Wounds that time can’t heal: Life satisfaction and exposure to traumatic events

Journal Pre-proofs Wounds that Time Can’t Heal: Life Satisfaction and Exposure to Traumatic Events Alessandro Bucciol, Luca Zarri PII: DOI: Reference:

S0167-4870(19)30208-9 https://doi.org/10.1016/j.joep.2019.102241 JOEP 102241

To appear in:

Journal of Economic Psychology

Received Date: Revised Date: Accepted Date:

15 April 2019 11 November 2019 15 November 2019

Please cite this article as: Bucciol, A., Zarri, L., Wounds that Time Can’t Heal: Life Satisfaction and Exposure to Traumatic Events, Journal of Economic Psychology (2019), doi: https://doi.org/10.1016/j.joep.2019.102241

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

© 2019 Published by Elsevier B.V.

WOUNDS THAT TIME CAN’T HEAL: LIFE SATISFACTION AND EXPOSURE TO TRAUMATIC EVENTS* Alessandro Bucciol

Luca Zarri†

University of Verona, Dept. of Economics Via Cantarane 24, 37129, Verona, Italy

*

The authors declare no conflict of interest, and this research received no specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

†

Corresponding author: Luca Zarri. Economics Department, Via Cantarane 24, 37129 Verona, Italy; Phone: +39 045 802 8101. Email: [email protected]

WOUNDS THAT TIME CAN’T HEAL: LIFE SATISFACTION AND EXPOSURE TO TRAUMATIC EVENTS Abstract In this study, by employing large-scale survey data from four waves of the US Health and Retirement Study (HRS), we explore the (potentially long-lasting) effects of individuals’ exposure to psychologically traumatic life experiences on their subjective well-being. To this aim, we exploit the richness of our dataset, that contains information about occurrence and timing of a set of extreme events out of individuals’ control that may leave a “scar” extending to their current levels of life satisfaction in general as well as with regard to specific life domains. Our findings indicate that having a close relative hit by a life-threatening illness or accident and, especially, having been victim of a serious physical attack or assault are negatively related to both general and domain-specific life satisfaction, even after controlling for personality traits. Next, life satisfaction is significantly lowered by being physically abused by a parent. Overall, we provide evidence that the effects of some traumatic events are persistent over time and mostly related to women. Surprisingly, the effects of child death are negligible also in the short term. Keywords: Adverse Life Events; Subjective Well-Being; Life Domains; Gender. JEL Classification: D91; I30.

2

“It has been said, ‘Time heals all wounds’. I do not agree. The wounds remain. In time, the mind, protecting its sanity, covers them with scar tissue and the pain lessens. But it is never gone” Rose Fitzgerald Kennedy

1. Introduction In the last decades, significant progress has been made within the extensive body of research that deems self-reported measures of subjective well-being or life satisfaction1 as valid and reliable indicators of well-being (Van Praag et al., 2003; Di Tella and MacCulloch, 2006; Metcalfe et al., 2011). Life satisfaction scores obtained from survey-based measures asking individuals how happy or satisfied they are with their life as a whole these days provide information on how respondents globally evaluate their present life. As noted by Blanchflower and Oswald (2004), self-assessed well-being is related to distinct factors such as circumstances, aspirations, comparisons with others, and a person’s baseline happiness or dispositional outlook. Last years’ empirical research has shown that subjective well-being has genetic roots (Bartels, 2015) and is associated with sociodemographic variables (income, gender, age, education, employment status, ethnicity, marital status; see, e.g., Clark and Oswald, 1994; Blanchflower and Oswald, 2004; and Graham and Chattopadhyay, 2013), personality traits (Hayes and Joseph, 2003; Weiss et al., 2008), relative earnings (Luttmer, 2005), parenthood (Stanca, 2012), natural disasters (Luechinger and Raschky, 2009) as well as a series of non-pecuniary factors including relational goods (Bruni and Stanca, 2008), freedom of choice (Verme, 2009), tax morale (Lubian and Zarri, 2011) and religiosity (Van Praag et al., 2010). Prior psychology and economics work in this area documented that a series of major negative life events – including unemployment, involuntary retirement, divorce, widowhood, disability and natural disasters – have strong effects on one’s current level of well-being 1

In this work, we use interchangeably the expressions “subjective well-being”, “happiness” and “life satisfaction”. As pointed out by Luechinger and Raschky (2009), even though “life satisfaction” and “happiness” may be viewed as constructs capturing slightly different aspects of well-being, there is high correlation between them and the two measures provide similar results (see on this also Di Tella and MacCulloch, 2005, and Van Praag et al., 2010).

