Community Wellbeing, Melbourne School of Population and Global Health, The University of Melbourne, Melbourne, VIC 3010, Australia; allison.milner{at}


Objectives To investigate the ‘adaptation’ versus ‘sensitisation’ hypotheses in relation to mental health and labour market transitions out of employment to determine whether mental health stabilised (adaptation) or worsened (sensitisation) as people experienced one or more periods without work.

Methods The Household Income and Labour Dynamics of Australia (HILDA) longitudinal survey was used to investigate the relationship between the number of times a person had been unemployed or had periods out of the labour force (ie, spells without work) and the Mental Component Summary (MCS) of the Short Form 36 (SF-36). Demographic, health and employment related confounders were included in a series of multilevel regression models.

Results During 2001–2010, 3362 people shifted into unemployment and 1105 shifted from employment to not in the labour force. Compared with participants who did not shift, there was a 1.64-point decline (95% CI −2.05 to −1.23, p<0.001) in scores of the MCS SF-36 among those who had one spell of unemployment (excluding not in the labour force), and a 2.56-point decline (95% CI −3.93 to −1.19, p<0.001) among those who had two or more spells of unemployment after adjusting for other variables. Findings for shifts from employment to ‘not in the labour force’ were in the same direction; however, effect sizes were smaller.

Conclusions These results indicate that multiple spells of unemployment are associated with continued, though small, declines in mental health. Those who leave employment for reasons other than unemployment experience a smaller reduction in mental health.

  • Unemployment
  • Mental Health
  • Cohort
  • Sensitisation
  • Mixed Models

Statistics from

What this paper adds

  • A substantial body of research shows that unemployment has a negative impact on mental health, while there is less research on the mental health outcomes of leaving employment owing to family reasons or sickness.

  • There is conflicting evidence about whether multiple periods without work, because of unemployment or for other reasons, have adverse effects on mental health.

  • This study investigated whether people adapted or became further sensitised to exits from employment.

  • In comparison with no shift out of work, the results suggest that the effect size for two or more periods of unemployment is greater than for one period of unemployment, supporting the sensitisation hypothesis. Leaving the labour force for other reasons was associated with a smaller decline in mental health.

  • Job seekers with a history of unemployment could be provided with additional support services to mitigate worsening mental health.

Evidence from longitudinal studies suggests that unemployment precedes a later decline in mental health.1–5 This relationship is influenced by a number of factors, including history of mental health problems and type of job held before unemployment, amount of time spent unemployed, prevalence of unemployment within a country and the presence of welfare and social support mechanisms.6–8

Past experiences of unemployment have also been shown to be important,8 and there is some evidence from studies in the UK and Europe that multiple experiences of unemployment can have detrimental impacts on life satisfaction and mental health.3 ,5 ,9 ,10 A longitudinal, nationally representative study of unemployment in the UK,9 found that first and second experiences of job loss were associated with the largest declines in individual mental health, while there was no decrease in mental health at the third experience of unemployment. Results also suggest that those who were previously economically inactive (not in the labour force (NILF)) became more sensitised (ie, had continued declines in mental health) with additional exits from work. The authors explain ‘sensitisation’ to unemployment using diathesis-stress models of depression, which posits that increasing exposure to a negative stimulus will result in an increased sensitivity (or response).11 In this case, multiple exposures to unemployment sensitised individuals and resulted in worse mental health. In contrast, the adaptation hypothesis suggests that individuals repeatedly exposed to an adverse experience (such as job loss) will become accustomed to this over time and have a smaller adverse reaction.9

A related body of research has examined the effect of being economically inactive (or NILF) on mental health.12–15 This group is similar to those who are unemployed in that they are also economically inactive, but differ in that they are not seeking work.15 ,16 The heterogeneous nature of the economically inactive category is another defining feature, as people in this group may be permanently sick, early retirees, in the home caring for children and relatives,16 or students.15 Research suggests considerable variation in mental health within those described as economically inactive.14 As a recent example, a study in the UK14 found that shifting into economic inactivity because of illness was associated with a significant decline in mental health, while moving out of employment for other reasons (eg, home caring for children and relatives, etc) had no significant influence on mental health.

