The Effect of Employment Contracts and Job Characteristics on
Health and Well-Being
Minka H. Haverkamp
Faculty of Economics, University of Groningen EBM877A20: Master’s Thesis Economics
Abstract
Individuals’ health is closely related to their productivity, making it for both employers and employees an important factor to achieve the optimal. The purpose of this thesis is to research if employment contracts and job characteristics influence the self-reported mental health and psychological well-being of
employees in the United Kingdom. Panel data from the United Kingdom Household Longitudinal Study (UKHLS) are used, offering representative data of workers in all four United Kingdom countries. Both self-reported mental health and psychological well-being effects are measured using a random effects probit model. Results reveal that no strong evidence is found for the effect of employment contracts on mental health, nor on psychological well-being. However, certain job characteristic aspects show to be closely related to the mental health and well-being of employees. Specifically, satisfaction with the amount of leisure shows to be positively related to mental health, while negatively to psychological well-being. Thereby, psychological well-being shows to be positively influenced by working at a larger company and working non-standard working hours. Next to that, worse psychological well-being results are found for individuals who work more overtime hours.
Contents
The Effect of Employment Contracts and Job Characteristics on Health and Well-Being ... 4
Literature Review ... 6
Theoretical Framework ... 9
Data and Methodology ... 11
Empirical Model ... 18
Results ... 20
Conclusion and Recommendation ... 26
References ... 29
Appendix 1 ... 33
Appendix 2 ... 34
The Effect of Employment Contracts and Job Characteristics on Health and Well-Being
As the importance of work for mental health has been widely acknowledged but the specific influences are less clear, it is essential to address gaps in the literature regarding employees and their health (Wilhelm et al., 2004). Mental health is defined as ‘a state of well-being in which the individual realizes his or her own abilities, can cope with the normal stresses of life, can work productively and fruitfully, and is able to make a contribution to his or her community’ according to the World Health Organization (WHO). This definition clarifies how important mental health is for an individual and why a favourable mental health outcome is needed in order to enjoy life and be productive. Private life and working life are, as they take most of the time of working individuals, two important factors influencing the state of mental health and psychological well-being.
In order to enjoy a fruitful life one should feel healthy and psychologically fine, which is closely related to your productivity at work. However, concerns about the mental health and psychological well-being of employees arise, as the pressure on workers performance and flexibility is increasing in most European countries (Cottini and Lucifora, 2013). Lower mental health or well-being reduces employees productivity and increases costs, for both firms and individuals. In the meanwhile the economy globalizes extremely leading to a continuously developing market, making it hard for employees to keep up. Work, working conditions and contractual arrangements are seen as essential contributors to social inequality in health within and across generations, although they have received less attention from health researchers than other aspects of socioeconomic position (Burgard and Lin, 2013).
One of the concerns nowadays is that the nature of temporary contracts has changed.
However, considerable concerns about both part-time employment and temporary contracts are relevant as they offer fewer financial and career benefits and less employment protection (Rodriquez, 2002).
The growth of service orientated work and automatization together with the decline in
manufacturing work has significant consequences for working conditions. These global developments have made ‘traditional’ sources of working conditions less relevant to modern working practices, they appear to have increased the scope for psychosocial job stressors and their consequent effects on health (Cappelli et al., 1997). This makes working conditions an interesting source for researchers in order to find out the effect on individuals and societies.
Literature Review
Part-time working is steadily increasing across Europe, allowing individuals to create a balanced path between work and their private life (Smith et al., 2002; Messenger, 2011; Beham et al., 2019). Earlier research reports different findings when looking at satisfaction of part-time work, compared to full-time work. Focussing on gender, greater job satisfaction for women working part-time is found by Beham et al. (2019), a negative relation for males but a positive relation for females is found by Montero and Rau (2015) and a negative effect for both sexes compared to individuals working full-time is found by Karatuna and Basol (2017). These are interesting findings as job satisfaction is positively related to individuals’ health and well-being, according to Satuf et al. (2018). However, working part-time has a negative effect on individuals’ health if this is involuntary (Joyce et al. 2010).
Lepinteur (2019) found that job and leisure satisfaction increases by working shorter workweeks (working less than full-time), resulting in higher health levels. Individuals working less than full-time might have more time to invest in their health. Additionally, using the British Household Panel Study Robone et al. (2011) find a positive health outcome for employees who are satisfied with the number of working hours. When looking at individuals who are unsatisfied with the number of working hours, unfavourable health outcomes arise. Few studies did look into part-time working and both mental health and psychological well-being. Robone et al. (2011) found comparable results for mental health and psychological well-being, with a few remarkable differences. Psychological well-being is for example negatively related to having a part-time job when there are children, talking about men in particular. This might be explained by a rather old-fashioned family view, where the family relies on the husband’s income.
changed into a less traditional direction during the last years. In addition, Pirani and Salvini (2014) found a negative relationship between temporary employment and self-related health for the Italian workers, especially when it is prolonged over time. Empirical evidence also shows this result for Spanish male workers with a temporary contract, suffering from poorer mental health compared to Spanish male workers with a permanent contract (Bartoll et al., 2019).
Compared to permanent contracts, Rodriquez (2002) finds that German individuals with a fixed-term contract are 42% more likely to self-assess a poor mental health while in contrast no health effect is found for individuals from Britain. Furthermore, for both men and women a positive relationship between health and temporary contracts is found by Silla et al. (2002) when they are highly educated. This might be due to the fact that they have less stress of their temporary contract as they have more choice
(opportunities) on the labor market. The negative relationship between health and a temporary contract for individuals with a low level of education might be explained by this as well (Robone et al., 2011). Conversely, almost no evidence is found for Australian workers who experience a negative mental health effect due to a temporary contract, not for a casual nor a temporary contract. This result holds for both men and women, even when they did not complete high school (Richardson et al., 2012). LaMontagne et al. (2014) did not find any significant changes in mental health following transitions from stable
permanent employment to temporary employment.
result is also found by Cottini and Lucifora (2013), as well as less favourable health outcomes due to psychological demanding jobs and no fixed working hours. Regarding the number of employees working at a firm, mixed results are found with just positive (males) or just negative results (females) by Robone et al. (2011). They also found a negative relationship for both health and well-being when looking at overtime hours, as this results in less time for individuals to invest in their health.
