Economic insecurity: A socioeconomic determinant of mental health (2024)

• •

  Effects of both subjective and objective economic insecurity are analysed within a causal model.

• •

  Fear of unemployment is the most damaging form of economic insecurity, particularly for males.

• •

  The negative effect of insecurity is experienced equally throughout the income distribution.

• •

  Effects observed regardless of unemployment fears being realised.

• •

  Estimates which do not control for simultaneity potentially understate the size of the effect.

Biasδ=βu1−βuδVar(ε)Var(F)

(1)

3.1. Data and selection of sample

The data used within the analysis comes from the BHPS (Institute for Social and Economic Research, 2010). All eighteen waves of available data are used to build the dataset. We measure the proportion of time spent unemployed within the last three years, and form a dummy variable indicating if the individual has a spell of unemployment within the next year. The need to include these three lags and one lead restricts the sample to the period covering 1993–2007. It was not possible to extend the sample beyond this point due to a discontinuity in the way that income is measured in the BHPS and the Understanding Society study, which replaced the BHPS after 2008. The need for a minimum of four consecutive periods of data for an individual also reduced our potential sample by approximately half. The focus for the analysis is working-age (16–64) sample members. Descriptive statistics for all variables are contained within Table A.1.

A number of additional sample restrictions are imposed in order to allow for a distinct effect of economic insecurity on health to be identified. Firstly, sample members must be part of the ‘primary’ workforce – that is employed on a permanent full-time contract. This restriction is imposed to address the possibility of individuals voluntarily or knowingly selecting into insecure employment, although it must be recognised that individuals may select into secure employment. However, the purpose of this restriction is to form a sample of employees who could reasonable expect security in their employment relationship, so selection into secure employment is a lesser concern. A further restriction is that the sample is limited to individuals either without a partner or whose partner is not suffering from either work-related or financial insecurity. Although intra-household transmission of economic insecurity is an interesting aspect that has not been thoroughly investigated to date, the aim of the current analysis is to accurately identify the health effect of economic insecurity on the immediate sufferer of insecurity. As a result of these sample restrictions 224 male and 949 female respondents are lost from the potential sample. By focusing on a section of the workforce which would traditionally be considered secure, it is likely that we underestimate the incidence of economic insecurity within the population and any associated mental health effects.

3.2. Dependent variable

The dependent variable in the analysis comes from responses to the GHQ-12. The questions cover aspects of mental functioning and emotional difficulties (see Appendix 2 for the question wording). Responses to the individual questions within the GHQ-12 are scored on a scale ranging from 0 (substantial decrease in symptoms compared to usual) to 3 (substantial increase in symptoms compared to usual). The twelve scores are then summed to form a Likert scale from 0 to 36 capturing a single dimension of mental wellbeing. In keeping with the relevant literature, this score has been reversed such that the scale is increasing in mental wellbeing. Additionally, the scale has been standardised to allow coefficients to be interpreted as standard deviations from the mean.

3.2.2. Subjective measurement of economic insecurity

Two subjective measures of economic insecurity are included in the analysis. The first is formed using individuals' expectations regarding their financial situation over the next twelve months. Individuals indicating that they expect their financial situation to worsen (originally on a 3-point scale) are coded as 1, while others are coded as zero. This subjective measure is called financial insecurity.

The second subjective measure comes from a question capturing individuals’ levels of satisfaction with their current job security.² Responses are given on a 7-point scale. These have been recoded into a binary variable such that those expressing any dissatisfaction with their current level of job security are coded as 1. This subset is called work-related economic insecurity.

Work-related economic insecurity is distinct from conceptualisations of job insecurity which involve only the risk of involuntary employment termination. The key distinction being that information other than the probability of job loss, such as the financial and non-financial consequences of employment volatility, is assumed to be contained within responses. Such a conceptualisation is broadly similar to the cognitive and affective insecurity distinction presented by Anderson and Pontusson (2007).

To investigate the validity of this assumption, we compare responses on two related questions. Within two waves of the BHPS respondents were asked “In the next twelve months how likely do you think it is that you will become unemployed?” as an additional question. Responses were given on a four-point scale ranging from “1 Very likely” to “4 Very Unlikely”. These responses are presented in Fig. 1 along with the proportion of respondents within each category that were classified as suffering (or not) from work-related economic insecurity, as defined above.