This study seeks to examine sensitivity and adaptation in the Australian context by examining the relationship between the number of times a person has been without work (also referred to as a ‘spell’ or period without work) and mental health. Understanding the effect of multiple spells without work is particularly relevant given shifts towards ‘flexible’ or temporary (casual) forms of labour in Australia17 and other countries.18 A movement towards fixed-term contract and casual work has created concerns about job security19 and the possibility that more workers will spend some periods (or multiple periods) out of the workforce.20 This study complements recent research on the relationship between job quality, unemployment and mental health using the same dataset,21 while still making its own unique contribution. First of all, our study separates out the effects of first and second periods without work on mental health. Second, it examines the influence of other reasons for leaving the labour force (eg, early retirement, illness, to care for family) on mental health.

We examine two groups in particular, those who transition from employment to NILF (which comprises those with a long-term health condition or disability, those completing full-time study, those who care for children or engage in home duties, those who are retired or voluntarily inactive and other reasons); and those who transition from employment to unemployment (those who remain in the labour force and are actively seeking re-employment). Those who move into NILF have also been described under the category of ‘economically inactive’.16 Both unemployment and NILF categories are assessed to investigate mental health outcomes associated with different reasons for leaving the workforce. The research question of our study asks, does shifting into unemployment or NILF multiple times have increasingly larger and detrimental impacts on the mental health of the working population compared with not shifting into unemployment or NILF?


Data source

The Household Income and Labour Dynamics of Australia (HILDA) survey is a longitudinal, nationally representative study of Australian households established in 2001. It collects detailed information annually from over 13 000 individuals within over 7000 households.22 The survey covers a range of dimensions, including social, demographic, health and economic conditions using a combination of face-to-face interviews with trained interviewers and a self-completion questionnaire. Although data are collected on each member of the household, interviews are only conducted with those aged >15 years.

The wave 1 survey began with a large national probability sample of Australian households occupying private dwellings. The homeless and those living in institutions were excluded from the initial survey, although people who subsequently moved into institutions remain in the sample.22 The response rate in wave 1 was 66%. Attrition at wave 2 of the survey was 13.2%.22 Of the group included in wave 1, interviews were carried out with 11 993 people in wave 2, 11 190 people in wave 3, 10 565 people in wave 4, 10 392 people in wave 5, 10 085 people in wave 6, 9628 people in wave 7, 9354 people in wave 8, 9245 people in wave 9 and 9002 people in wave 10; 7460 (53% of the initial wave 1 sample) people were interviewed in all 10 waves. A greater number of people were interviewed in each wave than were originally included in wave 1 because some non-respondents were successfully interviewed in later waves. Further, interviews were sought in later waves with all people in sample households who had turned 15 years of age. Additional people were added to the panel as a result of changes in household composition. For example, if a household member left his or her original household (eg, children left home, or a couple separated), an entire new household joined the panel. Inclusion of these new households is the main way in which the HILDA survey maintains sample representativeness.22

Inclusion criteria for participants

All annual waves between 2000 and 2010 were included in these analyses. To be included in the study, participants had to be employed and at some subsequent time move from employment to either unemployment or NILF.

Outcome variable

The Mental Component Summary (MCS) of the Short Form 36 (SF-36) measure was used as the primary outcome measure. The component summary score comprises four scales: mental health, role emotional, vitality and social functioning.23 A higher score on this scale reflects better mental health. The SF-36 is a widely used self-completion measure of health status, and has been validated for use in the Australian population, and to detect within-person change over time.23 The use of SF-36 in the HILDA survey has been shown to be psychometrically sound, with good internal consistency, discriminant validity and high reliability. The mean score on the MCS in HILDA is approximately 49.8 (SD 10.3) with a minimum of 4.4 and a maximum of 73.9. The overall measures of mental health in HILDA are roughly equivalent to normative data from Australian national health surveys.23 A clinically and socially meaningful change in mental health has been described as a difference of 5 points or more on the SF-36 (Ware et al, cited by Butterworth and Crosier23).