3

(Clark and Oswald, 1994; Lucas, 2007a; 2007b; Luechinger and Raschky, 2009; Yap et al., 2012; Anusic et al., 2014), with the strength of these effects depending on the life events considered in the analysis (Luhmann et al., 2012). Drawing on this rapidly expanding line of inquiry as well as on evidence from medicine and psychology studies indicating that past exposure to traumatic events is linked to one’s dispositions and well-being in later life (see Section 2 on this), our central conjecture in this study is that individuals’ current levels of general and domain-specific life satisfaction are negatively related to extremely adverse, psychologically traumatic events out of their control they passed through in their life. To this aim, we contribute to the existing literature by addressing the following questions: do these traumatic events resemble more a temporary “wound” that time can heal, as reminded to us by an old proverb, or do they leave instead a long-lasting “scar”, as suggested by Rose Kennedy’s opening quote? Does the association between traumatic events and life satisfaction depend on the domain of life satisfaction we refer to? Are there significant gender differences, with regard to this relationship? The main contributions of this study can be summarized as follows. First, we consider a set of extremely adverse life events out of individuals’ control (i.e., child death, physical attack/assault, physical abuse by a parent, life-threatening illness or accident that hit the respondent or a close relative)2 that arguably are associated with life satisfaction mainly via psychological channels. In contrast, most of prior research on the effects of major life events on subjective well-being examines life events – such as job loss, disability, divorce, widowhood, childbirth, retirement and natural disasters – that are often intertwined with significant worsenings in individuals’ socio-economic and/or physical health conditions (see Section 2 on this).3 Second, unlike many studies we refer to in Section 2, our dataset also

2

As pointed out by Lucas (2007a), “Certain people may be predisposed to experience life events, and these predisposing factors may be responsible for their happiness levels being lower than average” (p. 76). Therefore, in order to further mitigate selection concerns, in our analysis we also control for personality traits, by considering the “Big Five” taxonomy as well as further relevant traits (see Section 3 on this).

3

To rule out alternative interpretations, our empirical strategy also includes individuals’ socio-economic and health characteristics among the controls. As we make clear in Sections 3 and 4, we control for individuals’ socio-economic conditions by considering their education, occupation, income, financial and real wealth as

4

allows us to shed light on the relationships between timing of events and life satisfaction. Our intuition is that the aforementioned events may be serious enough to leave a scar on respondents’ life satisfaction even when they occurred during their formative years, i.e., some decades before, so that current self-evaluations of well-being, more or less subconsciously, incorporate the psychological burden of traumatic events. Third, in order to provide a richer and more comprehensive picture of the relationships between life events and life satisfaction, we investigate the links between traumatic events and life satisfaction also in relation to distinct areas of life (i.e., domains). Compared to prior work, this allows us to provide a more fine-grained analysis of the effects of different life events on general life satisfaction as well as on different and relevant life domains. Even though so far domain satisfaction can be viewed as a relatively under-investigated topic within the subjective well-being literature (Rojas, 2006; Bardo and Yamashita, 2014), in many parts of the world – as noted by Diener (2006) – there is growing consensus over the use of indicators of subjective well-being to inform policy debates and assessments of specific aspects of well-being and ill-being should prove helpful to policymakers beyond global measures. Finally, we shed light on the gender dimension, by exploring differences between men and women in their relationship between exposure to extremely adverse events and general and domain-specific life satisfaction.4 For our analysis, we use large-scale longitudinal data from the four 2006-2012 waves of the US Health and Retirement Study (henceforth HRS).5 This survey, collecting information

well as their macro-region of residence (by including region fixed effects in all our models) and we control for health characteristics by considering self-reported health status plus three functional limitations scores. 4

Easterlin and Sawangfa (2009) point out that an advantage of a domain-based approach to life satisfaction is that “judgments on domain satisfaction reflect both subjective factors of the type emphasized in psychology and objective circumstances stressed by economics” (p. 71). As they observe, while there is no complete agreement over which domains of life are conceptually preferable, most scholars agree that some domains are of major importance. Their analysis indicates that the overall patterns of happiness can be explained quite well by using four domain-specific measures of satisfaction (that is, finances, family life, work and health).