Open in a separate window

Fig. 1

Comparison of Unemployment Expectations and Work-related Insecurity Classification (1996 and 1997 only).

The key feature of Fig. 1 is that a large group of individuals express dissatisfaction with current job security despite perceiving a low likelihood of unemployment. These individuals are contained within the third and fourth shaded columns. This suggests responses to the question regarding satisfaction with current job security, and, as such, our work-related insecurity variable, likely involves more than just an assessment of becoming unemployed. It should be noted that an employment relationship can be involuntarily terminated without unemployment necessarily occurring if another job can be found quickly. Although such job loss is still likely to be costly since replacement jobs could be expected to be inferior. Therefore, Fig. 1 does not capture the full complexity of the relationship between insecurity and employment volatility, but it is presented to highlight a conceptual distinction between unemployment perceptions and work-related insecurity which may influence comparisons with studies which use an alternative question wording.

3.3. Model specification

The specification used to identify a lower bound for the effects of economic insecurity on mental health takes the form:

where H_it is the standardised GHQ-12 score for individual i at time t, I is the dummy variable indicating exposure to economic insecurity (work-related, financial, qualifying income loss, or ESIGB), P is a continuous variable capturing the proportion of time spent unemployed in the past 3 years, and F is a dummy variable which equals one if the individual had a spell of unemployment (regardless of duration) in the next 12 months. X is a vector of standard controls (percentiles of equivalised household income, existing medical condition dummy, education dummies, age bands, marital status dummies, number of children, industry of employment dummies, log of hours worked, and employer size dummies). The individual-specific intercept is given by α, η is the time dummy, and ε represents the idiosyncratic error. All standard errors are clustered at the individual level such that they are robust to arbitrary heteroscedasticity and within-subject autocorrelation.

The model above is first estimated using the fixed effects (FE) estimator. This approach requires that the dependent variable is cardinal in nature, which is evidently not the case when using the GHQ-12 score. However, such an approach is common within the relevant literature (Clark et al., 2010, Rohde et al., 2016, Watson, 2015), and a benefit of this strategy is that results are readily comparable to those produced by the instrumental variables (IV) estimation.³

A valid comparison between the FE and IV results is vital if a lower bound for the effect of economic insecurity on mental health is to be identified. The FE results are assumed to suffer from simultaneity bias which results in the coefficient being smaller (in absolute terms) than would otherwise be found. IV results, which do not suffer from simultaneity bias if sufficiently strong instruments are employed, should exhibit larger negative effects of insecurity on mental health. Thus, the FE results can be considered to be a conservative estimate of the ‘true’ effect size.

To enable IV estimation, variations in levels of insecurity within individuals’ reference groups are used. These instruments are constructed at the 2-digit occupation (17 categories), 2-digit industry (25 categories), and region (12 categories) levels. In each case the mean level of the relevant form of economic insecurity is used. For example, in an IV regression with financial insecurity as the dependent variable, all individuals within London will have the mean level of financial insecurity in London as one of the exogenous instruments. Two further instruments would be assigned based on the individual’s industry and occupation. A minimum sample size restriction is applied to form the instruments. In each case, the first percentile of category size is used as the cut-off. For every year the mean level of insecurity is calculated with each industry category having at least 35 observations, each occupation category having at least 30 observations, and each region category having at least 140 observations.

The rationale behind the instruments is that changes in mean insecurity at the region, industry, or occupational level will only impact on individuals’ mental health by altering expectations about their own employment or financial situation – i.e. individuals will form expectations by taking cues from their economic reference groups. Importantly, awareness of the mean levels of insecurity will not impact on individuals’ current mental wellbeing directly. Any effect will occur only by altering their economic insecurity – i.e. their level of concern for their future situation. Results from an ancillary regression are presented in Table A.2 where the instruments were shown to be both jointly and individually insignificant (p>0.1) in a regression on GHQ-12 score. These results suggest that the instruments are correctly excluded from the outcome equation of interest since we fail to reject the null hypotheses that aggregate insecurity has no direct effect on mental health.