Exposure variables

The key exposure of interest was the number of times a person had moved from employment to unemployment, or periods of being NILF (ie, spells without work). Each spell without work was measured from the year in which a person moved from ‘employment’ to ‘unemployment’ or to ‘NILF’ and ended when a person was re-employed. Hence, we explicitly examined the transition from employment to NILF or unemployment, excluding transitions such as NILF into unemployment. Measurement of mental health may follow the start of a spell without work by up to 12 months. The approach to coding can be seen in online supplementary table S1.

The selection of covariates was drawn from past research on the relationship between unemployment and health.8 These included age (≤34 years, 35–54 years, ≥55 years), employment arrangements (permanent, casual or precarious, fixed-term contract/labour hire, and self-employed), occupational skill level (low, low–medium, medium–high, high as based on the Australian Standard Classification of Occupations), educational history (high school not completed, high school completed, certificate/diploma, bachelor degree or above), equivalised household income and long-term health condition (yes or no).

Statistical analysis

Descriptive statistics were used to examine the characteristics of the sample, the number of transitions out of employment and mean scores on the MCS. The relationship between the number of times a person had been without work over the period 2001 to 2010 and mental health as scored through the MCS (the outcome variable) was investigated using multilevel linear regression models specified across three levels (response, individual, household). Multilevel models include both fixed and random parts to allow for circumstances when observations are clustered, such as in the present analysis where there are multiple time points for individuals (level 2) who are nested within households (level 3). Level 1 covariates varied by follow-up period and included time, spells of unemployment or NILF and all other predictor variables. At level 2, fixed-effects terms were included to control for time-invariant individual-level characteristics such as sex and country of birth. At level 3, a random intercept was included to control for any residual variation in the outcome variable that existed across households. The statistical model allowed the investigators to assess whether the experience of moving into unemployment or NILF once or twice was associated with a change in mental health (ie, the outcome, individual MCS score) compared with not moving from employment into unemployment or NILF.

The model specified equal variances for random effects (all covariances zero) for the variance–covariance structure and took into account both the longitudinal and complex sampling aspects of the research design. Several models were fitted to the data. In the first and most parsimonious univariate model, the shifts into unemployment/NILF from employment were regressed onto mental health scores (ie, the MCS) in the same year (transition from the labour force and mental health measured in the same year). After this, the model was expanded to include other covariates. As in the previous model, a random intercept was specified for individuals (level 2) and household (level 3) to account for heterogeneity in the relationship between spells without work and mental health.

A descriptive analysis was undertaken to describe the possibility of baseline mental health differences upon first entry into the HILDA study among those would go on to experience no time out of work, one spell or two spells without work (through examination of the mean scores of the MCS). A further analysis described major reasons in the NILF category (retirement, home duties and childcare, study and illness or injury) and mean scores of the MCS. All statistical analyses were completed using Stata statistical software.24


Over the period of the HILDA survey, 21 280 people were considered eligible for the study (table 1). From this initial group, a subset of people moved from employment into either NILF or unemployment. These people contributed an average of six observations to the analysis. Of this sample, 15.8% (n=3362) of people moved into NILF once, and a further 2.4% (n=512) experienced two spells of NILF (eg, employment → NILF → employment → NILF). The main reasons for moving into NILF were retirement (36.7%), home duties and childcare (29.7%), study (9.8%), and illness or injury (10.1%). A smaller proportion of people who were initially employed moved into unemployment once (5.20%, n=1105). Only 99 people who were employed experienced two (or more) spells of unemployment.

Table 1

Description of movement out of the labour force in the HILDA survey, 2001–2010

The mean MCS score of people who transitioned into NILF (47.90; 95% CI 47.64 to 48.13) was higher than in those who transitioned into unemployment, who had a mean MCS score of 44.86 (95% CI 44.28 to 45.43). The mean MCS score of those who experienced no transition from employment to unemployment or NILF was 49.30 (95% CI 49.23 to 49.37).