Loscocco and Spitze (1990) were one of the first who focused on differences in gender while looking at working conditions and psychological well-being. They found that work-related social support has a positive effect on the well-being of both men and women doing similar factory jobs, but no effect on stressful working conditions. Looking at more recent research about psychological well-being and working conditions, comparable results are found as for health with a few exceptions. For men, not working fixed hours or during the day has a positive effect on health but a negative effect on well-being. For women, working at home is positively related to health, while it is negatively related to well-being (Robone et al., 2011). In addition, Joyce (2010) found a positive effect on both mental health and
Theoretical Framework
According to Grossman (1972) “Health can be viewed as a durable capital stock that produces an output of healthy time. It is assumed that individuals inherit an initial stock of health that depreciates with age and can be increased by investment” (p. 223). The core of this paper lies in 1972 when a model of demand for the commodity ‘good health’ is published, designed by Grossman. As stated in the quote above, health can be seen as a durable stock that individuals possess. Main items of this model are that the shadow price of health depends on many more variables aside from the price of medical care and this shadow price rises with age. Furthermore, the stock of health depreciates faster at a higher age and higher educated people are assumed to be more effective producers of health (Grossman, 1972). Individuals are able to improve their stock of health by investing their own time and investing goods such as exercising, dieting and housing. This results in a trade-off individuals face between time and monetary spending against leisure and consuming other goods. Note that a higher health does not result in a higher
productivity according to Grossman, but does result in more hours which you can effectively spend on work as you are not ill.
The difference in stock of health between today (𝐻𝑡) and the next period (𝐻𝑖𝑡 + 1) is determined
by the investment in the current period (𝐼𝑡) minus the depreciation of the health stock in the current
period (𝛿𝑡 𝐻𝑡):
𝐻𝑖𝑡 + 1 – 𝐻𝑡 = 𝐼𝑡 – 𝛿𝑡 𝐻𝑡 (1) The rates of depreciation are assumed to be exogenous, but may differ with age as explained above. However, investment in the current period can be influenced by the individual. The model of Grossman argues that investment in health does not only depend on medical care spending (𝑀𝑡), but also on job
characteristics and employment conditions at work (𝑍𝑡) and a batch of observable and unobservable personal characteristics (𝐺𝑡 and 𝑒𝑡):
Clearly, this model extends the model that investment in health only depends on spending on medical care. The stock of health is part of the personal characteristics in this model, just as the time spent on improving your health (𝑇𝐻𝑖). This is an important factor which needs to be considered, as it may influence the health level of individual 𝑖 (Grossman 1972). The variable is assumed to be an unobservable individual variable in this model, which can be approached through 𝑒𝑡. Since Grossman (1972) implies that spending on medical care is not the only investment individuals can do to keep their health level decent, other factors are added to the model. Job characteristics and employment conditions have an impact on individuals health according to earlier research, therefore they are added to the model as well (Warren et al. (2004), Datta Gupta and Kristensen (2008), Richardson et al. (2012) for example). In addition, other variables such as housing, exercising and leisure time could be part of the model as well.
Combining equation (1) and (2) results in the difference in the stock of health between this period and the next period, which is defined by the investment in health minus the depreciation in health. In formula form:
𝐻𝑖𝑡 + 1 – 𝐻𝑡 = 𝐼𝑡 (𝑀𝑡, 𝑍𝑡, 𝐺𝑡, 𝑒𝑡)– 𝛿𝑡 𝐻𝑡 (3)
The resulting formula shows us that the next periods stock of health (current period) depends on the investment in the current period (former period), the individuals’ job characteristics and working conditions and personal characteristics:
Data and Methodology
Part of Understanding Society data are used for this research, namely the United Kingdom Household Longitudinal Study. This study offers high quality longitudinal data from members of representative households who are followed over many years. The data consists of a large number of subjects, including childhood, work, social life and behavioural aspects. Participants are part of the
households which are followed, including children till elderly. Understanding Society started in 1991 with a survey of British households, the British Household Panel Survey (BHPS). This survey ended in 2009 and was followed by the UKHLS, starting in 2009 with the first wave and ending in 2018 with wave 9. The total survey consists of approximately 40,000 households, from which almost one-fourth was part of the BHPS study as well. Although the UKHLS interviewed the participants annually, even waves differ from the odd ones. Since the interest for this research lies in a combination of personal aspects, working life and health only the even waves are used, resulting in information collected every two years. The model consists of almost 18,000 households and approximately 35,000 individuals aged between 16 and 65 years old1. Note that the UKHLS study includes all four UK countries; England, Wales, Scotland and Northern Ireland.
This paper pursues partly the paper of Robone et al. (2011), which conducted a research of self-assessed health and psychological well-being in combination with contractual conditions and working characteristics. Following Robone et al. (2011), the sample consists only of working employees,
unemployed participants are removed from the sample. The panel is unbalanced and new entrants are part of the survey, leading to a remaining panel of almost 20,000 females and broadly 15,000 men. Reported results are represented by gender, as it makes it easy to distinguish the founded results and it is standard in studies over health. The variables and their definitions are represented in table 1.