4.1. Comparison of trends

Given the difficulty of capturing economic insecurity within a single measure it is worthwhile to examine a range of indicators. Fig. 2 illustrates the mean level in each of the measures of economic insecurity. It is apparent that all measures show a decline in the level of economic insecurity until around 2001. This decline coincides with an extended period of macroeconomic stability, with the national unemployment rate, which can considered to be the objective risk of unemployment, also declining. Similar results showing declining levels of economic insecurity have been found for Australia (Rohde et al., 2014), although in the US it appears that economic insecurity increased over the same period (Hacker et al., 2014).

Open in a separate window

Fig. 2

Mean Levels of ESIGB and Subjective Measures of Economic Insecurity.

Beyond 2001 the trends differ. The full ESIGB, qualifying income drops, and financial insecurity show a slight increase until 2007. The only measure that does not display any clear increase beyond 2001 is the measure of work-related insecurity. The trend in this measure declines steadily. However, in all periods within our sample (1993–2007) work-related insecurity affects the greatest proportion of the sample compared to the other insecurity measures.

Our sample period predates the Great Recession, which would be expected to alter the trends in economic insecurity within Fig. 2. However, the period during and following the Great Recession is also of interest. As such we present some roughly comparable figures within the shaded area of Fig. 2 and in Table 1.⁴

A striking feature of the shaded area in Fig. 2 is the substantial increase in financial insecurity between 2007 and 2008. The proportion of individuals who expected their financial situation to worsen increased from around 5% to over 15%. Table 1 suggests that this rate was sustained, or even increased, until as recently as 2014. We do not observe an increase of this scale in the other available measures of economic insecurity, although it does appear that insecurity also increased in these measures following the Great Recession.

Due to the reasons outlined in Section 3.1. we are unable to include the period beyond 2007 within the current study. The analysis within the rest of the paper focuses on the period from 1993 to 2007.

4.2. Fixed effect regression

Table 2 gives the results from the FE regressions for each measure of economic insecurity. FE is used to control for confounding variables and individual heterogeneity. Both future orientated measures of economic insecurity (work and financial) result in a significant negative effect on mental health in males and females. For males, being dissatisfied with current levels of job security reduces mental wellbeing by 0.316 of a standard deviation from the mean. The equivalent figure for females is 0.171. Using Canadian data, Watson (2015) finds a negative effect on male mental health of 0.14 standard deviations from the mean, and 0.09 for females. For both genders the size of this effect is lower than those reported in Table 2 although the pattern of work-related insecurity affecting males to a greater extent is observed.

Gender differences are not observed for the other insecurity variables. When an individual expects their financial situation will worsen within the next year this reduces mental health by roughly 0.18 of a standard deviation for both males and females. In line with Knabe and Rätzel (2011), there is no statistically significant (at the 5% significance level) effect from past unemployment when current work-related insecurity is controlled for.

Table 2 shows that subjective elements of economic insecurity have a stronger negative impact on mental health than the objective measures used. There are no statistically significant effects associated with either experiencing a 25% or more decrease in household income or the full ESIGB index. For both males and females, it appears that exposure to perceived risk is more harmful than the realisation of risk. This is an important point since individuals may suffer a negative effect on mental health regardless of the perception of risk being either justified or realised. Therefore, these results suggest that the negative effects of economic insecurity may be hidden from conventional measurement by objective indicators.

The analysis of Rohde et al. (2016) enables some comparison across a range of economic insecurity measures affecting mental health. Rohde et al. (2016) find that financial insecurity is slightly more damaging to mental health than work-related insecurity. This differs from the results in Table 2 where work-related insecurity is clearly more damaging than financial insecurity in males, although Rohde et al. (2016) do not separate their sample by gender. It is particularly informative that Rohde et al. (2016) also find no statistically significant effect of a drop in household income of 25% or more. Rohde et al. (2016) impose an additional condition that income during the period in which the loss is incurred must be below the household-specific mean income over the entire sample period. The results in the last four columns of Table 2, along with those in Rohde et al. (2016), suggest that transitory income shocks have little impact on mental health. One reason for this may be that such shocks are the result of planned transitions. Alternatively, the psychological distress resulting from an unplanned, but expected, loss may occur prior to the event, and as such may be more likely to be captured by the financial insecurity variable.