Women were slightly more likely to transition into NILF than men, who, in contrast, were likely to move into unemployment (table 2). Younger people were more likely than older age groups to become unemployed, while older workers more frequently transitioned into the NILF category. Those with long-term health conditions were more likely to move into the NILF category.

Table 2

Baseline sample description of those who transitioned from employment to not in the labour force (NILF) (n=3362), unemployment, (n=1105) or had no shift (n=21 280) in the HILDA survey, 2001–2010

Compared with no shift from employment, the first shift into NILF was associated with a 1.16-point decline (95% CI −1.38 to −0.95; p<0.001) in mental health scores before adjustment for other variables (univariate analysis), and less than a 1-point decline (95% CI −1.13 to −0.70, p<0.001) after adjustment in the multivariate analysis (table 3). Two shifts into NILF (ie, employment → NILF → employment → NILF) was associated with 1-point decline (95% CI −1.51 to −0.48, p<0.001) in mental health scores in the univariate analysis, and a 0.71-point decline (95% CI −1.22 to −0.19, p value 0.007) after adjusting for other variables (multivariate analysis). Table 3 also indicates that being older and male was associated with better mental health than being ≤34 years or being female, in both the univariate and multivariate analyses. There was little evidence of an association between occupational skill level or employment arrangement after adjustment for confounders.

Table 3

Association between movement into not in the labour force (NILF) and per unit change in scores on the Mental Component Summary (MCS) of the Short Form 36 (SF-36), 2001–2010, univariate and multivariate analyses

Analyses show that the first transition into unemployment was associated with 1.85-point decline (95% CI −2.26 to −1.44, p<0.001) of the MCS in the univariate analysis compared with no shift and a 2.71-point decline (95% CI −4.08 to −1.34, p<0.001) at two spells of unemployment (table 4). After adjustment in the multivariate analysis, there was a 1.64-point decline (95% CI −2.05 to −1.23, p<0.001) in MCS scores among those who experienced one spell of unemployment, and a 2.56-point decline (95% CI −3.93 to −1.19, p<0.001) among those who experienced two or more spells of unemployment compared with no shift. The results for the other predictors are similar to those displayed in table 3. Men and older people had overall better mental health than women or those aged ≤34 years. Higher household income was associated with better mental health. Those with long-term health conditions had poorer mental health, and those working in casual jobs had worse mental health than those working in permanent jobs.

Table 4

Association between movement into unemployment and per unit change in scores on the Mental Component Summary (MCS) of the Short Form 36 (SF-36), 2001–2010, univariate and multivariate analyses

Differences in the effects of multiple experiences of unemployment and NILF are summarised in figure 1. The first transition into NILF was associated with 1.9% decrease in scores of the MCS (in the same year) compared with no shift (48.9–48 on the MCS) and a 1.4% decrease in mental health scores at the second spell of NILF (48.9–48.2 on the MCS) compared with no shift. The decline in mental health was greater for unemployment than for movement into NILF. The first spell of unemployment was associated with a 3.3% decrease in scores of the MCS (48.8–47.1 on the MCS) and a 5.2% decrease in scores on the MCS at the second spell of unemployment (48.8–46.3) compared with no shift. There was no significant difference between mental health scores associated with the first and second move into unemployment (as seen in the overlapping CIs).

Figure 1

Differences in the effects of multiple experiences of unemployment and not in the labour force (NILF).

Descriptive analysis of the baseline differences in mental health upon entry into the HILDA survey indicated that those who experienced one (47.10; 95% CI 46.04 to 48.17) or two (46.30; 95% CI 44.19 to 48.40) spells of unemployment had lower scores on the MCS upon entry into the survey than those that did not experience this shift (48.99; 95% CI 48.80 to 49.17). Investigation of major reasons for NILF showed that those who were retired in the sample had a MCS score of 52 (95% CI 51.7 to 52.4), those in home duties or childcare had a score of 47.3 (95% CI 46.8 to 47.8), those who were studying had a MCS score of 46.4 (95% CI 45.7 to 47.1) and those who were without work had a MCS score of 37.1 (95% CI 36.2 to 38). Owing to the small sample size, it was not possible to investigate differences based on number of transitions or to assess other reasons for exit.


