Unexpected employment shocks and their influence on health and time use

(1)

LEIDEN UNIVERSITY

Unexpected employment

shocks and their influence on

health and time use

A quantitative research on the effects of employment shocks on (mental) health

and leisure time allocation in The Netherlands

Joshua K eet

20-7-2020

Joshua Keet (s1526251)

Supervisor: Jim Been (second reader: Eduard Suari-Andreu) Faculty of Governance and Global Affairs

(2)

1 Abstract

This research aims to find the difference between unemployment and unexpected unemployment, when it comes to the effects to health. To look further into the mechanisms the analysis also focuses on the allocation of leisure time. Employing a quantitative research design using the LISS panel data from The Netherlands, this research exploits the (un)employment probabilities of the respondents. The wide array of health and time use categories allows to analyze the effects of (unexpected) job loss on both physical and mental health and behaviour. The results suggest that unexpected unemployment deteriorates health more compared to unemployment. Looking at the allocation leisure time, the unexpected unemployed watch more television and allocate less time to home production.

Acknowledgements

I would like to thank Jim Been for all the advices and supervision during this research, I will never forget our fantastic (online) meetings. With this thesis as the concluding piece of my studies, I would also like to thank the teachers of the study of Public Administration who have taught me in these years.

Special thanks to Emiel de Groot for reading and giving text-suggestions along the way. And lastly, I would like to thank my partner Evianne Brouwers. Working and studying closely side by side during this period only brought us closer together.

(3)

2

1. Introduction

Do we work to live, or do we live to work? One might say the first. We work in order to afford to buy the necessities of life like shelter and food. After we have obtained the essentials, we start buying the things that make us happy. So, a job provides us with both the essentials in life and more. As Maslow (1943) would put it, our need to survive is at the core of motivation. Hence, we are motivated to work in order to survive. However, we allocate a large amount of time in our lives to working. Consequently, the psychological impact of working is therefore much bigger than just providing us with the things we need. A job also links people to the ‘outside world’, because people who work usually report a larger feeling of connection to the economic and social welfare of their communities (Blustein, 2006; Bowe, Bowe & Streeter, 2000). Being employed also fulfills the need of self-determination. Self-determination refers to the feeling of authenticity. This feeling is usually characterized by the experience of control over the direction of one’s life (Ryan & Deci, 2000). To say we live to work might be too extreme, but to a high extent we find satisfaction in having a job. This feeling of satisfaction we get from working and contributing to society plays a large role in our health, especially our mental health (Faragher, Cass & Cooper, 2005). The feeling of satisfaction can also be described as the opposite of the monetary value of work, namely non-monetary value of work. Both monetary and non-monetary value of work have an influence on your health.

The connection between health and having a job also raises questions. If a person is unemployed, do we observe a deteriorating health of this person both physically and mentally? The short answer is yes. Results suggest that job loss has a negative impact on a person’s health (Michaud, Crimmins & Hurd, 2016). This negative effect further strengthens the notion that we indeed, to some extent, live to work. With that in mind, we can go a step further. What if a person unexpectedly loses their job due to, for example, a recession, do we still see a negative effect on this person’s health? And if there is a negative effect, is this effect bigger because it is a sudden unforeseen job loss? It is also equally interesting to study what the effect on health is if a person unexpectedly keeps their job. Does their health suddenly increase, both physically and mentally? This thesis focuses on this unexpected job loss or job keep and its effect on (mental) health and leisure time. To further analyze the potential mechanisms at work, it is also interesting to look at how these persons allocate their time in case of unexpected job loss or job keep. Does

(6)

5 unexpected job loss means more time is allocated to home-production and sports? And when people unexpectedly keep their job, do they allocate their time differently?

Because this thesis is written in times of the Covid-19 pandemic, it is appropriated to treat this crisis as an example of an unexpected way to lose one’s job. Within three weeks after the outbreak in the United States, 16 million people lost their job (The Guardian, 2020). Since the virus was spreading rapidly, these people will not have foreseen that they would lose their jobs. The United States is not the only country that suffers from a sudden large amount of unemployment due to Covid-19. Thus, due to this pandemic and the predicted recession that might follow it, it becomes more relevant to look what unexpected job loss means for these people their health (Goodman, 2020).

To look further upon the effects of unexpected job loss and job keep on (mental) health and time allocation, this thesis uses a central question:

What is the effect of unexpected job loss or job keep on health, both physically and mentally, and how can this be explained by time allocations?

1.1. Method of data collection and analysis

This research employs a quantitative research design. The data is derived from the LISS (Longitudinal Internet Studies for the Social sciences) panel administered by CentERdata (Tilburg University). The LISS panel is a representative sample of Dutch individuals who participate in monthly internet surveys. The panel is based on a true probability sample of households drawn from the population register. Households that could not otherwise participate are provided with a computer and internet connection. A longitudinal survey is fielded in the panel every year, covering a large variety of domains including work, education, income, housing, time use, political views, values and personality. This thesis uses the core studies about health, income and social inclusion & leisure from the LISS panel to conduct quantitative research on.

(7)

6 The method of this research is a quantitative regression analysis. With help from the panel data, it is possible to estimate causal effects. Within these regression there is a differentiation made between expected and unexpected shocks of (un)employment. This method is comparable to the method of Stephens (2004), he only used this method on consumption whereas this research uses it on health and allocation of leisure time. As Stephens (2004) already showed, he finds different effects for unexpected (un)employment shocks in comparison to expected shocks. It therefore is highly interesting to differentiate between those groups. Because it seems that expectations are important for consumptions levels. That is why this research looks if the same results are found in The Netherlands for health and allocation of leisure time. Besides these regressions, it is also interesting to see whether the results are homogeneous or differ for different groups. The different groups this research focuses on are mentioned in paragraph 3.2.

1.2. Academic relevance

Many studies show that the unemployed have worse physical and mental health compared to those who are employed. These studies do so a few subjective and/or objective measures, (Clark and Oswald, 1994; Blanchﬂower, 1996; Korpi, 1997; Winkelmann and Winkelmann, 1998; Laporte, 2004; Hamilton, Merrigan & Dufresne, 1997). This thesis is academically relevant because it combines a variety of subjective and objective measures to come to a more overall image of the relation between unemployment and health and allocation of leisure time.