Independent Variables
The UKHLS provides rich information about the health of the individual respondent. This paper uses the general health variable (sah) in order to measure the self-assessed health of the individual. The question asked is ‘In general, how would you say your health is…?’, after which the respondent could choose poor, fair, (very) good or excellent. This question is repeated in every wave, creating valuable information over de self-assessed mental health status of each individual through the years. Respondents
Table 1
Variable definitions
Variable Description
Health
Self-reported Health Categorical variable with 1 if poor, 2 if fair, 3 if good or very good, 4 if excellent GHQ Likert scale, with psychological well-being where 0 is the worst and 36 the best
Employment Contracts
Part time job 1 if working part time, 0 if not Temporary job 1 if temporary job, 0 if not
Job Characteristics
No standard hours 1 if working rotation shifts or during evenings/nights, 0 otherwise Overtime hours Number of overtime hours during the week
Work autonomy 1 if a lot/some autonomy over job tasks, 0 if a little or none Working from home 1 if working from home on a regular basis, 0 otherwise Flexible working 1 if flexible working arrangements at work, 0 otherwise
Company size Categorical variable ranging from 1-9, from respectively 1-2 to >1000 employees Managerial duties 1 if manager, supervisor or foreman, 0 otherwise
satisfied with leisure 1 if satisfied with amount of leisure, 0 otherwise dissatisfied with leisure 1 if dissatisfied with amount of leisure, 0 otherwise
Personal Characteristics
Age Age at year of the wave
Divorce 1 if divorced or separated from partner, 0 otherwise Married 1 if married or having a civil partner, 0 otherwise
Widow 1 if widowed, 0 otherwise
White 1 if individual is white, 0 otherwise
Household size Number of people in the household including respondent Children 1 if 1 or more kids in the household
Income Log of gross monthly income in pounds
Education 1 if qualification level of individual is A level (or comparable to A level) or higher Professional work 1 if professional social class at work, 0 otherwise
Technical work 1 if technical social class at work, 0 otherwise
who only participated in one of the waves are excluded from the sample, as the lagged variable of self-assessed health is also part of the model. By adding the lagged variable of self-self-assessed health, concerns about reverse causality are reduced (Gupta and Kristensen, 2008; Robone et al., 2011).
In order to measure the psychological well-being of the individuals, the shortest form of the General Health Questionnaire (GHQ) is used. This questionnaire consists of 12 questions, giving quick, reliable and sensitive information about the individuals overall psychological well-being (Jackson, 2007)2. The respondent answers 12 questions which they can score from 0 to 3, respectively from ‘not at all’ to ‘much more than usual’. In order to combine the overall score of the GHQ, the Likert scale is used: ‘one of the most fundamental and frequently used psychometric tools in educational and social sciences research’ according to Joshi et al (2015). In order to maintain the same range the Likert scale is rescaled, with 0 representing the worst outcome and 36 representing the best. Just as with mental health, the lagged variable of the dependent variable Likert (likert) is added in this model to overcome causality problems (Gupta and Kristensen, 2008; Robone et al., 2011). In order to control for the time it takes to observe the effect of employment conditions and job characteristics on health and well-being, these variables are added in the lagged form as well in both models (Bartley et al. 2004).
By choosing for two models, the difference between individuals’ mental health status and
psychological well-being can be shown. Mental health is reported as here and now, asking individuals at a certain time how they would rate their general health level. The psychological well-being score is
achieved by asking 12 questions, covering anxiety and depression, social dysfunction and loss of
confidence. These questions refer to the past, making individuals think about their average feelings of the past few months.
Dependent variables
For the purpose of looking into the effect of employment conditions on health and well-being, two dummy variables are added to the model; having a part-time job (jbpart) and having a temporary job (jbtemp). Having a part-time job refers to respondents who work less than 30 hours a week. A temporary job refers to respondents who have a job, but not a permanent contract. Both temporary and part-time jobs seem to have an impact on mental health and psychological well-being, according to numerous studies (Rodriquez, 2002; Robone et al., 2011; Bartoll et al., 2019 for example).
The UKHLS offers a lot of information about employment conditions, job characteristics and personal characteristics. By considering earlier research and findings (Robone et al., 2011; Pirani and Salvini, 2014; Warren et al., 2004), a selection of multiple variables covering employer conditions are made. Working in rotation shifts or during the evening/night is covered by the dummy variable ‘no standard hours’(wktime), as this might influence the mental health and psychological well-being status of an individual. This is also the case with overtime hours, which are indicated by the number of overtime hours during a normal week (ovthrs). Next to that, working from home on a regular basis is added to the model as dummy variable (jbhome), as a different work environment might have an impact on
individuals’ work satisfaction, well-being and health. Freedom at work is added to the model with the autonomy variable (jbaut), representing 1 for individuals who have a lot or some autonomy over their work tasks and 0 for individuals who have a few or none. Having flexible working arrangements such as working term-time, working flexi-time and/or job sharing with other colleague(s) are included in a dummy variable (jbflex), just as being classified as a manager, supervisor or foreman (jbmngr).
The effect of part-time employment might be different for individuals who are satisfied with their working hours than from individuals who are not satisfied with the amount of working hours. The
mostly or somewhat satisfied are included in the satisfied variable and individuals who are completely, mostly or somewhat dissatisfied with their amount of leisure are part of the dissatisfied leisure time variable (satlein). Note that there is also a large in-between group, which is neither satisfied nor
dissatisfied with their amount of leisure. In addition, the number of employed at the workplace (jbsize) of the respondent are included as a categorical variable, divided into 1-2, 3-9, 10-24, 25-49, 50-99, 100-199, 200-499, 500-999 and more than 1000 employees.
Personal characteristics
A lot of research is already done regarding mental health and psychological well-being when looking at an individuals’ personal circumstances. These variables are added to the model under the heading personal characteristics, since they serve as control variables in the model. As the model permits for a non-linear relation with self-assessed health and psychological well-being, control variables for age, age squared and age cubic (age, age2 and age3 respectively) are part of the model. Additionally, the model controls for marital status by including being married or having a civil partner (married), being a widow (widow) or being divorced or separated from a partner (divorce). Thereby, control variables are added for race (white), with 1 for white individuals and 0 otherwise and for high education (hiqual), with 1 for education equal/similar to A level or higher and 0 for less. Likewise, a dummy variable to control for children (children) is part of the model, with 1 for having one or more children and 0 for no children. Household size is added as well (hhsize), representing the number of people in the household including the respondent. The logarithm of income (logincome) is added as the monthly gross income of the individual.
individuals’ job and the kind of work they are performing. The UKHLS uses the coding frame from the BHPS to construct a measure of the employment relations and conditions of occupations. This measure (National Statistics Socio-Economic Classification) is closely associated with the social economic group and the kind of work of the individual3. In order to control for the social class of the present job, five dummy variables are added to the model: professional work (workprof) represents positions that cover all types of higher professional work, technical work (worktech) represents positions that cover types of higher technical work, manual skilled work (workskilnm) represents positions that exist of non-manual skilled work, while non-manual skilled word (workskil) represents non-manual skilled work and unskilled work (workunsk) represents positions consisting of unskilled work.