As discussed in Section 3.2.1, liquid financial wealth holdings are often minimal amongst those suffering a qualifying income loss. This suggests that asset poverty may be a major risk factor associated with the ESIGB, since sufferers of economic insecurity are more likely to have little or no protection from household income volatility. Therefore, if such a loss was unexpected it is plausible that these individuals would target credit markets to meet any shortfall in income. Such a scenario could cause chronic psychological distress as individuals manage the resulting burden of debt over multiple periods.

To provide some context as to the relative size of the effect of insecurity, it is useful to make a comparison with other statistically significant variables. Suffering from one or more existing health problems has a negative effect on mental health of approximately 0.12 and 0.15 for males and females, respectively. Therefore, at the individual level either work-related or financial insecurity is more damaging to mental health than suffering from an existing health problem. A further comparison can be made to a major life event - separating from a partner. For males and females, separating from a partner has a larger negative effect on mental health than all forms of insecurity. The negative impact for males is approximately 0.56 standard deviations from the mean, and for females around 0.28.

It is also useful to employ the MDD concept to assess the magnitude of coefficients. The MDD is the threshold level of change in an outcome that an individual would perceive as either beneficial or harmful and which would lead to the individual altering how they manage the issue (Schunemann et al., 2005). Across many studies this threshold has been observed to be between 0.3 and 0.5 standard deviations from the mean (Norman et al., 2004). Therefore, this level can be considered to be a valid approximation in the absence of clinically observed values. By this measure the gender differences become more significant for work-related insecurity since for males the coefficient is above the MDD threshold, while for females it is below this threshold. Consequently, male sufferers may change their behaviour in response to insecurity. This could include seeking clinical treatment or altering other behaviours, for example, taking a period of time off work. It is also possible that a change in behaviour aimed at coping with psychological distress may produce further negative health outcomes, such as obesity and hypertension (Ferrie et al., 1995, Ferrie et al., 2002, Niedzwiedz et al., 2017, Rohde et al., 2017b).

To investigate if exposure to work-related economic insecurity did change behaviour we conducted a series of t-tests. We found a statistically significant difference in the number of hours worked by men who had suffered from work-related economic insecurity in the current (P<0.01) or previous (P<0.01) period compared with those who had not suffered insecurity. The insecure worked fewer hours. These differences were not found in the female sample. Furthermore, a higher incidence of smokers was observed amongst men who had suffered from work-related insecurity in the previous period (P<0.05). This difference was not observed in the period within which insecurity was suffered. Statistically significant differences in the incidence of smoking were found in both periods for the female sample (P<0.01). Although not unequivocal, these tests provide some support that work-related economic insecurity is associated with changes in behaviour which were more pronounced above the MDD threshold.

Pertinent to the analysis is the test of a “breadwinner hypothesis” conducted by Watson (2015). An alternative test of this “breadwinner hypothesis” was carried out in the current analysis by limiting the sample to only those identified within the BHPS as being legally or financially responsible for the household (2,159 males and 1,009 females). As with Watson (2015), this resulted in the coefficient for work-related economic insecurity becoming insignificant for females at the 5% significance level (p=0.068), while the equivalent coefficient for males became larger (β=−0.330) and remains statistically significant (p<0.01). Additional checks showed little difference in the proportion of household income being provided by male and female household heads. As such, it may not be the “breadwinner” role in itself that causes the gender differences that are observed in many studies of job insecurity, although it cannot be ruled out that the responsibilities of this role impact differently on males and females.

The results in this analysis support the suggestion that the source of this difference between genders involves more than financial aspects. Clark (2003) and Strandh et al., (2013) show that differing societal expectations regarding labour force participation of males and females can alter the negative impact on wellbeing resulting from unemployment. It may also be the case that such societal expectations influence the extent of psychological distress resulting from a fear of unemployment.

One gender difference shown in Table 2 which contradicts existing literature is that anticipation of unemployment has a statistically significant negative effect on males, but no effect on females. This result is apparent across all measures of economic insecurity that are used within the analysis. Conversely, Clark and Georgellis (2013) show a negative impact on GHQ score for both males and females prior to unemployment being experienced. This effect is larger for females. It should be noted, however, that the specifications in Table 2 differ from Clark and Georgellis (2013). In particular, the inclusion of current work-related economic insecurity would be expected to capture a similar effect to anticipation of unemployment, since work-related insecurity includes a cognitive assessment of the likelihood of suffering involuntary termination of employment.