It is very plausible to accept that unemployment affects an individual’s health. However, there might also exist a reverse causality, or health and employment status might be caused by unobserved factors such as genetics, culture or lifestyle. Researches in the literature of economics typically attempt to solve this reverse causality problem by exploiting exogenous unemployment events, such as ﬁrm closures or mass lay-offs (Bockerman and Ilkmakunnas, 2009; Kassenboehmer and Haisken-DeNaw, 2009; Salm, 2009; Schmitz, 2010). This thesis contributes to this existing literature by not using this exogenous unemployment events, but by taking a different kind of approach. It uses the expectations of people to look at unexpected job loss, normally caused by these exogenous unemployment events. By using expectations the findings will be more generalizable than firm closures. You take a more representative selection of the population, and this increases the generalizability of the conclusions.

(8)

7 This thesis conducts further research on the topic of unexpected job loss, as mentioned by Stephens (2004). This research is about the effect from unexpected job loss on a person’s consumption. This effect on consumption, already has an effect on someone’s physical and mental health. As previously stated, consumption which is health related declines. This could lead to postponing costly operations and other medical care, which lead to a decreasing physical health. Also, the loss in income and therefore not consuming what one want to consume, can lead to a decrease in mental health. So, less consumption caused by job loss has an effect on a person’s physical and mental health. This thesis acknowledges this, and aims to broaden this view by looking directly at a person’s health both physical as mental. By doing this, both the decrease in consumption as the psychological aspects of losing one’s job are incorporated in the results. To bring both these aspects together for unexpected job loss and job keep is new and therefore fill a gap in the literature.

Additionally, most studies on the eﬀect of job loss on health behaviour suﬀer from various shortcomings. The studies focus on correlations and concentrate on case studies, but do not deal with the endogeneity of job loss, or do not draw on longitudinal data (Roelfs, Shor, Davidson & Schwartz, 2011). This thesis not only focuses on the causality between unexpected job loss and health, but also attempts to reveal certain mechanisms between the spending of leisure time of the employed and unemployed. This inclusion of the allocation of time may be an important input factor in the production of health, according to Grossman (1972). The combination of using expectations to predict economic behavior (unemployment & health investment) and the time-use component gives a unique insight into the way the allocation of time, health investment and health interact with each other in case of unexpected job loss or job keep.

1.3. Societal relevance

From the perspective of society, this research is relevant because it looks beyond the direct costs of unemployment such as unemployment insurance (UI) benefits. In addition, there is a connection between health insurance and one’s consumption on health. This thesis combines these two institutions. On the one side it looks indirectly at the effects the UI and health insurance in The Netherlands, because those two institutions ensure that the unemployed is still able to consummate at a certain level. Why this thesis looks indirectly at this effect, is because only data from The Netherlands is included. This means there are no different levels of UI and

(9)

8 extreme different levels of health insurance incorporated in the data. A comparison with other results from the United States can shine more light on this subject.

On the other side, this research showcases some direct effects on the costs of health care for society. When people become less healthy, they need more health care and this costs more for society. When people unexpectedly lose their job and with it a larger deterioration of their health, it can bring more costs to society than initially thought. Especially during recessions, or the current Covid-19 pandemic, a large amount of people lose their job without a chance to anticipate and prepare for it. Knowing this, a government can anticipate even higher health care costs during a recession or a pandemic than initially thought.

1.4. Structure thes is

This thesis is structured as follows. In chapter 2, the relevant literature, a theoretical and an institutional framework are provided. Chapter 3 explains the methodological choices that are made. Chapter 4 entails the analysis with the results from the regressions. And lastly, chapter 5 provides a discussion of the results and a conclusion in which the research question is answered. Additionally, limitations, policy recommendations and further research possibilities are also discussed in chapter 5.

(10)

9

2. Literature review

This literature review is composed as follows. To start off, the literature on job loss and income and the influences on health investments are described briefly. Then the effects of unemployment on health is discussed more elaborately. Health is distinguished in two categories: physical and mental health. After the distinction between physical and mental health, it is appropriate to discuss the literature that combines the effects of unemployment on both physical and mental health. Then, the effects of job insecurity and income volatility on health gives more insight in the unexpected job loss effects on physical and mental health. And to see more of the mechanisms underlying the effects, literature on the time-use of the unemployed is discussed. Then the theoretical framework brings the described literature together and provides some hypotheses. This section concludes with theoretical influence from institutions and an institutional framework of the Netherlands related to the health of the unemployed.

2.1. Effects of une mployment on health investment

Since the focus of this research is on the non-monetary value of work, the monetary value of work and with it the effects of unemployment on health investment are described briefly. Ando & Modigliani (1963) came up with the “life cycle” hypothesis of saving, which suggests that we smooth consumption over our life time to maximize our utility. The life cycle model then shows that in our early life we loan to consume more and we save when we get older to make sure we can consume when we are retired. But, this life cycle can be interrupted with unemployment which means private savings or loans need to cover for this period or consumption has to go down. Which plan is adopted is very dependent on personal circumstances, your earnings history and tenure are important factors in this decision (Jacobson, LaLonde & Sullivan, 1993). Indeed, Cutler & Lleras-Muney (2007) find a causal impact from income to health which opts that most of the time investments in health is dropped. Kristensen & Andersen (2018) nuance this finding by looking at different health services, and conclude that it depends on whether user charge is needed or not. This means that private health expenditure versus public health expenditure also plays a part here. To look at the private health expenditure, the Grossman model in paragraph 2.7. shines more light on the subject. The public health expenditure is discussed through the health insurance in The Netherlands in paragraph 2.8.3. Aguiar & Hurst (2005) add to the literature that the unemployed substitute this consumption loss with more home production.

(11)

10

2.2. Effects of une mployment on phys ical health

If one loses his or her job, this has a deep effect on one’s life. Because unemployment does not only reduces income in both the short and long run (Jacobson, Lalonde & Sullivan, 1993), but also results in a variety of other negative life issues. Examples of this are decreased life satisfaction (Knabe & Rätzel, 2011), increased risk of divorce (Charles & Stephens, 2004) and further job losses (Stevens, 1997), as well as negative consequences for the children of those affected (Lindo, 2011). On top of this, there is a large amount of evidence suggesting that when you are unemployed, you tend to have more unhealthy behaviours than those who are employed (Henkel 2011; Roelfs et al., 2011). However, it is not very clear whether this evidence represents the causal effect of unemployment, or a reverse causal effect, or a spurious correlation (Marcus, 2014).