Table 2 shows a summary of the variable statistics, including the mean, standard deviation (SD), the minimum value (min) and the maximum value (max). The panel shows almost equal self-reported health for both genders, just as the general health questionnaire score. Looking at the self-reported mental health variable of this sample, an average of ‘(very) good’ is found for both genders. When looking at psychological well-being, males score on average 1 point higher than females, with an average score of 26 against 25 respectively. In general, both genders score between sufficient and excellent for mental health and psychological well-being.
Table 2
Descriptive statistics
Female Male
N= 19,557 N= 15,290
Variable Mean SD Min Max Mean SD Min Max
Self-reported Health 3.054 .590 1 4 3.081 .574 1 4
GHQ 25.037 5.130 0 36 26.178 4.670 0 36
Part time job .725 .446 0 1 .429 .495 0 1
Temporary job .064 .247 0 1 .049 .216 0 1
No standard hours .243 .429 0 1 .291 .454 0 1
Overtime hours 3.121 5.614 0 1 4.429 6.751 0 80
Work autonomy .707 .455 0 1 .749 .434 0 1
Working from home .148 .355 0 1 .197 .398 0 1
Flexible working .651 .477 0 1 .575 .494 0 1 Company size 5.047 2.395 1 9 5.292 2.398 1 9 Managerial duties .364 .481 0 1 .461 .498 0 1 Satisfied w leisure .526 .499 0 1 .531 .499 0 1 Dissatisfied w leisure .342 .474 0 1 .329 .470 0 1 Age 42.49 10.83 17 65 42.30 10.743 18 64 Divorce .139 .346 0 1 .0785 .269 0 1 Married .568 .496 0 1 .623 .485 0 1 Widow .013 .114 0 1 .005 .072 0 1 White .895 .307 0 1 .883 .032 0 1 Household size 3.036 1.259 0 14 3.133 1.384 1 15 Children .895 .500 0 1 .440 .496 0 1 Income 7.436 .638 1.466 9.726 7.826 .608 .916 9.845 Education .742 .448 0 1 .722 .428 0 1 Professional work .053 .223 0 1 .082 .274 0 1 Technical work .406 .491 0 1 .421 .494 0 1
Skilled work
non-manual .295 .460 0 1 .140 .348 0 1
Skilled work manual .219 .413 0 1 .312 .463 0 1
Unskilled work .023 .150 0 1 .039 .193 0 1
higher with 6,4%, while males score below this border with 4,9%. However, differences in gender appear when looking at having a part-time job, where almost three quarters of the women respond with ‘yes’, while not even half of the males have a part-time job4. When looking at having a temporary job the table reports a few more women compared to men. Furthermore, women tend to have more often 1 or more children, do less professional but more non-annual skilled work and seem to have fewer managerial duties compared to men. When looking at education, 72,4 % of the females and 72,2% of the males have on average a qualification comparable with A level or higher. This is higher than the national average, which is 54,89 according to the OECD. Note that both genders score a standard error of more than 40%. The other variables are representative for the national average of the UK.
Empirical Model
The first model consist of the independent variable health (h*it) and dependent variables
subdivided into the lagged health variable (hit-1), a set of other lagged variables (xit-1), time-invariant elements (αi) and the error term (εit)5. As the health variable is categorical and ordered (‘Excellent, Very
good, Good, Fair or Poor’), a random-effects ordered probit model is used to estimate the relationship between health and the independent variables. Endogeneity bias may be a concern in this model in the form of reverse causality, where health and or well-being influences employment conditions and job characteristics instead of the other way around. In order to overcome this burden the approach of previous literature is followed, by including one period lagged variable of self-assessed health and psychological well-being to the model (Gupta and Kristensen, 2008; Robone et al., 2011). Additional lagged variables are added since there might be a time-lag in noticing the effect of these independent variables on health. The questions asked and the answers given are focused on the current time, the lagged variables allow some time to observe the effect on health. Combining this results in the following formula:
4 .725 versus .429 for females and males respectively, while the standard errors do not vary that much with .446 and .495 respectively.
ℎ
∗𝑖𝑡= 𝛾ℎ𝑖𝑡 − 1
+
𝛽’𝑥𝑖𝑡 − 1+ 𝛼
𝑖+
𝜀
𝑖𝑡 (5)Where i stands for the number of individuals in the sample, ranging from 1 to N individuals and t stands for the number of waves in the survey, ranging from 2 to N. The error term is assumed to be normally distributed and uncorrelated with the set of lagged- and the time-invariant variables. Since there is no natural scale for the latent variable, the variance of the error term is restricted to one. Note that h*it is a latent variable as the variable indicates self-reported health, while the observed variable is hit:
ℎ𝑖𝑡 = 𝑗 𝑖𝑓 𝑚𝑗 − 1 <
ℎ
∗𝑖𝑡 < 𝑚𝑗 (6) Where 𝑗 = 1,..,4, 𝑚0 = −∞, 𝑚4 = +∞ and 𝑚𝑗 − 1 ≤ 𝑚𝑗The probability of observing the particular category of self-assessed health reported by individual 𝑖 at time 𝑡 (ℎ𝑖𝑡) becomes:
𝑃𝑖𝑡𝑗 = 𝑝 (ℎ𝑖𝑡 – 𝑗) = 𝛷 (𝑚𝑗 − 𝛽’𝑥𝑖𝑡 − 1 −
𝛾ℎ𝑖𝑡 − 1 – 𝛼
𝑖) − 𝛷 (𝑚𝑗− 1 − 𝛽’𝑥𝑖𝑡 − 1 −𝛾ℎ𝑖𝑡 −
1 – 𝛼𝑖) (7)
Which is conditional on a set of independent variables and the individual effect, while assuming that the error term is normally distributed.