5

As noted by Lucas (2007a), two important methodological advantages of large and representative samples, over alternative designs, are that even rare events are sampled and that, since the questionnaires often

5

on a large sample representative of the US population aged 50 or more, is an important tool for addressing our research questions as it conveys information on current levels of subjective well-being (including both general and domain-specific life satisfaction measures) and extremely adverse events occurred in the past, including information on timing of the events as well as on specific events occurred more than three decades before during childhood or adolescence, i.e., when the respondent was less than 18 years old.6 Our major results can be summarized as follows. We detect a reduction in life satisfaction after the occurrence of a life-threatening illness or accident that hit a close relative (partner or child) and, in particular, after a serious physical attack or assault that hit the respondent. Surprisingly, passing through the death of a child is not significantly associated with life satisfaction, not even in the short term. As to extremely adverse events experienced during childhood or adolescence, life satisfaction is significantly lower when the respondent was physically abused by either parent. Next, we show that the effects of two relevant traumatic events (being victim of physical attack/assault and physical abuse) are also persistent over time. Further, our assessment of life satisfaction through domain-specific measures (considering satisfaction with home, leisure, family life, city, finances and health) indicates that the traumatic events that turn out to be more significantly associated with general life satisfaction also matter for several, but not all, domains. In general the effects turn out to be gender-specific, as they hit women much more than men, especially on those events that involve a close relative (child or partner) rather than oneself. The remainder of the paper proceeds as follows. Section 2 provides a selective review of the relevant strands of literature, with a special focus on studies exploring the relationship between current subjective well-being and prior adverse life experiences. Section 3 describes our data and methodology. Section 4 contains our key findings and Section 5 concludes.

include a variety of issues, demand characteristics are unlikely to have much of an effect (see, on this, also Anusic et al., 2014). 6

Notice that age is limited in the 50-80 range, since, as a further restriction, we decided to focus on individuals not older than 80, as the elderly are over-sampled in HRS. In addition, as Bucciol and Zarri (2015) noted, such individuals may find it difficult to recall past personal events, especially when such events occurred in their early life.

6

Finally, supplemental information lists the raw variables used to identify life satisfaction and traumatic events.

2. Related Literature This paper lies at the crossroads of two areas of research. One is the fast-growing “economics of happiness” literature and, in particular, the line of inquiry that examines the links between happiness and negative life events. The other is the huge cross-disciplinary research area – that has been developing mainly in psychology, psychiatry and medicine, but more recently also in economics – dealing with the effects of past exposure to trauma and other major life events on current decision-making and well-being. Our study seeks to contribute to these streams of literature by providing empirical evidence on the relationship between adults’ life satisfaction and a series of extremely adverse life events they experienced over time throughout their personal history. Over the last decades, studies on the determinants of subjective well-being have become a growingly important and active research area that crosses several scientific disciplines, including economics, psychology and medicine (Bartels, 2015). Using UK data on mental distress, Clark and Oswald (1994) find that the effect of being jobless is significantly and negatively related to well-being (see on this also Blanchflower and Oswald, 2004). The detected effect is quantitatively large and robust across varied specifications. They also offer evidence that individuals who have been unemployed for a long time have lower levels of distress than those who have recently become jobless. By means of comprehensive panel data on Germany, Bonsang and Klein (2012) show that, in line with Clark and Oswald’s (1994) established result, involuntary retirement is negatively associated with life satisfaction, whereas voluntary retirement has a small positive effect on life satisfaction. Based on US data, Blanchflower and Oswald (2004) document that marital status is very important for happiness, as both being separated and being widowed are significantly and negatively associated with reported happiness (see on this also Lucas et al.’s, 2003, longitudinal study on the effects of marital transitions on life satisfaction). Oswald and Powdthavee (2008) provide evidence based on British and German micro data that people exhibit partial hedonic