Additional regressions interacting the economic insecurity variables with income, unemployment experience, and unemployment outcomes within the next 12 months were performed. Only one of these individual interactions was statistically significant at the 5% level (financial insecurity and unemployment experience in females), and in every case the interactions were jointly insignificant. Therefore, the decision was made to exclude these interactions from the main results. However, the lack of statistically significant interactions may still be informative. Firstly this shows that the negative effect of work-related and financial insecurity is experienced regardless of future unemployment outcomes. More importantly, the negative effects of work-related and financial economic insecurity on mental health do not appear to change depending on the sufferer’s position within the household-income distribution. Such a finding is consistent with results using Australian data reported by Rohde et al. (2017a) which showed that the sensitivity of health to insecurity varied little with increased income. Thus, suggesting that threats to an individual’s relative status, rather than a risk of poverty, is the principal factor in the insecurity and health relationship.

4.3. Instrumental variables regression

Table 3 reports the results of the IV estimation. The purpose of this stage of analysis is to test the hypothesis, based on Geishecker (2012), that the coefficients on the economic insecurity variables from the FE analysis (Table 2) underestimate the effects of economic insecurity on mental health due to simultaneity bias. This prediction can be readily applied to work-related economic insecurity. In the case of financial insecurity this is less clear as the source of the insecurity is not known, although this could conceivably result from an expectation of employment volatility. The objective insecurity measures are based on the experience of past events, and there is no reason to believe that current mental health will affect past events. Hence, simultaneity bias was neither suspected nor supported in these cases and the results are not reported in Table 3.

For males, the hypothesis that the FE results underestimate the negative effects of work-related economic insecurity on mental health is supported. Diagnostics tests indicate that the instruments are sufficiently strong (Kleibergen-Paap F=24.32) to identify a coherent set of parameters (). Therefore, it is likely that this relationship does suffer from simultaneity bias. The coefficient is almost three times larger in the IV regression compared to the FE results, which is roughly equivalent to the bias found by Geishecker (2012) using a sample which pooled the genders. Furthermore, insecurity becomes the largest negative factor in the model and the anticipation of unemployment variable is no longer statistically significant, although we fail to reject the null hypothesis that this coefficient equals the value found in the FE regression (P>.1). It is not possible to make direct comparison of the bias size with other studies which do not control for potential simultaneity bias. However, it could be assumed that these studies underestimate the size of the negative effect.

The results for financial insecurity in males are less conclusive. Although the absolute value of the coefficient is larger in the IV model compared to the FE model, the coefficient is no longer statistically significant. This could be interpreted as there being no simultaneity bias. Such a finding would suggest that the sources of insecurity are not common to the work-related insecurity measure and that the FE result provide evidence of a causal relationship.

For females, the results appear to be affected by weak instruments, as reflected in the Kleibergen-Paap statistic, and so the coefficients and standard errors may be unreliable. To understand why we find little evidence of simultaneity in the insecurity and mental health relationship for females it is worthwhile to consider the social norm effects of unemployment (Clark, 2003). As Geishecker (2012) illustrates (Eq. 1 above), simultaneity bias in the insecurity and mental health relationship increases as the wellbeing derived from current employment increases. Social norms may influence the current in-job wellbeing. If this is higher for males than females (βuMale>βuFemale), as found by Clark et al. (2010), then this implies that we could expect simultaneity bias to be more substantial within the FE regression for males.

A further implication of social norm theory is that the reference group effect may be weaker for females compared to males. Our identification strategy assumes that individuals will respond to insecurity within their reference group. However, given the result of Clark et al. (2010) that the social norm of employment differs between genders, then the extent to which males and females respond to variations in the reference group will also differ. It is possible that males are more responsive to variations in the reference group as a result of social norms, and so the instruments are more effective.

¹This contrasts with the broader research on the effect of aggregate economic shocks and overall health. Some papers (Ruhm, 2000, Tekin et al., 2017) find that economic shocks can generate better health, although negative effects are found for mental health outcomes.

²The BHPS question asks “How satisfied would you say you are with the job security in your present job?”.

³Performed in Stata using XTIVREG2 (Schaffer, 2015).