Since we are dealing with a non-linear categorical dependent variable SAH and the intuition from linear regression models does not hold for nonlinear models, the estimated coefficients cannot be
Results
Self-Assessed Health
The results of the random effects probit regression on self-assessed health are shown in table 3, while the corresponding marginal effects are presented in table 4. For all results the corresponding coefficients and standard errors are shown for females and males, respectively. Note that the results of table 3 only have qualitative content, as the dependent variable SAH for the model of self-assessed health
Table 3
Random effects model for self-assessed health
Female Male
Coefficient SE Coefficient SE
Part time job .012 .021 -.007 .023
Temporary job .036 .037 -.038 .048
No standard hours -.030 .021 -.030 .023
Overtime hours -.0001 .002 .001 .002
Work autonomy .009 .020 .016 .025
Working from home .029 .027 .021 .030
Flexible working -.009 .021 .0002 .024 Company size .004 .004 -.007 .004 Managerial duties .030 .021 .006 .024 Satisfied w leisure .192*** .028 .122*** .031 Dissatisfied w leisure .001 .028 -.029 .032 Age -.006 .032 -.085** .038 Married .045* .025 .017 .029 Divorced -.024 .032 -.024 .044 Widowed -.164** .079 .270* .147 White .087*** .029 .018 .033 Household size .026*** .009 .008 .010 Children .002 .025 -.061** .029 Income .083*** .018 .126*** .022 Highly educated .032 .022 .043 .025 Professional work .223* .123 .118 .136 Technical work .113 .118 .073 .132
Skilled work non-manual .097 .118 .034 .133
Skilled work manual .066 .120 .034 .133
Unskilled work .020 .131 .114 .142
With *p < .05. **p < .01. ***p < .001
is non-linear. In order to say something about the magnitude of the coefficients, table 4 is added representing the marginal effects of the variables.
Table 3 and 4 show that having a part-time job or having a temporary job is slightly negative related to self-assessed health for males, while it is slightly positive related to females. When looking at part-time work, this might have to do with the fact that females choose more often for part-time
employment while males prefer to work full-time. Thereby, the standard error is bigger for both genders
Table 4
Marginal effects for self-assessed health
Female Male
Variable Coefficient SE Coefficient SE
Part time job .003 .005 -.002 .005
Temporary job .008 .008 -.0001 .011
No standard hours -.007 .004 -.007 .005
Overtime hours -.0003 .003 .0004 .0003
Work autonomy .002 .005 .004 .006
Working from home .006 .006 .005 .007
Flexible working -.002 .005 .0001 .005 Company size .001 .001 -.001 .001 Managerial duties .006 .005 .001 .005 Satisfied w leisure .043*** .006 .027*** .007 Dissatisfied w leisure .0001 .006 -.007 .007 Age -.001 .001 -.019* .008 Married .010* .006 .004 .007 Divorced -.005 .007 -.005 .010 Widowed -.036* .018 .061* .033 White .019*** .007 .004 .007 Household size .006*** .002 .002 .002 Children .001 .005 -.014* .007 Income .018*** .004 .028*** .005 Highly educated .007 .005 .009* .006 Professional work .051* .027 .027 .031 Technical work .025 .026 .017 .029
Skilled work non-manual .021 .026 .013 .030
Skilled work manual .015 .026 .008 .030
Unskilled work .004 .029 .026 .032
With *p < .05. **p < .01. ***p < .001
and no significant relationship is found, implying little or no relationship in this sample (0,3% for females while -0,2% for males, while both have a standard error of 0,5%). This holds also for the temporary job coefficients where a negative sign was expected for both genders, but is only found for males. Females seem to worry less about their temporary employment, which might be due to males earning the cost (0,8% for females with a similar standard error, while -.01 for males with a standard error of 0,11%).
Working from home, having more autonomy at work and being a manager, foreman or supervisor result in a more favourable health outcome as expected. These positive health results can be explained by more freedom and options to cope with work, as there is room for arranging time more appropriately when working at home for example. Also consistent with earlier findings is the health effect on
employees who work during the evening/night or in rotation shifts, which result in an unfavourable health outcome for both genders. The effect of working in a firm with more employees results for women in higher self-reported health, while in adversely a negative result is found for men. This is in contrast to earlier findings, where the effect was slightly negative for females and just positive for males. Although these findings differ, very small effects are found in general when looking at firm size effect.
For both females and males is a positive, significant relationship found for satisfaction with the amount of leisure and self-assessed health. Females tend to report a higher self-assessed health level when they are satisfied with the amount of leisure compared to men, with a 4,3% against 2,7% higher health level respectively. As explained by Grossman (1972), leisure is part of the factors affecting the stock of health and can be used to invest in health. In contrast, males who are unsatisfied with their time of leisure report a worse health outcome with 0,7%, but with a standard error of also 0,7%. When looking at
Psychological Well-Being
The results of the random effects probit regression on psychological well-being are shown in table 5. Some results correspond to the outcomes of self-assessed health, but some interesting differences are found as well. At first sight, it is remarkable that for males more differences are found when comparing with self-reported health. This holds mostly for the personal characteristics variables. When looking at the
Table 5
Random effects model for psychological well-being
Female Male
Variable Coefficient SE Coefficient SE
Part time job -.022 .017 .007 .018
Temporary job .024 .031 .005 .040
No standard hours .034* .018 .011 .019
Overtime hours -.001 .001 -.002* .001
Work autonomy -.011 .017 -.001 .020
Working from home -.019 .017 .016 .025
Flexible working .010 .017 .004 .020 Company size .002 .003 .008** .004 Managerial duties -.016 .018 .010 .002 Satisfied w leisure -.14*** .024 -.136*** .025 Dissatisfied w leisure .053** .024 .051* .027 Age -.008 .002 .047 .031 Married -.041* .021 -.062** .025 Divorced .041 .028 0.020 .037 Widowed .023 .025 -.134 .120 White .031 .025 .074*** .027 Household size -.018** .008 .002 .008 Children .012 .020 .031 .024 Income .008 .015 .053*** .018 Highly educated .026 .018 .020 .021 Professional work .073 .103 -.104 .113 Technical work .101 .097 -.146 .120
Skilled work non-manual .112 .099 -.088 .111
Skilled work manual .090 .099 -.168 .110
Unskilled work .111 .110 -.188 .118
With *p < .05. **p < .01. ***p < .001
different results contractual conditions and job characteristics, I find some notable different outcomes for both genders.