7

adaptation to disability. Based on two nationally representative panel studies, Lucas (2007b) shed light on the negative and long-lasting effects of disability on life satisfaction. Luhmann et al.’s (2012) meta-analysis integrates longitudinal data from 188 publications and examines the short- and long-term effects on subjective well-being of major family and work events, showing that different events differ in their effects on well-being, though these effects are not a function of the alleged desirability of events (see on this also Luhmann and Eid, 2009). Recent research has explored the links between natural disasters and subjective wellbeing. Luechinger and Raschky (2009) investigate the effects on life satisfaction of major floods for 16 European countries in the period from 1973 to 1998 and detect a significant and robust negative impact of floods on life satisfaction. Calvo et al. (2015) focus on pre- to postdisaster changes in happiness of 491 women affected by Hurricane Katrina, one of the most devastating and deadliest natural disasters in US history. They find that, for most respondents, happiness was lower one year after the event but was not significantly different from pre-disaster four years later. Next, several studies reveal that also crime and/or property victimization experiences negatively affect individuals’ current levels of subjective wellbeing (see, e.g. Cohen, 2008, for the US; Hanslmaier, 2013, for Germany; Staubli et al., 2014, for Switzerland; Mahuteau and Zhu, 2016, for Australia).7 Since in this paper we focus on extremely adverse life events out of individuals’ control that arguably are associated with current life satisfaction through psychological channels, the work that is closest to ours includes Frijters et al. (2011), Bellis et al. (2013), Moor and de Graaf (2016) and Mahuteau and Zhu (2016). Frijters et al. (2011) present Australian data examining the links between major positive and negative life events and life satisfaction, including serious personal illness or injury, death of a close relative (without distinguishing between death of a spouse and death of a child) and being victim of a property crime. Their findings indicate that, after two years, individuals adapt to a large extent to many negative events, but far less to death of a close relative and personal illness or injury. Based on UK 7

It is worth noting, however, that in general works on crime victimization (like the aforementioned studies and other research we refer to below, in this section) have no data about long-term effects of victimization on subjective well-being, as, with few exceptions, in these surveys respondents are asked if they had been victim of a robbery or other crime during the previous 12 months.

8

survey data, Bellis et al. (2013) find that subjective well-being is strongly associated with negative childhood experiences, that they assessed by means of dichotomous childhood measures aimed at capturing how happy one’s childhood was and how violent was one’s home life as a child. Mahuteau and Zhu (2016) rely on Australian panel data and show that property crime and (especially) physical violence are negatively correlated with subjective well-being. They also compare the well-being effects of these crime victimization experiences with three further major negative life events and document that physical violence is far more distressing than being fired. Using cross-sectional European Values Study data, Moor and de Graaf (2016) detect a strong negative effect of bereavement regarding close relatives (father, mother or child) on happiness. This relationship also exists in the case that the death of a close relative occurred more than 10 years before.8 Our work also speaks to the voluminous interdisciplinary body of research investigating the hedonic, emotional and behavioral effects of exposure to trauma over time. In medicine and psychiatry, several studies have shed light on the well-known “post-traumatic stress disorder” (Yehuda, 2002). As pointed out by Bernile et al. (2017), recent medical research indicates that traumatic experiences physiologically alter brain development and function and, in turn, influence subsequent behavior with regard to both economic and non-economic decision-making (see on this also Smith and Smith, 2010). Carmil and Breznitz (1991) document that exposure to traumatic life experiences produces enduring effects on political attitudes, religious identity and future orientation. Holman and Silver (1998) find that passing through traumatic experiences is associated with long-term psychological distress (see on this also Bunce et al., 1995). Frazier et al. (2001) examine the links between positive and negative life changes and post-traumatic distress on a sample of recent female sexual assault survivors. They show that positive changes were reported soon after the assaults and increased over 8

Also Oswald et al. (2015), in their experimental study with UK university students on happiness and productivity, examine the link between major real-world shocks and happiness. One of their experiments indicates that students who have recently suffered family tragedies (aggregated by using a single “bad life event” variable) are less happy than those who did not pass through similar happiness-shock experiences in their recent past. The effect of bad life events on well-being declines over time. However, it is important to note that this work covered recent bad life events only, as respondents had to declare whether they experienced one of the listed family tragedies within a 0-5 years range.