To dive deeper in whether there is a causal effect between unemployment and physical health, Marcus (2014) estimates the effect of involuntary job loss on smoking behaviour and body weight using German data. To start with smoking behaviour, Marcus finds that involuntary job loss leads to an increase in the chance to smoke by 3 percentage points. For the body mass index he finds that unemployment leads to a slight increase of 0.3 kg on average. In comparison to other studies on smoking and body weight (Ruhm 2005; DeCicca & McLeod 2008), the increase in smoking seems rather large, while the increase in body mass index is rather small. The eﬀects on smoking and body mass index are smaller than comparable ﬁndings for the United States (Falba, Teng, Sindelar & Gallo, 2005; Deb, Gallo, Ayyagari, Fletcher & Sindelar, 2011), which according to Marcus (2014) might be explained by the more generous unemployment insurance in Germany.

Another research is conducted by Jónsdóttir & Ásgeirsdóttir (2014). This research looks further into the effect of involuntary and unexpected job loss on body weight. The research is conducted in Iceland and relies on data of 2007 and 2009. The reason what makes this research more interesting for this thesis, is the fact that a recession took place between both measuring points. It is therefore a simulation of people who lost their job unexpectedly, because it is unlikely that many people would predict in 2007 that they would lose their job in the next year. So, to compare the 2007 and 2009 data of unemployed people and their body mass index is, in a way, an indirect measure of the effect of expected versus unexpected job loss on body weight. Of

(12)

11 course, it needs to be noted that this only gives an indication because this distinction was not made in the research itself. Then on to the results themselves. Jónsdóttir & Ásgeirsdóttir find that the effect of job loss during the recession actually decreased weight, especially for women (2014). All the models they used showed a smaller increase in body mass index when a person loses his or her job, this finding is also supported by Ruhm (2000) who concluded that people have more healthy weights in recessions. However, those results are inconsistent with other studies between unemployment and body weight who show decrease in body mass index (Ruhm 2005; Falba et al., 2005; DeCicca & McLeod 2008; Deb et al., 2011; Marcus, 2014). There are also studies that show more unhealthy weights (under- and overweight) are related to recessions (Böckerman et al., 2007; Charles & DeCicca, 2008). So, the economic climate plays a fundamental role in the effect of unemployment on body weight. The literature, however, does not have a clear answer on how a recession affects body weight. But, when studies examine unemployment and involve macroeconomic conditions, they imply that many aspects of health improve in a recession (Ruhm, 2003; Economou, Nikolaou & Theodossiou, 2008).

The literature about the effects of unemployment on health also looks at different ways of measuring health than just body mass index or weight. For example, the relationship between unemployment rates and mortality rates. Sullivan & von Wachter (2009) concluded that there are higher mortality rates for unemployed workers in the United States in the 1980s, with evidence implying that these results are associated with the decrease in income. Unemployment have been found to lead to significantly increases in mortality rates in Sweden and Denmark (Eliason & Storrie 2009; Browning & Heinesen 2012). Eliason & Storrie (2009) found with Swedish register data that the risk of mortality among men increased by 44% during the first four years after they lose their job. Interestingly, they found no effect of unemployment on female mortality.

There are also other ways to measure health in the literature. A great example of this, is the study done by Black, Devereux & Salvanes (2015). Besides smoking, they looked at cholesterol and blood pressure levels of unemployed people in Norway. They found little evidence to suggest unemployment has any effect on blood pressure for men. However, they do find that the cholesterol levels of men are higher, three or four years after they lost their job. Their findings among women, show evidence that blood pressure decreases after unemployment and that these effects last even seven years after they lost their jobs. The cholesterol levels of women who lost

(13)

12 their job, compared to women who did not, show that unemployment leads to higher cholesterol levels for the first four years after displacement.

In the paper of Michaud et al. (2016), the relationship between physical health and job loss is analyzed using biomarker data on a sample of respondents aged 59–70 for whom the authors have detailed job history going back eight to sixteen years. They find no relationship between the closing of businesses and physical health, whether health was measured using biomarkers or self-reported conditions. But, the authors do find a robust effect of layoffs on physical health, measured using biomarkers. A potential mechanism that might explain the difference in results between layoffs and business closures, according to Michaud et al. (2016), is psychological stress. When a single employer is being let go off, this hits the worker harder than when a whole firm is being laid off. This is because the latter is most likely due to reasons that are beyond a single worker’s control, and makes it easier to move on. This therefore brings less psychological stress. Finally, the authors find that the effect of a layoff is stronger for those who expected to remain in the labor market longer. This is especially interesting for this thesis, because it suggests that layoffs that are unexpected and disrupt career plans may have stronger effects on health.

There are also a lot of other different ways to look at the relationship between unemployment and other forms of physical health. Examples of this are sleeping behaviour (Van Cauter & Spiegel, 1999; Ferrie et al., 2007; Gangwisch et al., 2007; Patel & Hu, 2008; King et al., 2008) and drug use (Platt, 1995; French, 2001; DeSimone, 2002).

Regarding sleep behaviour the research of Antillón, Lauderdale & Mullahy (2014) looked at sleeping behaviour in the U.S. and macroeconomic unemployment. They found that when the unemployment in a state increased, so did the time people sleep. Moreover, when state employment increased, the number of people suffering from sleeplessness declined. This is also in general what the literature holds on the relationship between unemployment and sleeping behaviour: you sleep more and are less likely to experience sleep problems when you are unemployed. The relationship between drug use and unemployment is looked upon by Compton, Gfroerer, Conway, & Finger (2014) using data from the U.S. They find that unemployment is associated with illicit drug use and drug use disorders. This is consistent with other literature.

(14)

13

2.3. Effects of une mployment on mental health

Losing one’s job can be stressful. All sorts of psychological reactions to this stressful situation are possible. For instance, one feels uncertainty and a loss of social identity and role (Kasl & Jones, 2000). As discussed earlier, besides income loss there are also non-monetary costs to losing one’s job. So, what are these non-monetary costs? Jahoda’s latent deprivation model (1981 & 1988) suggests that unemployment is very destructive because individuals lose hidden work benefits. These hidden beneﬁts entail ﬁve experiences that are crucial for your mental health: it gives your day time structure; regular interaction with your peers; contact with people who have the same goals; identity and status; and forced activity. The main provider of these experiences is considered to be work. This explains why unemployment brings stress to one’s life, even after controlling for the loss in income.

Clark (2003) suggests that there is even a sixth experience that needs to be considered as a hidden beneﬁt of work. This sixth experience entails that working for a living means complying with the social norm to work. Experimental research suggests that compliance with social norms is a way of gaining social approval. Social approval positively affects the self-related emotions and self-concept (Christensen, Rothgerber, Wood & Matz, 2004; Cialdini & Goldstein, 2004). So, those who do not or cannot comply with the social norm to work are ‘robbed’ of this hidden benefit and might even be exposed to social disapproval. This social disapproval is another contribution that can reduce the well-being of unemployed.