Starting with employment conditions, we find that males tend to score better well-being when having a part-time or, surprisingly, a temporary job (0,7% and 0,5% respectively). Women score a worse psychological well-being result when working part-time, in contrast to the mental health model (-2,2% against 0,03%). The better work-life balance due to part-time working does not overrule the negative effects, which may consist of fewer financial and career benefits and less employment protection. Women with a temporary job show again no worse outcome, expecting them to worry less about their contract (2,4%). However, no significant results are found regarding contractual conditions and large standard errors are shown. This indicates that there is no evidence in this sample for a strong correlation between having a part-time job or being temporarily employed and psychological well-being.
Looking at job characteristics, some remarkable findings show up when comparing with self-assessed health. Working in rotation shifts or during evenings/nights results in a higher well-being score for both genders, being significant for women. They reach a higher psychological well-being of 3,4% when working in rotation shifts, against 1,1% for males. This shift in sign might be due to better conditions regarding rotation shifts, leading to less demanding work. Working more overtime hours results for females in a 0,1% and for males a 0,2% lower well-being, being significant for males. Particularly working more hours when already working a standard amount of hours might be
unfavourable for the state of mind, demanding more from the individual at the cost of own-time. While women show a small positive effect for both models regarding the size of the firm with 0,1% and 0,2% respectively, men report a positive and significant outcome when looking at psychological well-being in contrast to mental health. Working at a larger firm increases their well-being with 0,8%, which might be due to more social talks and support.
outcome of the mental health model, where both genders’ health outcome is positive when satisfied with their leisure. As health is a state of current being, psychological well-being is seen as a state of current living. This difference may partly explain the effect of leisure on both states, as currently being satisfied with the amount of leisure increases your state of mental health. However, an individuals’ psychological well-being refers to a more complex relationship regarding satisfaction with the amount of leisure as this refers to the state of living. If an individual has not achieved a balance between challenging and
Conclusion and Recommendation
This research looks at the effect of employment conditions and job characteristics on self-reported health and psychological well-being among employees of the UK. The rich source of information comes from the United Kingdom Household Longitudinal Study (UKHLS), from which four two-year waves are used covering data from 2010 until 2018. No evidence is found regarding the impact of contractual conditions on mental health and psychological well-being in this sample. However, in line with earlier studies this research shows that job characteristics have an impact on individuals’ health and well-being. This information might be useful to improve the mental health level and psychological well-being of employees by closely looking at certain job characteristics. Different outcomes for males and females reveal different preferences for each gender.
Interesting outcomes show up when looking at the effect of job characteristics on both health and well-being. Satisfaction with leisure appears to increase the mental health level of females with 4,3% and 2,7% for males. In contrast, lower well-being scores arise for females and males who are satisfied with the amount of leisure when looking at psychological well-being, decreasing the score with 14% and 13,6% respectively. Thereby, females and males who are unsatisfied with the amount of leisure score a higher well-being score with 5,3% for females and 5,1% for males. More colleagues appear to have a positive effect on well-being, with a significant result for males of 0,8% higher well-being score when working at a larger firm. This result was negative for males when looking at mental health (-0.1%) and slightly positive for females in both the mental health and well-being model (with 0,1% and 0,2%
respectively). Working in rotation shifts increases females’ well-being score with 3,4%, while increasing it with 1,1% for males. This result was negative for both genders in the mental health model (-0,7% for each). No or little evidence of the impact on health and well-being was found for the other job
characteristic variables in this sample.
on health, I choose to conduct a more general research which can be used more broadly. In addition, by comparing self-reported health and psychological well-being in one study, differences are easily
highlighted among these kinds of feelings. This also holds for the differences between females and males, for both health and well-being. As far as I am aware, recent literature using UKHLS panel data did look into health or well-being and certain work related effects, but not specific to employment conditions and job characteristics. Thereby, this study pursues partly the paper of Robone et al. (2011), who investigated the impact of contractual conditions and working conditions on the health and well-being of employees by using the British Household Panel Study. By using the subsequent study of the Society Survey for this research, more accurate information can be found about employees and their vision on health and well-being. Differences in results can be used to understand the transformation and preferences of the labor force, for both males and females.
Important shortcomings arise in this study. First of all, no evidence for the effect of contractual conditions on mental health and psychological well-being is found for this sample. This does not imply that this holds for each country, industry or sample. Further investigation is necessary in order to make strong conclusions about the impacts. Secondly, the study is based on population-averaged effects of employment conditions and job characteristics on self-assessed mental health and psychological well-being. Therefore, the results found in this research refer to the average UK resident and the effects on specific employees may vary widely. Thirdly, the focus of this study lies on the mental health and well-being of employees by looking at the past twelve months of their working life. This relative short-term view may not hold for the longer term, where the impact might be different or even vanished. Additional research should be done in order to find the effects of employment conditions and job characteristics in the long run. Also, different results may be achieved for different countries, which may be due to cultural differences or economic development state dissimilarity for example.
moment, due to nitrogen rules, higher focus on animals rights and farmer protests. However, the data was not sufficient enough to work this out. A short research on industries can be found in appendix 3. Next to that, I wanted to include interaction terms as well (for example working part-time * having children) since this would have broadened the research. Due to the discussions of economists about the purpose of
interaction terms and the possible calculation problems, I decided to leave them out.
References
Bartley, M., Sacker, A., Clarke, P. (2004). Employment status, employment conditions, and limiting illness: prospective evidence from the British household panel survey 1991-2001. Journal of Epidemiology & Community Health, 58, 501-506. https://doi.org/10.1136/jech.2003.009878.
Bartoll, X., Gil, J., Ramos, R. (2019). Temporary Employment, Work Stress and Mental Health Before and After the Spanish Economic Recession. International Archives of Occupational and
Environmental Health, 92, 1047-1059. https://doi.org/10.1007/s00420-019-01443-2.
Beham, B., Drobnic, S., Präg, P., Baierl, A. (2019). Part-time work and gender inequality in Europe: a comparative analysis of satisfaction with work-life balance. European Societies 21(3), 378-402. https://doi.org/10.1080/14616696.2018.1473627.
Burgard, S.A., Katherine, L.Y. (2013). Bad jobs, bad health? How work and working conditions contribute to health disparities. American Behavioural Scientists, 57(8), 1105-1127.
https://doi.org/10.1177/0002764213487347.
Cappelli, P., Bassi, B., Katz, H., Knoke, D., Osterman, P., Useem, M. (1997). Change at work. Oxford University Press.