9

time, whereas there was a decrease in negative changes. Next, the relations with distress were stronger for negative changes. Edwards et al. (2003) assess the long-term mental health effects of different forms of childhood maltreatment and document an overall trend for poorer mental health as the number of abuse types increased. Based on UK panel data, Metcalfe et al. (2011) present quasi-experimental evidence of an international spillover of terrorism, showing that the September 11 attacks lowered levels of subjective well-being – assessed through a measure of mental distress – in a different country. Their findings suggest that major negative life events might produce detrimental effects on individuals’ well-being through the psychological channel (i.e., via increased fear and anxiety towards life) and, due to media coverage, today this may well occur also when individuals do not live in the country directly hit by the event. Metcalfe et al. (2011) also document that the negative effects lasted until the end of November in 2001 and then dissipated (see also Clark et al.’s, 2019, study, based on US data, on the (short-run) psychological effects of terrorism on well-being). Cornaglia et al. (2014) examine the links between crime and a detailed measure of mental well-being in Australia and detect a large decrease in mental well-being immediately after violent crime victimization. However, they also show that this effect dissipates very quickly. Next, they observe no impact of property crime victimization on mental health. Misheva (2016) employs one wave of twin data from Australia and shows that traumatic events such as being assaulted, being raped and being involved in an accident are negatively associated with emotional well-being. However, she also finds that these effects dissipate over time. Hentschel et al. (2017) also use Australian data and analyze the impact of major life events, compared to personality traits, on the stability of affective well-being. Their results indicate that financial worsening and serious personal injury/illness turn out to have the highest effects.

3. Data We implement our analysis using data from the Health and Retirement Study (henceforth HRS), a large-scale longitudinal survey on a representative sample of the US population aged 50 or more. The survey is run every two years, since 1992, by the Institute for Social

10

Research of the University of Michigan.9 Although primarily interested in studying the health conditions of the elderly population, over time the HRS questionnaire expanded to incorporate further topics. Nowadays the HRS provides detailed information on, e.g., employment, assets, and housing. A special psycho-social module, introduced in 2004, refers to such issues as the relationship with others, individual personality, and the occurrence of lifetime traumatic events in the past. In the HRS design, the module interviews in every wave a rotating half of the sample. As a result, the same individual answers to its questions every four years. The psycho-social module is the source of our key variables, most of which were available only between years 2006 and 2012 (crucially for our research, questions on lifetime traumatic events were excluded since the 2014 wave). Due to these constraints on the HRS design, our analysis is based on four waves, from 2006 to 2012. The final dataset is made of 13,006 observations on 6,503 individuals for which we have full information on all the variables under benchmark investigation. For each individual we have precisely two observations, one four years after the other (in 2006 and 2010 or in 2008 and 2012). The lack of a higher number of observations per individual limits within-individual variability10 and prevents us from properly exploiting fixed-effect methodologies. We then opt for random-effect models with Mundlak correction to proxy for fixed effects. The Mundlak (1978) approach consists in incorporating in the specification variables measuring the individual-specific average of the time-varying explanatory variables. Like fixed effects, these variables help to control for potential omitted variables. For each individual we have two observations on most of the variables listed in Table 1, that we can group in five broad categories: a) life satisfaction, b) lifetime traumatic events, c) personality, d) health, and e) socio-demographics. The exact wording of the variables

9

The datasets used in the analysis were derived from the 2006-2012 waves of the Health and Retirement Study, a public domain resource downloadable here (accessed March 20 2019): https://hrs.isr.umich.edu/data-products/access-to-public-data. Derived data are available from the corresponding author upon request. All the analyses in this study are performed using the software StataTM Special Edition v. 15.0.