Before we look at some quantitative researches, it is good to take notice of a qualitative one first. To get a deeper understanding of the mechanisms at work here. Björklund, Söderlund, Nyström & Häggström (2015) is a great example of this. Although they only looked at young unemployed men in Finland, the results show universal insights into what one feels when you are unemployed. Björklund et al. (2015) dived this in three categories: 1) slowly losing one’s foothold, 2) a feeling of shame and guilt and 3) flight from reality. To begin with “slowly losing one’s foothold”, the interviewed men find it hard to get any sort of structure and order in their lives. So, the thing they are missing is time and structure. Daytime and nighttime become more fluent for these men, since there is nothing to go to or be in the morning. The second category of “a feeling of shame and guilt”, is about the men not contributing to society. This brings anxiety

(15)

14 about if they will ever find a job. Then there is the loss of identity that comes with unemployment. Respondents feel stress and pressure because they don’t have a career path or big plans in their lives. Additionally, the financial situation of being unemployed provided extra stress, these worries are not about getting the necessities in life, but more about the more luxurious. Then the last category by Björklund et al. (2015) is “flight from reality”. This, among other thing, is about boredom.. This could lead to feeling of meaningless, which again makes the men feeling more sidelined from society. This also led to having no desires to do daily life tasks like cooking or cleaning. Here again, the day has no goal for these men and as a consequence they are just wandering about.

Thus, in theory the relation between unemployment and mental health is very clear. Yet, empirically this is not always the case. Fenwick & Tausig (1994), for example, simultaneously studied the effects of aggregate unemployment and work conditions on “satisfaction” and “stress”. They find that job security, which is lower when the unemployment rate is higher, increases satisfaction but has no effect on stress. Browning, Dano & Heinesen (2006) also find mixed evidence for a causal link of unemployment to ill mental health. Using a large and representative sample of Danish workers, they ﬁnd evidence that there is no impact of unemployment on hospitalization for stress-related diseases for men. Lastly, Björklund (1985) describes why it is hard to look at the effect of unemployment on mental health. He only found a strong interaction effect of age for mental symptoms.

There are also some studies that do show an effect of unemployment on mental health. An example is the research by Gathergood (2013). He estimated the causal impact of unemployment on individual psychological health using an instrumental variable research design on data from the United Kingdom. He finds that, most of the time, the people who become unemployed already showed worse psychological health before they got unemployed. Although this is the case, the unemployment event in itself worsened the individual psychological health. Also interesting to notice is that Gathergood (2013) found that the existence of local unemployment is damping the impact of unemployment on the individual psychological effect. We already saw the theory of local labour markets and their effect on health in the study by Charles & DeCicca (2008). They did not find an effect however of local unemployment rates on body weight, but Gathergood (2013) does find an effect on psychological health. The variation across different

(16)

15 relevance groups is also good to notice, Gathergood (2013) saw that age and gender effects lead to this variation to the impact of unemployment across individual’s psychological health. Another example is a study by Tefft (2011). This study looks at the links between unemployment rates and how many times depression and anxiety is being searched for on the internet. The results are interesting, since Tefft (2011) finds a significant effect between unemployment and the depression search index. Meaning that this is more searched for when the unemployment rate rises. But anxiety is not being significantly searched more when the unemployment rate rises. This again shows that the mechanisms are not easily exposed, and that they can also be of a different direction you would suppose.

The use of medication for anxiety and depression (Frese & Mohr, 1987; Stankunas, Kalediene, Starkuviene, & Kapustinskiene, 2006) is also a popular in the literature on mental health. Álvaro, Garrido, Pereira, Torres, & Barros (2019) researched this phenomenon using data from Spain. They find that unemployment leads to a lower self-esteem, and this is related to more depressive symptoms. Interestingly, this relationship is observed for men, but not for women.

2.4. Effects of une mployment on phys ical and mental health

The separation between physical and mental health and the last two paragraphs is merely a theoretical one. A decline in mental health can also lead to physical outcomes as well. If you look for example what stress for short-term effects has on the physical health and mental health, the intertwining of the effects becomes very clear. As said earlier, losing one’s job creates stress about your financial position and non-monetary attributions like the value of work. Brunner & Marmot (1999), researched the influence of psychological factors on physical health. They firstly state that stress brings short-term effects on the physical as well as the mental health. They secondly started recognizing that people’s social and psychological circumstances contribute to their health in the long term. The psychological circumstances they refer to are: “chronic anxiety, insecurity, low self-esteem, social isolation, and lack of control over work”. These psychological circumstances both affect your mental as well as your physical health. They also state that this makes biological sense. Because our bodies have developed in such a way that they automatically respond to emergencies. Stress brings stress hormones, and these affect the cardiovascular and immune systems. Brunner & Marmot (1999) state that hormones in the short

(17)

16 term may save lives. The heart rate rises, more blood flows through your muscles and anxiety and alertness both increase. This makes sure we can deal better with a physical threat for a brief moment. But this stress level should decrease, otherwise this response of our bodies is activated for too long and too often. This can increase all sort of health issues, like “depression, increased susceptibility to infection, diabetes, high blood pressure and accumulation of cholesterol in blood vessels walls” (Brunner & Marmot, 1999).

Björkland et al. (2015) is, again a good starting point. The young unemployed men interviewed, started to drink considerably more and some also started to smoke. The pattern that the authors found was that being drunk sometimes eased their concerns. This shows that the psychological factor of not working heavily influenced their decisions about their physical health. The interviewees reported that they do not much more than “sleeping, watching TV and partying with friends, while clearly expressing that they would like to manage their physical health better, for example by training and exercising”.

This qualitative research shows that it is very plausible that certain mechanisms are at place when looking at the physical and mental health of the unemployed. They use their time differently in comparison to the employed, by watching more TV and ‘hangout’ with friends, but not necessarily do more ‘healthy’ things with their time like exercising or voluntary work. More on this subject is discussed in paragraph 2.6. For now it is good to look at quantitative research that took both the physical as well as the mental health of unemployed people into account. Romeu Gordo (2006) has done a quantitative research into this subject in Germany. For many individuals unemployment is something that is very hard to emerge from, especially for the older population. The chances of re-employment also decrease the longer you stay unemployed. And this situation can cause health problems to people who cannot accommodate to the situation. Romeu Gordo (2006) uses a longitudinal study (called German Socio-Economic Panel) to analyze the effects on health satisfaction. She uses short- and long-term effects as well as the effect of re-employment on health satisfaction. This health satisfaction is a combination of both the physical and mental health, but of course is measured subjectively. Her main conclusions are differentiated between men and women. And rightly so, because for men unemployment has a significant negative effect on their health satisfaction in the short term while for women Romeu

(18)

17 Gordo (2006) finds no effect. An unemployment spell of two years and longer has a significant negative effect for both men and women. And re-employment has a significant positive effect for also men and women, this effect does not differ if you control for the time of the unemployment spell.