⁴The data prior to 2008 and after 2008 come from two different datasets – the BHPS and Understanding Society, respectively. Although there is some overlap in respondents and questions, the two datasets are not strictly comparable. Thus, while we focus on the BHPS in the regression analysis below, we have added the later data in Table 1 for illustrative purposes.

Table A.1

Descriptive statistics.

Table A.2

The Effect of Economic Insecurity on Standardised GHQ-12 Score including Exogenous Instruments (Fixed Effect Regression).

• Anderson C.J., Pontusson J. Workers, worries and welfare states: social protection and job insecurity in 15 OECD countries. European Journal of Political Research. 2007;46(2):211–235. [Google Scholar]
• Banks J., Blundell R., Smith J.P. Understanding differences in household financial wealth between the United States and Great Britain. Journal of Human Resources. 2003;38(2):241–279. [Google Scholar]
• Benach J., Vives A., Amable M., Vanroelen C., Tarafa G., Muntaner C. Precarious employment: understanding an emerging social determinant of health. Annual Review of Public Health. 2014;35(1):229–253. [PubMed] [Google Scholar]
• Bender K.A., Theodossiou I. The unintended consequences of flexicurity: the health consequences of flexible employment. Review of Income and Wealth. 2017 Forthcoming. [Google Scholar]
• Bossert W., D’Ambrosio C. Measuring economic insecurity. International Economic Review. 2013;54(3):1017–1030. [Google Scholar]
• Clark A., Oswald A. Unhappiness and unemployment. Economic Journal. 1994;104(424):648–659. [Google Scholar]
• Clark A. Unemployment as a social norm: psychological evidence from panel data. Journal of Labor Economics. 2003;21(2):323–351. [Google Scholar]
• Clark A., Georgellis Y., Sanfey P. Scarring: the psychological impact of past unemployment. Economica. 2001;68(270):221–241. [Google Scholar]
• Clark A., Knabe A., Rätzel S. Boon or bane? Others' unemployment, well-being and job insecurity. Labour Economics. 2010;17(1):52–61. [Google Scholar]
• Clark A., Georgellis Y. Back to baseline in Britain: adaptation in the British household panel survey. Economica. 2013;80(319):496–512. [Google Scholar]
• Ferrie J.E., Shipley M.J., Marmot M.G., Stansfeld S., Smith G.D. Health effects of anticipation of job change and non-employment: Longitudinal data from the Whitehall II study. BMJ: British Medical Journal. 1995;311(7015):1264–1269. [PMC free article] [PubMed] [Google Scholar]
• Ferrie J., Shipley M., Stansfeld S., Marmot M. Effects of chronic job insecurity and change in job security on self reported health, minor psychiatric morbidity, physiological measures, and health related behaviours in British civil servants: The Whitehall II study. Journal of Epidemiology and Community Health. 2002;56(6):450–454. [PMC free article] [PubMed] [Google Scholar]
• Geishecker I. Simultaneity bias in the analysis of perceived job insecurity and subjective well-being. Economics Letters. 2012;116(3):319–321. [Google Scholar]
• Goldberg D.P., Gater R., Sartorius N., Ustun T.B., Piccinelli M., Gureje O., Rutter C. The validity of two versions of the GHQ in the WHO study of mental illness in general health care. Psychol Med. 1997;27(1):191–197. [PubMed] [Google Scholar]
• Green F. Unpacking the misery multiplier: How employability modifies the impacts of unemployment and job insecurity on life satisfaction and mental health. Journal of Health Economics. 2011;30(2):265–276. [PubMed] [Google Scholar]
• Hacker J., Huber G., Nichols A., Rehm P., Schlesinger M., Valletta R., Craig S. The Economic Security Index: A New Measure for Research and Policy Analysis. Review of Income and Wealth. 2014;60:S5–S32. [Google Scholar]
• Hagenaars, A., de Vos, K., & Zaidi, M.A. (1994). Poverty Statistics in the Late 1980s: Research Based on Micro-data. Luxembourg.
• Hampson, E., Soneji, U., Jacob, A., Mecu, B., & McGahan, H. (2017).Mental Health and Employers: The Case for Investment. Retrieved from 〈https://www2.deloitte.com/uk/en/pages/public-sector/articles/mental-health-employers-review.html〉.