This study found that multiple transitions out of work were associated with significant (albeit small) declines in self-reported mental health scores. The transition from employment into unemployment was associated with more than a 1-point decrease in the mental health component summary (MCS SF-36) score, and over a 2-point decrease at the second spell of unemployment in a 10-year period. This pattern suggests that people were ‘sensitised’ to two spells of unemployment and that this was associated with worsening mental health. However, it is worth noting that this may not be equivalent to a clinical change in mental health, which has been described as a difference of 5 points or more.23 In comparison, movement into NILF was associated with less than a 1-point decrease in the MCS SF-36 score. A decrease in mental health was less noticeable at the second movement than the first, which suggests support for the ‘adaptation’ hypothesis for those transitioning into NILF.

Previous studies on the effects of multiple exposures to labour force transitions have produced mixed results.3 ,5 ,9 ,10 These mixed results may be because past studies used measures of life satisfaction (rather than harder mental health outcomes) and did not explicitly focus on a sample who were employed before the shift.3 ,5 ,10 The most recent study on the effect of multiple labour market transitions (specifically on mental health)9 found support for the sensitisation hypothesis. Those who had been employed before job loss adapted (eg, mental health stabilised), while those who were previously economically inactive (NILF) became more sensitised with additional unemployment spells. This finding corroborates evidence from a longitudinal study in the USA, which found that a history of unemployment predicted depressive symptoms in later waves.25

Our study differed from previous research9 because it explicitly examined the shift from employment to unemployment or to NILF (rather than from unemployment to NILF). Further, our exposure variable was restricted to those who indicated they were employed before the shift. This approach allowed a more rigorous examination of the directionality of the relationship between transitions out of work and mental health. Despite this, our study cannot explicitly rule out the possibility that the relationship between unemployment and mental health may be bi-directional, in that mental health affects future likelihood of transition out of work.26

The results of our study suggest that shifts into the category of ‘NILF’ was associated with a small decline in mental health at the point of the first exit and a stabilisation of mental health at the time of the second transition from work (ie, adaptation). This difference is probably connected to the reasons people had for leaving employment.1 ,2 The heterogeneity in the category of NILF was demonstrated through descriptive analyses, which showed that those who were retired had a notably higher score on the MCS than those who were out of workforce because of home duties or child care, school or of illness. Owing to sample size restrictions, we were unable to examine the differences in first and second exits from employment by categories of NILF, or to examine the range of other reasons for moving into the NILF category. However, results at the general NILF level indicated different associations with the MCS depending on the number of shifts a person made into NILF. Based on this, we suggest that the first time a person leaves employment into NILF may represent a disruption to life roles and work identity, and hence be associated with a decline in mental health.1 ,2 At the point that a person leaves the workforce for the second time, adaptation to the non-work role might have occurred, and there is less decline in mental health.

Leaving the workforce into unemployment is associated with a continued decline in mental health from the first to the second spell of unemployment. There are a number of explanations for why the second spell is associated with worse outcomes than the first spell. For example, subsequent periods of unemployment might have a greater financial impact, resulting in lower income and subsequent psychological distress.27–29 There are also a number of psychological factors that might explain the decline in mental health, including a loss of time structure, social contact, collective purpose, status and esteem, and activity.27 ,28 ,30 Some research suggests that past unemployment ‘scars’ individuals, making them more vulnerable to the negative effects of future job loss.3 This result aligns with the sensitisation hypothesis, which argues that repeated exposure to an adverse event will result in an increased stress response.9 ,11 This model also recognises that the effects of unemployment are dependent on internal factors such as a person's existing level of vulnerability and appraisal processes—that is, the degree to which individuals perceive external factors such as unemployment as stressful. From this perspective, those who have been unemployed previously may appraise the loss of a job more negatively and be more sensitised to the adversities associated with unemployment (such as lower income and loss of social roles), resulting in worse mental health outcomes. It is also possible that there is a health selection effect, in that people who become unemployed are more likely to have worse baseline mental health.26 Sensitivity analysis supports this idea but suggests that differences in MCS scores of those who experienced one or two spells of unemployment are not significantly different from those of people who remain employed throughout the survey.