10

Up to 91% of the total variability among the traumatic event variables is between-group rather than withingroup.

11

belonging to groups a)-d) is reported in the Supplemental Information. For sake of comparability, all variables in categories b)-d) take values in the 0-1 range. Our key variables are those on life satisfaction, that we assess through an aggregate indicator of general life satisfaction (in our benchmark analysis) as well as through domainspecific measures (on six different domains such as satisfaction with home, leisure, family life, city, finances and health).11 All these measures are ordinal variables taking values on a Likert scale from 1 to 5, with higher values indicating higher life satisfaction. General life satisfaction is on average worth 3.593 with a median of 4, although the distribution of answers covers the whole range of possible values (see Figure 1). Respondents are also more likely to be satisfied with their house and leisure activities, than with their finances or health conditions. This evidence may partly reflect the fact that our sample period incorporates the recent economic and financial crisis, and that the individuals involved in the study are relatively old.

Figure 1. Life Satisfaction The lifetime traumatic event group includes four events that may have occurred at any point in life (the death of a child, a serious physical attack or assault, and a life-threatening illness or accident involving the respondent or her partner/child) and one event that occurred 11

The domain-specific measures are available only in the 2008-2012 waves, with the exception of the measure on finances that is available since 2006. This means that only for general life satisfaction and satisfaction about personal finances we have exactly two observations available per individual.

12

in the distant past, i.e., before age 18 (physical abuse by either parent). In principle, this list of extremely adverse events could be expanded and include five more events that HRS collects (natural disasters, fired in a combat, partner/child addicted to drugs or alcohol, years of school repeated, parents addicted to drugs or alcohol). However, even though we consider these events in a robustness check, we decided to exclude them from the main analysis due to endogeneity concerns, as we fear that the occurrence of these traumatic events might not have been out of our respondents’ control.12 From Table 1 we learn that the events, although infrequent, are not rare: they arise in between 6.0% (for physical attacks/assaults) and 28.7% (for life-threatening illnesses/accidents hitting the spouse or a child) of the sample. These figures are in line with Bucciol and Zarri (2015), who use the same dataset to study the correlation between traumatic events and stock holding. It is worth noting that the frequency of these events is probably higher than in the population, because individuals in our sample are generally older and, therefore, have had more time to experience events. Importantly, for the four events that may arise at any point in time, we know the year in which they happened (on average, about 20 years before the interview): we exploit this information on the timing of events to see whether events occurred many years before are correlated with life satisfaction. We distinguish between traumatic events experienced within 10 years, between 11 and 30 years, and over 30 years before to have a roughly balanced distribution of the answers. Unfortunately we cannot go more into detail as the distribution of the events is rather sparse. Indeed, in some periods we would observe too rare occurrence of some events (especially on physical attack) if we considered narrower year ranges. The personality group includes a set of nine indexes that can be derived from the HRS questionnaire, and are meant to assess the following aspects of personality: openness to experience, conscientiousness, extraversion, agreeableness, neuroticism (these five items denote the well-known “Big Five” taxonomy of personality; see on this Costa and McCrae, 1992), cynical hostility, anxiety,

12

For instance, an individual could have a lifestyle that, more or less subconsciously, makes it more likely for her to be exposed to a natural disaster, or she could undertake a behavior that induces a partner or child to get addicted. This gives rise to self-selection, that might bias our estimates.