• Institute for Social and Economic Research British Household Panel Survey: Waves. 2010;1–18:1991–2009. [Google Scholar]
• Knabe A., Rätzel S. Scarring or scaring? The psychological impact of past unemployment and future unemployment risk. Economica. 2011;78(310):283–293. [Google Scholar]
• Layard, R., Chisholm, D., Patel, V., & Saxena, S. (2013). Mental Illness and Unhappiness. CEP Discussion Papers dp1239.
• Niedzwiedz C.L., Katikireddi S.V., Reeves A., McKee M., Stuckler D. Economic insecurity during the great recession and metabolic, inflammatory and liver function biomarkers: analysis of the UK household longitudinal study. Journal of Epidemiology and Community Health. 2017 [PubMed] [Google Scholar]
• Norman G.R., Sloan J.A., Wyrwich K.W. The truly remarkable universality of half a standard deviation: Confirmation through another look. Expert Review of Pharmacoeconomics and Outcomes Research. 2004;4(5):581–585. [PubMed] [Google Scholar]
• Office for National Statistics (2018). Contracts that do not guarantee a minimum number of hours: April 2018. Accessed on 24 July 2018. Retrieved from 〈https://www.ons.gov.uk/employmentandlabourmarket/peopleinwork/earningsandworkinghours/articles/contractsthatdonotguaranteeaminimumnumberofhours/april2018〉.
• Parente P.M.D.C., Santos Silva J.M.C. A cautionary note on tests of overidentifying restrictions. Economics Letters. 2012;115(2) [Google Scholar]
• Rohde N., Tang K.K., Osberg L., Rao P. Economic insecurity in Australia: who is feeling the pinch and how? Economic Record. 2014;91(292):1–15. [Google Scholar]
• Rohde N., Tang K.K., Osberg L., Rao P. The effect of economic insecurity on mental health: Recent evidence from Australian panel data. Social Science Medicine. 2016;151(Supplement C):250–258. [PubMed] [Google Scholar]
• Rohde N., Tang K.K., Osberg L., Rao P. Is it vulnerability or economic insecurity that matters for health? Journal of Economic Behavior Organization. 2017;134(Supplement C):307–319. [Google Scholar]
• Rohde N., Tang K.K., Osberg L. The self-reinforcing dynamics of economic insecurity and obesity. Applied Economics. 2017;49(17):1668–1678. [Google Scholar]
• Ruhm C.J. Are recessions good for your health? The Quarterly Journal of Economics. 2000;115(2):617–650. [Google Scholar]
• Schaffer, M.E. (2015). XTIVREG2: Stata module to perform extended IV/2SLS, GMM and AC/HAC, LIML and k-class regression for panel data models. Retrieved from 〈http://EconPapers.repec.org/RePEc:boc:bocode:s456501〉.
• Schunemann H.J., Puhan M., Goldstein R., Jaeschke R., Guyatt G.H. Measurement properties and interpretability of the Chronic respiratory disease questionnaire (CRQ) COPD: Journal of Chronic Obstructive Pulmonary Disease. 2005;2(1):81–89. [PubMed] [Google Scholar]
• Stauder J. Unemployment, unemployment duration, and health: Selection or causation? The European Journal of Health Economics. 2018 [PubMed] [Google Scholar]
• Strandh M., Hammarstrom A., Nilsson K., Nordenmark M., Russel H. Unemployment, gender and mental health: the role of the gender regime. Sociology of Health and Illness. 2013;35(5):649–665. [PubMed] [Google Scholar]
• Tekin E., McClellan C., Minyard K.J. Health and health behaviors during the great recession: A note on drinking, smoking, obesity, and physical activity. Review of Economics of the Household. 2017 [Google Scholar]
• Theodossiou I. The effects of low-pay and unemployment on psychological well-being: A logistic regression approach. Journal of Health Economics. 1998;17(1):85–104. [PubMed] [Google Scholar]
• Watson B. Dalhousie University; 2015. Exploring the Relationship Between Economic Insecurity and Health Outcomes.〈http://hdl.handle.net/10222/56324〉 (PhD Thesis) [Google Scholar]
• Watson B., Osberg L. Healing and/or breaking? The mental health implications of repeated economic insecurity. Social Science Medicine. 2017;188:119–127. [PubMed] [Google Scholar]