Another research done with the same German longitudinal study is Schmitz (2011). The results Schmitz (2011) states are, interestingly, not in line with the study by Romeu Gordo (2006). Schmitz used exogenous entries into unemployment and found no effect of unemployment on health. Other reasons for unemployment that are not exogenous show a negative effect of unemployment on health. Schmitz, however, nuances this finding by saying that reversed causality might contribute to this result. The exogenous distinction Schmitz (2011) makes, is different than the approach of Romeu Gordo (2006). This approach of Schmitz takes the possible endogeneity of unemployment into account, to take the reversed causality out of the equation. He also states that “it can be assumed that unemployment first reduces mental health before it deteriorates the overall health status” (Schmitz, 2011). But he finds no evidence of a negative effect of being unemployed on mental health. Why it is hard to find an effect of unemployment on health, is the fact that Schmitz (2011) uses closure of a plant. This reason is exogenous, but the only thing this can establish is a non-negative effect. This is because you cannot rule out that the health of someone declined because they got fired or due to other endogenous reasons. Salm (2009) and Böckerman & Ilmakunnas (2009) also do not find evidence that unemployment has a significant effect on health. Salm (2009) also uses the same exogenous strategy as Schmitz (2011), namely a business closure, and also looked into individuals characteristics to separate voluntary and involuntary job loss. He also finds no effect even when controlling for specific groups by gender, marital status, race, education, and previous conditions. This therefore raises the question if the health of the unemployed is inferior to the employed because of reverse causality. The conclusion of Salm (2009) is what he calls a “cautious” one. The absence of a significant effect of unemployment on health might be because losing your job does not lead to a decrease in your health.

Böckerman & Ilmakunnas (2009) also use longitudinal panel data, but then from Finland instead of Germany. They find just as Salm (2009) that there is no significant effect of unemployment on

(19)

18 the health of individuals. The health status of the unemployed is already worse than those of the employed and when using difference-in-difference models, the change in health is not significantly dissimilar. Therefore they conclude that unemployment can be a veil for health issues. But, when the researchers looked at long-term unemployment, they do found negative effects on a person’s self-assessed health. The results for Finland, can have the same explanations as for Germany.

Empirical research on this topic has been done in many ways, and tell a different story. Schiele & Schmitz (2016) again use the German longitudinal panel data, and found something else than Schmitz (2011). The method Schiele & Schmitz (2016) used was a quantile regression method, this method makes it possible to look with one explanatory variable at different outcomes and vary in the explanatory power of this variable. Thus, the authors were able to look at physical and mental health and tell them apart. This is more sophisticated then the self-assessed health indicators and the health satisfaction. To estimate causal effects of unemployment, Schiele & Schmitz (2016) focused on job loss due to plant closure. They found that the adverse effects of unemployment on physical health are significantly negative and are of a mentionable size, but only for people who had poor physical health right before they became unemployed. The effects on mental health are along the same line, although not significant. An example of an indicator for physical health that only had a very small medium term effect, is BMI. This again shows that perspective matters, the effects of unemployment on health are very hard to grasp, differ among studies and are heavily influenced by which health indicators you are measuring with. And Schiele & Schmitz (2016) also point out that because they looked at a plant closure, the health effects of people that voluntary quit their job and become unemployed are absent. Because it was their own choice, but if you get fired the health effects might even be worse because it has to do everything with you instead of a whole company that gets shut down.

To look at involuntary unemployment, Schaller & Stevens (2015) have done research on this subject. They found that unemployment that is involuntary is associated with significant decreases in self assessed physical and mental health and reports of anxiety or depression. For chronic diseases (heart diseases, diabetes, high cholesterol, etc.), they found no significant evidence that the chances of those increase if you become involuntarily unemployed.

(20)

19

2.5. Effects of job insecurity and inco me volatility on health

Because this thesis takes the expectations of losing one’s job into account, it is important to also touch upon the subjects of job insecurity and income volatility and their implication on one’s health. Job insecurity and income volatility are in essence what this expectations asks: do you feel your job is secure and with it your income? There has been done some research into both of these phenomena and their effects on health. To start off, job insecurity will be looked upon first.

In the psychological literature job insecurity is a known source for poorer health and well-being (Burchell, 1994, Nolan et al., 2000, Wichert, 2002, Cheng & Chan, 2008). Caroli & Godard find that this is true on almost all health categories when you do not account for the endogeneity of job insecurity (2016). But when they do, they find effects on fewer health categories. So, the effects of job insecurity on physical health are maybe not as robust as initially thought. For mental health Green (2011) finds that the initial effects are robust. He even compares 100% job insecurity to unemployment, and finds that the effects are the same for someone with average employability. The last finding of Green (2011), is :when one has no job insecurity there is more life satisfaction. So, when a person does not expect to lose their job, this person already could have better mental health than the person who does expect to lose their job. As expressed above, to look at income volatility is an extended form of looking at job security. Because it all comes down to consumption, is this level of consumption sustainable for you in the future? And when it is not, what are the impacts on health? A recent study by Adeline, Crèvecoeur, Fonseca & Michaud (2019) shows that long- and short-term income volatility are associated with decreasing health and well-being. Their findings for the long-term volatility are of bigger magnitude for a person’s health and well-being than the short-term volatility. This has all to do with the life-cycle model, a short-term income volatility can set you off your life-path for a short amount of time with the opportunity to return to it. While a long-term income volatility has a permanent effect on your life-cycle and means that you consume less now and also will in the future. This idea is probably more offsetting than a short-term income volatility that gives hope of improvement to return to a higher consumption in the future. Adeline et al. (2019), however, do warn for the existence of reverse causality in their results even though they tried to tackle this problem by controlling for life-time factors such as disability. The assumption that the life-cycle model plays a part in the effects is strengthened by the study of Halliday

(21)

20 (2017). He finds that earnings growth has a positive relationship with health for married persons. And he also finds that the income of their spouse protected married women against deteriorating health. This shows that married women feel their life-cycle is secured by their spouse, and that their own income volatility will not harm their consumption level. Halliday (2017) did, however, find no evidence between income volatility and health for single persons.