13

anger-in and anger-out.13 We decided to control for personality traits as a growing body of research has been showing that subjective well-being is associated with personality characteristics (see, e.g., Hayes and Joseph, 2003, and Weiss et al., 2008). The category on health includes self-reported health status, plus three scores related to functional limitations: Mobility and Strength (MS), Activities of Daily Living (ADL) and Instrumental Activities of Daily Living (IADL). All scores are measured in such a way that, the higher the score, the fewer functional limitations. We take account of these variables as a potential further channel of correlation with life satisfaction. The idea is that the correlations we may eventually find between life satisfaction and traumatic events are mediated, at least partly, by the occurrence of functional limitations due to the traumatic events. More specifically it could be that, after experiencing traumatic events, people face poor health conditions as well as functional limitations later in life (see, e.g., Lantz et al., 2005, on this). Variable Life satisfaction (general) Satisfied with house Satisfied with leisure Satisfied with family life Satisfied with city Satisfied with finances Satisfied with health Child death Within 10 years Within 11-30 years Over 30 years Physical attack/assault Within 10 years Within 11-30 years Over 30 years Illness/Accident Within 10 years Within 11-30 years Over 30 years Illness/Accident (others) Within 10 years Within 11-30 years Over 30 years Physical abuse (<18)

Median Mean Std. dev. a) Life Satisfaction (score) 4 3.593 0.960 4 4.199 0.865 4 4.163 0.864 4 3.927 0.900 4 4.064 0.896 3 3.272 1.128 3 3.403 1.022 b) Lifetime Traumatic Events 0 0.145 0.353 0 0.049 0.217 0 0.041 0.199 0 0.055 0.227 0 0.060 0.237 0 0.016 0.127 0 0.015 0.123 0 0.028 0.164 0 0.257 0.437 0 0.132 0.338 0 0.066 0.248 0 0.060 0.237 0 0.287 0.452 0 0.155 0.362 0 0.091 0.287 0 0.041 0.197 0 0.072 0.258

Minimum

Maximum

1 1 1 1 1 1 1

5 5 5 5 5 5 5

0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 (Continues in the next page)

Table 1. Summary Statistics

13

The difference between anger-in and anger-out is that anger-out refers to the predisposition to express angry feelings towards others, whereas anger-in is based on suppressing those feelings and holding them in.

14

Variable Openness Conscientiousness Extraversion Agreeableness Neuroticism Cynical hostility Anxiety Anger-in Anger-out Good health MS score ADL score IADL score Age Woman Non-white Immigrate Married High school College Employee Self-employed Income (k USD) Fin. wealth (k USD) Real wealth (k USD) Home

Median

Mean c) Personality 0.667 0.653 0.733 0.691 0.733 0.737 0.867 0.846 0.417 0.439 0.360 0.372 0.133 0.171 0.333 0.392 0.143 0.163 d) Health 0 0.478 0.833 0.774 1 0.963 1 0.969 e) Socio-demographics 67 66.436 1 0.592 0 0.147 0 0.060 1 0.693 0 0.195 0 0.112 0 0.306 0 0.093 47.025 72.680 19.488 150.416 148.970 291.149 1 0.865

Std. dev.

Minimum

Maximum

0.182 0.132 0.182 0.156 0.154 0.218 0.183 0.226 0.164

0 0 0 0 0 0 0 0 0

1 1 1 1 1 1 1 1 1

0.500 0.221 0.125 0.096

0 0.083 0 0

1 1 1 1

7.198 0.491 0.354 0.238 0.461 0.396 0.315 0.461 0.291 116.932 544.447 774.770 0.342

50 0 0 0 0 0 0 0 0 0 0.001 0.001 0

80 1 1 1 1 1 1 1 1 5,613.675 3.29e4 3.26e4 1

Note: 13,006 observations for all the variables except for the domain-specific ones that are available on fewer observations, between 9,381 and 12,964. The variables in italics are dummy. Table 1. (Continued) Finally, the socio-demographic group of variables includes standard information on age, gender, ethnicity, nationality, marital status, occupational status, and education of the head, plus income and wealth of the household. Notice that age is limited in the 50-80 range, since, as a further restriction, we decided to focus on individuals not older than 80, as the elderly are over-sampled in HRS. In addition, as Bucciol and Zarri (2015) noted, such individuals may find it difficult to recall past personal events, especially when such events occurred in their early life.14

14

Recent evidence documents that retrospective, self-reported information on childhood diseases, accidents, hospitalizations and socio-demographic information (including empirical studies using HRS data) has a high level of accuracy also when the reported events occurred several decades before (Havari and Mazzonna, 2011). Next, as to the timing of the major life events we examine in our analysis, we believe that considering