In general, other predictors in the NILF and unemployment analysis were similar. Older age groups had better mental health than younger age groups. This aligns with past research showing a general improvement in self-reported mental health up until the age of 70 years.31 Consistent with our study, past research demonstrates that those with long-term physical and mental health conditions also reported worse mental health,32 ,33 and those with a higher overall household income had better mental health.34 Our results also suggest that obtaining high school education was associated with marginally worse mental health than those not completing high school. However, as the effect size was relatively small and the p value just above statistical significance we would encourage readers to interpret this with caution.

The limitations of this study include a lack of data on whether people leave employment voluntarily, and employment status before entering the HILDA survey. Another limitation is that we could only examine two shifts in employment status, and that the data reported were not sensitive to when the labour market transition occurred, or movements in and out of the labour force within years of the survey. For example, a person might have been employed for the 11 months before the survey, and then become unemployed just before the survey began, while another person might have been unemployed for the entire 12 months before the survey. There is some ambiguity in the interpretation of the NILF category, which may reflect a shift out of the workforce due to illness, leaving to care for children or retirement. It might be speculated that the first of these categories would have greater adverse impacts on mental health than the other two categories.14

Finally, there is potential measurement bias in MCS responses in HILDA, which has slightly different scores than reported in previous national surveys.23 It is also worth considering that there may be bias associated with the self-reported nature of the data on mental health, although, in general, this measure has previously been found to be reliable.23 Despite these potential problems, there are numerous strengths to the dataset that justify its use in this study. These include its considerable sample size and longitudinal design, which meant that multiple time-varying exposures could be assessed against a commonly used measure of mental health. In addition, the data could be used to investigate numerous potential influences on the relationship between transitions out of work and mental health.

Although overall changes were small, this study suggests that mental health worsens with successive experiences of unemployment and that the decline in mental health is greatest at the second exit from work. Shifts out of the labour market to ‘NILF’ were associated with a small decline in mental health at the first spell, but stabilised at the second spell. This suggests that that those with repeated exposure to unemployment may be particularly at risk of mental health problems after job loss. Job seekers with a history of unemployment could be provided with additional support services to mitigate worsening mental health. At a macro-level, there is a need to consider population-level impacts of unemployment in national economic and health policy. Future research needs to focus on designing ways to measure the times at which adverse work-related events such as unemployment have the greatest impact on mental health. Research might also benefit from assessing whether work-related psychosocial characteristics and quality moderate the relationship between leaving employment and mental health.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Contributors AM conceived the article, retrieved data, conducted the analysis and wrote the initial draft. ADL and AP contributed to the conception and design of the initial draft. MJS contributed to analysis and drafting. ADL and MJS contributed to interpretation of results. All authors made substantial contributions to the final draft.

  • Funding The National Health and Medical Research Council Capacity Building Grant in Population Health and Health Services Research (ID 546248) provided salary support for AM. The HILDA Project was initiated and is funded by the Australian Government Department of Social Services (DSS) and is managed by the Melbourne Institute of Applied Economic and Social Research (Melbourne Institute).

  • Disclaimer The funding source had no involvement in the study design, collection, analysis and interpretation data, the writing of the report or in the decision to submit the paper for publication. The findings and views reported in this paper are those of the author and should not be attributed to either DSS or the Melbourne Institute.

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement The use of the data was approved by the Australian Government Department of Social Services (DSS). This paper uses unit record data from the Household, Income and Labour Dynamics in Australia (HILDA) Survey.

Request Permissions

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.