Cottini, E., Lucifora, C. (2013). Mental Health and Working Conditions in Europe. ILR Review, 66(4), 958-988. https://doi.org/10.1177/001979391306600409.
Grossman, M. (1973). On the concept of health capital and the demand for health. Journal of Political Economy, 80(2), 223-255. https://doi.org/10.1086/259880.
Jackson, C. (2007). The General Health Questionnaire. Occupational Medicine, 57(1), 79. Hhtps://doi.org/10.1093/occmed/kql169
Joshi, A., Kale, S., Chandel, S., Pal, D.K. (2015). Likert scale: Explored and explained. British Journal of Applied Science & Technology, 7(4), 396-403. https://doi.org/10.9734/bjast/2015/14975.
Joyce, K., Pabayo, R., Critchley, J.A. (2010). Flexible Working Conditions and Their Effects on Employee Health and Wellbeing. Cochrane Systematic Review, 2.
https://doi.org/10.1002/14651858.cd008009.
K. Wilhelm, V. Kovess, C. Rios-Seidel, A. Finch. (2004). Work and Mental Health. Social Psychiatry and Psychiatric Epidemiology, 39, 866-873. https://doi.org/10.1007/s00127-004-0869-7
Karatuna, I., Basol, O. (2017). Job satisfaction of part-time vs. full-time workers in Turkey: the effect of income and work status satisfaction. International Journal of Value Chain Management, 8(1), 58-72. https://doi.org/10.1504/ijvcm.2017.10003559.
LaMontagne, A.D., Milner, A., Krnjacki, L, Kavanagh, A.M., Blakely, T.A., Bentley, R. (2014). Employment Arrangements and Mental Health in a Cohort of Working Australians: Are Transitions From Permanent to Temporary Employment Associated With Changes in Mental Health? American Journal of Epidemiology, 179(12), 1467-1476. https://doi.org/10.1093/aje/kwu093.
Lepinteur, A. (2019). The shorter workweek and worker wellbeing: Evidence from Portugal and France. Labour economics, 58, 204-220. https://doi.org/10.1016/j.labeco.2018.05.010.
Loscocco, K.A., Spitze, G. (1990). Working Conditions, Social Support and the Well-Being of Female and Male Factory Workers. Journal of Health and Social Behavior, 31(4), 313-327.
https://doi.org/10.2307/2136816.
Messenger, J.C. (2010). Working time trends and developments in Europe. Cambridge Journal of Economics, 12(2), 295-316. https://doi.org/10.1093/cje/beq022.
Montero, R., Rau, T. (2015). Part-time work, job satisfaction and well-being: evidence from a developing OECD country. The Journal of Development Studies, 51(4), 370-385.
https://doi.org/10.1080/00220388.2014.963567.
Norton, E.C., Wang, H., Ai, C. (2004). Computing interaction effects and standard errors in logit and probit models. The Stata Journal, 4(3), 154-167. https://doi.org/10.1177/1536867x0400400206.
Norton, E.C., Ai, C. (2003). Interaction terms in logit and probit models. Economics Letters, 80(1), 123-129. https://doi.org/10.1016/s0165-1765(03)00032-6.
OECD (2010-2018), Adult education level. https://data.oecd.org/eduatt/adult-education-level.htm.
OECD (2010-2018), Part-time employment rate. https://data.oecd.org/emp/part-time-employment-rate.htm.
OECD (2010-2018), Temporary employment. https://data.oecd.org/emp/temporary-employment.htm#indicator-chart.
Pirani, E., Salivini, S. (2015). Is Temporary Employment Damaging to Health? A Longitudinal Study on Italian Workers. Social Science & Medicine, 124,
121-131. https://doi.org/10.1016/j.socscimed.2014.11.033.
Richardson, S.S., Lester, L.H., Zhang, G. (2012). Are Causal and Contract Terms of Employment Hazardous for Mental Health in Australia? Journal of Industrial Relations, 54(4), 557-578. https://doi.org/10.1177/0022185612454974.
Robone, S., Jones, A.M., Rice, N. (2011). Contractual Conditions, Working Conditions and Their Impact on Health and Well-being. The European Journal of Health Economics, 12(5),
Rodriquez, E. (2002). Marginal employment and health in Britain and Germany: does unstable employment predict health? Social Science & Medicine, 55(6), 963-979.
https://doi.org/10.1016/s0277-9536(01)00234-9.
Satuf, C., Monteiro, S., Pereira, H. (2018). The protective effect of job satisfaction in health, happiness, well-being and self-esteem. International Journal of Occupational Safety and Ergonomics, 24(2), 181-189. https://doi.org/10.1080/10803548.2016.1216365.
Silla, I., Gracia, F.J., Peiró, J.M. (2005). Job Insecurity and Health-Related Outcomes among Different Types of Temporary Workers. Economic and Industrial Democracy, 26(1), 89-117.
https://doi.org/10.1177/0143831x05049404.
Smith, M., Fagan, C., Rubery, J. (2002). Where and why is part-time work growing in Europe? Routledge.
Warren, J.R., Hoonakker, P., Carayon, P., Brand, J. (2004). Job characteristics as mediators in SES-health relationships. Social Science & Medicine, 59(7), 1367-1378.
https://doi.org/10.1016/j.socscimed.2004.01.035.
Appendix 1
This appendix contains the data construction, explaining the road from original raw data to the sample used in the models. The original sample consisted of approximately 40,000 households, broadly 77,000 individuals, representing 41,500 females and 35,600 males. The data clean-up started with finding all the necessary variables, including self-reported health, GHQ 12, employment conditions, job
characteristics and personal characteristics. After adding them to a new file all the missing, unknown and inapplicable variables were excluded. In addition, the variables were cleaned up in order to meet the requirements; keeping only individuals aged 16-65 years old, generating the dummy variables, reshaping the self-reported health and GHQ 12 variables and adding age squared and cubic age variables.