2.6. Time Use

To look further into the mechanisms of being employed versus being unemployed, it is important to know how these groups spend their time. Being employed gives scarcity to your leisure time, while the unemployed have more leisure time. The question is then, does this extra leisure time makes you happier?

Aguiar, Hurst & Karabarbounis (2012) report that the literature up until 2012, indeed show leisure time is increasing for the unemployed. Another important notion that the authors give, is the intratemporal substitution between time and goods. Once you retire, you spend more time on producing final goods and therefore consumption drops in comparison to the same non-retired people. The same holds up for unemployed people. They allocate 30 to 40 percent of the time they used to work into home production, while leisure time absorbs about 50 percent of the time the unemployed used to work with sleeping and watching television being the largest categories this leisure time is being spent on (Aguiar, Hurst & Karabarbounis, 2013). And when the unemployed are being reemployed, they report decreases of 35 percent in time spent on cooking, housework, child care, and shopping (Krueger & Mueller, 2012). Another activity that is important when you are unemployed, is the time you spend on finding a new job. In some countries it is even required to have access to unemployment insurance. Krueger & Mueller (2010) show that the average unemployed worker in the U.S. spends 41 minutes per week on finding a job, this is more than the average unemployed worker in Europe. Aguiar, Hurst & Karabarbounis (2013) also show that the unemployed devote time to job searching, between two and six percent of the hours they used to work.

How does your time-use then relate to your health status? Krueger & Mueller (2012) report that the unemployed are sadder then the employed, they suggest that during activities in leisure time the unemployment plays a negative role in how these activities make you feel. This suggestion is

(22)

21 in line with other research, like Knabe, Rätzel, Schöb & Weimann (2010). Knabe et al. (2010) find that “Employed persons are more satisﬁed with their life than the unemployed and report more positive feelings when engaged in similar activities”. But Hoang & Knabe (2019) also show a negative relation between working and life satisfaction. While working is reported as “one of the least enjoyable experience of the day” by the employed, the unemployed allocate this time to activities that they enjoy. This phenomenon is called the time-composition effect (Hoang & Knabe, 2019).

There are also researches that look at the relation between time-use and health from a different perspective. Podor & Halliday (2012) find that better health is associated with more time spend on market- and home-production, which means less leisure time. They put their findings in the context of the model of time-allocation of Gronau (1980) and conclude that better health means higher efficiency in market-production and even a higher efficiency in home-production. So, healthier people work more, whether it is at home or at a job. That healthier people get more done at home, is also in line with Hamermesh & Lee (2007). They say, if you keep all other thing equal, that anything that raises efficiency in commodities at home makes people less stressed about time. This stressing about time is also present in the research of Giménez-Nadal & Molina (2015). They find that people that report less time spend on sleep, personal care, non-market work and leisure time for men, perceive their own health better. They also find that market-work is a substitution for the above time categories.

The last perspective of time-use on health is an important one for the theoretical framework used in this thesis. Namely that time is a resource for good health, because “resting, healthy behaviour, accessing health services, working, resting and caring all require time” (Strazdins, Welsh, Korda, Broom & Paolucci, 2016). Strazdins et al. find that people that are “time-poor” (with reports of more than 80 hours per week on care and work), or people that say they are often or always rushing spend less time on physical activity. Feeling rushed is also associated with lower self-rated and mental health. Du & Yagihashi (2017) show that when wage increases with 10%, people increase their time-use on healthy things like exercising and medical and personal care by 18,5 minutes per week. While time spend non-healthy things like watching television decrease with 46,3 minutes per week when wage increases with 10%.

(23)

22

2.7. Theoretical fra mework

To combine the literature, we need a theoretical framework to help understanding more of the mechanisms at place. To do this, a rather classic model will be used. Namely, the Grossman model (Grossman, 1972). Grossman states that an agent maximizes an arbitrary utility u with respect to consumption c and health g.

𝑢(𝑐, 𝑔)

In the equation u has the usual properties and u’> 0 and u’’< 0. The u is maximized under the following budget and time constraints:

𝑐 = 𝑤 ℎ + 𝑏

Where w is wage, h working hours, and b benefits if h = 0. But, since time is not free of constraints, we look at the total time available (24 hours per day) with the following equation:

𝑇 = ℎ + 𝑙 + 𝑠(𝑔_𝑡)

With T total time available, l leisure time, and s sickness time. We assume that health follows this equation:

𝑔_𝑡+1− 𝑔_𝑡 = 𝑖_𝑡− 𝛿𝑔_𝑡 Δ𝑠

Δ𝑔_𝑡< 0

And health investments i follows:

𝑖_𝑡= 𝑓(𝑐, 𝑙)

With f being an arbitrary function mapping consumption and leisure time into health investments.

(24)

23 1. h = 0 which decreases c assuming w+h > b. Hence, the agent is likely to decrease i after

unemployment ceteris paribus.

2. Drop in h, gives an increase in l. Hence, the agent is likely to increase i after unemployment ceteris paribus.

In conclusion, theoretically, the effect of unemployment on health is ambiguous and depends on the changes in consumption and leisure and how they produce health. This thesis empirically estimates the effect of unemployment on health and look upon the mechanism to which health changes, by analyzing the effect of unemployment on leisure time.

2.8. Effects of institutions and ins titutional f ra mework in The

Netherlands

In the theoretical framework, b stands for the benefits one receives. This requires to look a little closer at the effects of institutions and the institutions that are in place in the Netherlands that affect the results. The institutional framework consists of the unemployment insurance (UI) and health care insurance in The Netherlands.

2.8.1. Effect s of insti tuti ons

To come back to Schmitz (2011), he finds no causal effect of unemployment on health. Schmitz (2011) also provides two explanations why this is the case. First, it is the generous insurance benefits Germany has for the unemployed, especially when you look at the entitlement duration. This is very long in comparison to other countries, and therefore the financial impact of unemployment is smaller which only causes a little effect to the health status. The notion that health insurances mitigates the effects on both mental and physical health if someone gets unemployed is supported in the literature (Tefft, 2011; Ferrarini, Nelson & Sjöberg, 2014; O’Campo et al., 2015; Kuka, 2018).

Secondly, Schmitz (2011) notes that when you compare Germany to the United States for example, people do not lose health insurance in Germany when they lose their job. Healthcare is therefore still available in a large amount to the German people. Hence, the financial constraint in general and especially on the subject of health investments is low in Germany compared to the United States (see Sullivan & Von Wachter, 2009).

(25)

24

2.8.2. Une mpl oy me nt ins uranc e (U I)

Normally, when people become unemployed in The Netherlands they have the right to claim UI benefits. But, you will only receive this right if you have worked at least 26 of the last 36 weeks and if the job loss is not due to the employee. The reasons why job loss are due to the employee are voluntary quits and instant dismissals carried out by the employer. The duration of the UI benefits depends on the duration of previous work spells. The minimum duration is three months, this will get extended by one month for every working year. The maximum duration was 38 months for the people who worked 4 out of the last 5 years, but since 2016 this maximum has been reduced to 24 months. Additionally the extension of months is also reduced, you will receive a month for every year worked until 10 years, after this pint you will get half a month extension for every year worked. The replacement rate is 75% for the first two months of the last earnings, the absolute maximum has been set on € 3.100 per month. The replacement rate is reduced to 70% from the third month on with an absolute maximum of € 2.900 per month. Until 2016 the replacement rates were 70% for the total duration of the unemployment spell. But in the agricultural, industrial and constructional sectors, collective agreements make sure that employers complement the UI benefits to let the employees receive a 100% replacement rate. The duration of this replacement rates varies between the sectors. When one loses his or her job, their contribution to occupational pensions are stopped automatically or are reduced, this differs per sector in The Netherlands.

The UI benefits in The Netherlands can be called generous in comparison to other countries. Especially the drop between two months and one year is relatively small. After five years we see a replacement rate that is also one of the highest, only Austria, Belgium, Denmark and Luxembourg have higher replacement rates after five years. However, The Netherlands do have a high duration of unemployment spells. OECD (2019a) shows that only 6% of the unemployment spells in The Netherlands are shorter than a month, in the U.S. this is about 30%. And the unemployment spells that are longer than one year in the Netherlands are more than 40% of all unemployment spells, only in Belgium, Greece, Ireland and Portugal this number is higher.

(26)

25

2.8.3. Health i nsura nce

All Dutch citizens are required to purchase statutory health insurance from private insurers. The insurers are required to accept all applicants. The financing of the health care is primarily public, through premiums, tax revenues, and government grants. The health insurance standardly include physician, home nursing, hospital and mental health care, as well as prescription drugs. The insured pay premiums, annual deductibles, and coinsurance or copayments on selected services and drugs. The government finances the coverage for children up to age of eighteen.

In Europe, only Norway, Germany, Austria, and Sweden spend more on health per capita than The Netherlands, according to the OECD (2019b). In 2017, 10.1 % of the GDP was spend on health, which is slightly above the EU average of 9.8 %. The absolute spending is € 3,791 per person, which is, again, above the EU average of €2,884.

(27)

26

3. Methodology

This chapter discusses the methodological choices that are made in this thesis. First the data collection and operationalization of the variables are discussed. Then the method of analysis and lastly a reflection of the outcome of these choices on validity, reliability and generalizability.

3.1. Data collection & operationalization

As been said in the introduction, this thesis derives its data from the LISS panel in the Netherlands. The panel contains longitudinal data from 2007 until 2019 on many different subjects. The panel is based on a true probability sample of around 5,000 households drawn from the population register by Statistics Netherlands. The same households are questioned in every wave, but it is possible to drop out of the panel. This thesis uses the core studies about health, income, personality and social inclusion & leisure from the LISS panel to conduct quantitative research on. Important for this research design is to assess whether the transition into unemployment or staying employed is unexpected. The LISS questionnaire has the following question which this research exploits: Do you think that there is any chance that you might lose

your job in the coming 12 months? You can indicate this in terms of a percentage.

As discussed in the literature review, there are many ways to measure one’s physical and mental health. But since we are using a longitudinal panel survey, we are constraint by the questions in the survey. Here the scope of the survey is important, and the scope of the LISS panel is very broad. Some of the most important variables used in the LISS survey for this research are very much the same as the above described literature used. There is also a diversity between subjective health measures and objective health measures made by a doctor. For the physical subjective health measures, thedr are variables in the survey that feature in this research: subjective health (measured from 1 very bad to 5 very good), life satisfaction (measured from 0 very bad to 10 very good), body mass index, smoking and alcohol. Smoking and alcohol both consist of two variables. For smoking the questions are: have you ever smoked (dummy variable) and do you smoke now (dummy variable). For alcohol the question are: how did you have a drink containing alcohol in the last 12 months (1 for every day to 8 not at all) and on how many of the last week did you drink alcohol? The physical objective health measures that feature in this research are: high blood pressure or cholesterol measured by a doctor (dummy variables),

(28)

27 medication for sleeping problems or anxiety or depression prescribed by a doctor (dummy variables). To further look upon health behaviour the times per year the respondents visited a family physician or medical specialist in a hospital are also be included. To look at the mental health, the only variable that is possible to use from the LISS panel is the ‘satisfied with life’ question. This question asks on an eleven-point scale (zero to ten) how satisfied one is with life at the moment. This variable comes with a notion: many things affect this variable, all factors of life are included like housing and social contacts. The last inclusions are control variables where we do not expect to see a relation between them and unemployment: ability to walk 100 meters and to pick a coin from a table (both ranging from 1 without any trouble to 5 .

For the operationalization of the time use, the LISS panel survey also measures how people allocate their time. From the literature review it is clear to look at the time people watch television, sport, cook, clean (house commodities), repair, perform informal care and voluntary work. These categories all feature in the survey and therefore are used. To also further look at the time use, the variables of time spend listening to the radio, reading, following a course, gardening, shopping (online), going out, listening to music and travelling are present. The time use variables are measured in minutes spend on each category per week.

To control for demographic characteristics, some control variables are also be added to the regressions. As shown from the non-monetary (health) literature, it is very plausible that we see different results for men and women and older persons. When looking at the monetary side of the effect, having a partner may lead to more consumption smoothing, higher educated may lead to more income and less consumption constraint when being jobless, and being the main income earner when losing your job may lead to more consumption constraint when losing your job. That is why there are control variable like being female, which is a dummy variable with 0 for being male and 1 for being female. The same system with dummy variables is used for having a partner (0 for not having a partner, 1 for having a partner), having had high education (0 for not having had high education and 1 for having had high education), being fifty or older (0 for not being fifty or older, 1 for being fifty or older) and being the main income source for the household (0 for not being the main income source, 1 for being the main income source). High education in The Netherlands has been operationalized by distinguishing between having a degree from an applied science university (HBO in Dutch) or a university and not having a

(29)

28 degree from these institutions. For these control variables there are also separate regressions, these are discussed in the next paragraph. Furthermore, the regression uses control variables of age, change in the number of children and change in partner. The variable of age is split up in four categories: (1) between 25 and 34, (2) between 35 and 44, (3) between 45 and 54, and (4) between 55 and 64. When looking at the change in partner, the 0 stands for "no change", 1 stands for "single to married/relationship", 2 stands for "married/relationship to single", and 3 stands for "don't know".