Thereafter, the lagged variables were created. This resulted in a loss of approximately 20,000
Appendix 2
Table 6
Dependent variables
Variable Question
Self-reported health In general, how would you say your health is? GHQ 1. Lost sleep over worry
2. Constantly under strain 3. Unhappy or distressed
4. Could not overcome difficulties 5. Able to concentrate
6. Play useful part in things 7. Capable of making decisions 8. Face up to problems
9. Enjoy day-to-day activities 10. Reasonable happy 11. Losing confidence in self 12. Thinking of self as worthless
Table 6 represents the self-reported health variable question, as well as the 12 questions
concerning the GHQ 12 (psychological well-being) variable. These 12 questions cover the topics anxiety and depression, social dysfunction and loss of confidence, in order to get a clear view on the individual’s psychological well-being status. The answers individuals could choose regarding the health question are; poor, fair, (very) good or excellent. The answers individuals could give regarding the GQH 12 are a score from 0 to 3, ranging from ‘not at all’ to ‘much more than usual’. These scores are added together,
Appendix 3
The main research of this paper controls for doing professional, technical, non-manual skilled, manual skilled and unskilled work. However, there are other possible factors who are intertwined with employees and their contractual and working conditions who may influence mental health and
psychological well-being. Next to the kind of work which the employee is doing, the sector in which the firm of the respondent is operating is informative as well. Different collective labor agreements between sectors have direct influence on contractual agreements and job characteristics. Thereby, certain branches attract a certain type of people, having different priorities than individuals of other segments.
The extended model in this appendix consists again of the two dependent variables self-assessed health and psychological well-being. Self-assessed health is reported by answering the question ‘In general, how would you say your health is…?’, after which the respondent could choose ‘Excellent, Very good, Good, Fair or Poor’. In order to reduce concerns about reverse causality, the lagged variable of both SAH and psychological well-being are added to the model, just as in the original research. Psychological well-being is measured by using again the shortest form of the General Health Questionnaire, giving quick reliable and sensitive information about the individuals overall psychological well-being (Jackson, 2007). The score of each question from 0 to 3 (‘not at all’ to ‘more than usual’) are converted into the Likert scale, where 0 presents the worst outcome and 36 the best. In both models the lagged version of the independent variable is added, in order to reduce concerns about reverse causality. Employment condition variables and job characteristic variables are also part of the model in the lagged version, since it takes time to notice the effect on mental health and psychological well-being.
In order to find out if the sector where the firm/organization of the respondent is active has an effect on mental health and psychological well-being, the model is extended by adding different sectors. The data are collected by the UKHLS from wave two onwards, using the Standard Industrial
mainly make or do at the place where you work?’ The answers are merged and divided into 11 sectors, added as dummy variables to the model. The sectors, as well as the percentage of the individuals of this model who are part of that sector, are shortly described in table 7. Note that the ‘other sector’ variable is omitted because of collinearity. A collinearity test shows that these industries are not highly correlated to the control variables of kind of work. Just as with employment conditions and job characteristics are the sector dummies added in the lagged form, to observe the effect on mental health and psychological well-being.
Table 7
Sector distribution
N= 19,316 N= 14.949
Variable % Female % Male
Agricultural 0,22 0,65 Mining 0,14 0,53 Manufacturing 4,48 16,44 Electrowater 0,87 2,94 Construction 1,30 6,70 Sales 13,22 13,16 Transport 6,29 10,37 Services 25,3 32,04 Education 18,00 7,26 Socialwork 27,16 7,02 Cultural 2,86 2,69
Remaining % consists of 'other sectors'
Table 8
Random effects model for self-assessed health and psychological well-being
Self-assessed health Psychological well-being
Females Males Females Males
Coefficient SE Coefficient SE Coefficient SE Coefficient SE Agricultural -.036 .215 .027 .226 -.176 .179 -.414** .192 Mining -.226 .288 .195 .252 .014 .243 -.232 .212 Manufacturing -.098 .179 .219 .211 -.080 .147 -.231 .179 Electrowater .226 .197 .280 .217 -.080 .164 -.288 .184 Construction .188 .187 .212 .211 -.070 .156 -.238 .180 Sales .032 .174 .223 .211 -.112 .145 -.247 .179 Transport .087 .176 .237 .212 -.060 .146 -.214 .180 Services .069 .173 .195 .210 -.065 .143 -.212 .179 Education .029 .174 .198 .213 .011 .144 -.146 .181 Socialwork .015 .173 .166 .213 -.036 .143 -.099 .180 Cultural .039 .179 .154 .217 -.093 .148 -.309* .184 Age -.003 .032 -.095*** .035 -.022 .027 .065** .029 Married .053** .024 .040 .027 -.048** .020 -.056** .023 Divorced -.017 .031 -.004 .040 .034 .026 -.012 .033 Widowed -.162** .077 .210 .132 .021 .066 -.209* .109 White .098*** .028 .012 .030 .026 .024 .093*** .025 Household size .031*** .009 .007 .009 -.020*** .029 .005 .008 Children -.010 .024 -.052* .027 .011 .020 .019 .022 Income .079*** .016 .092*** .016 -.007 .013 -.032 .014 Highly educated .050** .021 .049** .023 -.045** .018 .006 .019 Professional work .319*** .117 .153 .130 -.012 .097 -.136 .108 Technical work .189* .112 .106 .127 .022 .097 -.156 .105 Skilled work nm .148 .112 .076 .128 .060 .093 -.086 .107 Skilled work man. .131 .113 .055 .127 .008 .094 -.164 .106 Unskilled work .088 .021 .116 .137 .025 .105 -.167 .114 Part time job .009 .021 -.008 .023 -.021 .017 -.005 .019 Temporary job .034 .035 -.023 .040 .020 .029 .044 .033 No standard hours -.018 .020 -.028 .021 .031* .017 .014 .018 Overtime hours -.001 .001 .001 .001 -.001 .001 -.003** .001 Work autonomy .018 .020 .031 .024 -.011 .017 -.009 .020 Working home .014 .028 .036 .030 -.009 .023 .030 .025 Flexible working -.007 .021 .002 .024 -.002 .017 -.006 .020 Company size .001 .002 -.007*** .003 .004* .002 .004* .002 Managerial duties .017 .021 .006 .023 -.006 .018 .014 .019 Satisfied w leisure .117*** .026 .121*** .028 -.146*** .022 -.125*** .023 Dissatisfied w leisure -.008 .028 -.019 .030 .049** .023 .046* .025 With *p < .05. **p < .01. ***p < .001