Then all these variables from the different databases are put together in one dataset, and since every respondent has a unique respondent’s number it is possible to look at the same person through different waves over time. Every regression has a different number of observations, since for some variables more respondents have filled in an answer than for other variables. To further specify the dataset, only people being aged between 25 and 64 are represented in the regressions. This distinction has been made because it is believed that between these ages you lead what is known as your “working life”. To further look at the effects, people that are not looking for a job or are in an (early) retirement are excluded, so the only people that are working or want to work are included in the regressions. This makes sure that the results are not biased with observations from people that don’t want to work due to for example being a stay-at-home parent or are retired. This gives truer effects of job loss on health and time use.

The following tables gives some summary statistics for every variable that is used in the regression:

(30)

29 Table 1. Summary statistics.

Variable Observations Mean Std. Dev. Min Max

Independent variables

Job loss 18,266 .021 .145 0 1

Unexpected job loss 15,215 .007 .073 0 1

Unexpected job keep 15,215 -.162 .239 -1 0

Health variables

Subjective health 31,423 3.116 .758 1 5

BMI-score 31,035 25.660 4.086 15.321 40.999

High blood pressure 30,014 .127 .333 0 1

High cholesterol level 30,014 .076 .265 0 1

Ever smoked 31,361 .575 .494 0 1

Smoking now 18,023 .356 .479 0 1

Alcohol use 31,357 4.491 2.172 1 8

Days of consuming alcohol last week 19,343 3.397 2.097 1 7

Using drugs 31,425 .061 .234 0 1

Medication for sleeping problems 31,286 .043 .203 0 1

Medication for anxiety or depression 31,286 .052 .223 0 1

Family physician visits 31,287 1.952 3.186 0 100

Medical specialist visits 31,289 1.129 3.060 0 100

Satisfied with life 35,619 7.388 1.411 0 10

Walking 100 meters 31,409 1.168 .552 1 5

Picking a coin from a table 31,403 1.085 .388 1 5

Time use variables

Informal care 33,326 2 8 0 168 Volunteer 33,324 1 5 0 168 Sport 33,125 112 146 0 840 Watching tv 33,006 1039 646 0 4200 Listening radio 32,995 1100 1197 0 5040 Reading 21,911 320 320 0 1890 Following a course 4,303 251 272 0 2100 House chords 24,790 350 366 0 2100 Gardening 19,763 248 326 0 2310 Repairing 1,458 204 317 0 2400 Cooking 24,856 312 232 0 1920 Shopping 17,593 176 147 0 960 Shopping online 23,250 51 113 0 6000 Going out 11,834 207 165 0 1208 Listening music 24,222 905 1016 0 5040 Travelling 1,737 147 167 0 1500 Control variables Age 41,893 46.599 11.248 25 64 Being female 33,420 .547 .498 0 1 Having a partner 33,438 .825 .380 0 1

Being high educated 36,225 .629 .483 0 1

Being fifty plus 41,893 .443 .497 0 1

(31)

30

3.2. Method of analys is

To analyze the data, this thesis uses a regression analysis in STATA on the LISS panel data to estimate causal effects. To turn the theoretical framework explained in paragraph 2.6. into an empirical specification we need the following regression equation:

∆𝑔𝑖𝑡 = 𝛾0+ 𝛾1[𝑗𝑜𝑏𝑙𝑜𝑠𝑠𝑖𝑡− 𝐸𝑖𝑡−1𝑗𝑜𝑏𝑙𝑜𝑠𝑠𝑖𝑡] + 𝑎𝑔𝑒′𝑖𝑡𝛾2 + ∆𝑋′𝑖𝑡𝛾3+ 𝑡′𝑡𝛾4+ 𝜀𝑖𝑡

Where 𝑔_𝑖𝑡 refers to health of individual 𝑖 at time 𝑡; 𝑗𝑜𝑏𝑙𝑜𝑠𝑠_𝑖𝑡 is a dummy variable that takes value one if an individual suffers an involuntary job loss between periods 𝑡 − 1 and 𝑡; 𝐸_𝑡 is an operator denoting the expectations an individual forms conditional on the information available at 𝑡; [𝑗𝑜𝑏𝑙𝑜𝑠𝑠_𝑖𝑡− 𝐸_𝑖𝑡−1𝑗𝑜𝑏𝑙𝑜𝑠𝑠_𝑖𝑡] is the unemployment shock at time 𝑡, which takes the values in the interval [0,1] if 𝑗𝑜𝑏𝑙𝑜𝑠𝑠_𝑖𝑡 = 1, and takes values in the interval [-1,0] if 𝑗𝑜𝑏𝑙𝑜𝑠𝑠_𝑖𝑡 = 0; 𝑎𝑔𝑒_𝑖𝑡 is a vector of age dummies; 𝑋_𝑖𝑡 is a vector of control variables including marital status and number of children in the household ; 𝑡_𝑡 is a vector of year dummies; and 𝜀_𝑖𝑡 is the error term. The same estimation is also used to look at the allocation of time of an individual.

This estimation is based on the previous work by Stephens (2004). He stated that subjective job loss expectations are highly significant predictors of subsequent job losses. Stephens only looked at working men, with information on their industry, occupation, tenure and job loss probability. He got his data from the Health and Retirement Study (HRS), conducted in the U.S. between 1992 and 1996. For the demographic control variables Stephens only found an effect of tenure on job loss, the longer your tenure is, the lower the chance is that you will be displaced in the following year. All the other demographic variables don’t show any significant effect. Then Stephens (2004) finds that if you have put in 100% of probability of losing your job, the chance you are displaced the following grows with 14,6% compared to one who thought he had a 0% probability of losing his job. This means that if you go from 0% to 10% probability of losing your job, the chances increase with roughly 1,5%. This estimation holds mostly if Stephens (2004) controls for demographic variables, the estimation only drops slightly to 12.2%.

So, Stephens (2004) shows that individual job loss expectations contain much information that goes beyond information found in demographic variables known to be related to the occurrence of job loss. It therefore is highly interesting to differentiate between expected and unexpected job

Unexpected employment shocks and their influence on health and time use