15

3.1. Research Questions With this work we aim to test a set of four hypotheses. Hypothesis 1. There is negative correlation between general life satisfaction and the occurrence of a lifetime traumatic event. As to Hypothesis 1, while we know from prior work that there might be a positive legacy of trauma (Tedeschi and Calhoun, 1996), we view it as more natural, based on the streams of literature reviewed in Section 2, to conjecture that exposure to major events and life satisfaction are negatively correlated. Regarding this hypothesis, Table 2 reports a first analysis comparing the median life satisfaction conditional on the occurrence of a traumatic event. For two traumatic events – physical attack/assault and being physically abused – the median life satisfaction with the event is one point lower (3 instead of 4). More generally, for all traumatic events a Mann-Whitney test rejects at common significance levels the null hypothesis that the distribution is identical in the two groups. The evidence, however, could be biased by the lack of controls for the sample characteristics in the two groups. Hypothesis 2. The correlation falls as the traumatic event is farther in time. Also with regard to Hypothesis 2, even though some works detect long-run effects of traumatic events (see, e.g., Moor and de Graaf, 2016, for bereavement regarding close relatives) and, as noted by Powdthavee (2009), longitudinal evidence from prior work has shown that individuals are unlikely to adapt completely to adverse events such as unemployment and disability, we believe that – in line with most of the studies examining the timing issue we refer to in Section 2 – it is natural to expect that the correlation between general life satisfaction and exposure to traumatic event falls as time unfolds.

three time windows (i.e., within 10 years, between 11 and 30 years, and over 30 years), rather than the specific years in which the events occur, further mitigate “recall bias” concerns.

16

Hypothesis 3. There is a negative correlation between domain-specific measures of life satisfaction and the occurrence of a lifetime traumatic event. As to Hypothesis 3, the main reason why we are interested not only in exploring general life satisfaction but also several domain-specific measures of satisfaction with life is that the latter arguably play an important complementary role. As pointed out by Van Praag et al. (2003), who developed a joint model based on satisfaction with life as a whole and on domain satisfactions, general life satisfaction can be viewed as an aggregate construct. In principle, as noted by Powdthawee (2009) with regard to disability, a given life event may have different effects on different life domains, so that the balance effect on overall life satisfaction may be difficult to measure and interpret. Therefore, compared to the general life satisfaction measure, domain-specific ones make it possible to see (i) which aspects of life, if any, are significantly correlated with (some of) the traumatic events under study and, when this is the case, (ii) whether the detected correlation is positive or negative. Prior work has documented that adverse life events such as, e.g., becoming disabled are negatively associated with life satisfaction in several domains (Powdthavee, 2009). As we made clear in Section 1, we believe that including also domain-specific measures of life satisfaction in our analysis allows us to provide a more comprehensive, richer picture of the relationships between the set of the adverse life events under study and individuals’ life satisfaction. In line with existing work on the effects of life events on different domains of life satisfaction (see e.g. Andreß and Brockel, 2007; Powdthavee, 2009; Georgellis et al., 2012), we conjecture that exposure to traumatic events and domain-specific life satisfaction are negatively correlated. Hypothesis 4. There are gender differences, with women being more affected than men by traumatic events involving a close relative (child or partner), rather than themselves. Hypothesis 4 draws from our expectation that women are more affected than men in their life satisfaction when traumatic events regard a close relative (child or partner), rather than themselves (see, on this, Moor and De Graaf, 2016): as recalled by Eckel and Grossman

17

(1998), prior work in psychology, sociology and political science clearly indicates that in non-economic settings women are more socially-orientated and less selfish than men. Eckel and Grossman’s (1998) study extends this finding to economic settings. Our analysis in Section 4 digs deeper into these issues and tests our hypotheses, also controlling for the observable characteristics of the respondent.

Traumatic Event Child death Physical attack/assault Illness/Accident Illness/Accident (others) Physical abuse (<18)

Median Life Satisfaction With Traumatic Without Traumatic Event Event 4 3 4 4 3

4 4 4 4 4

Test Statistic

P-value

5.242 10.877 8.170 4.118 10.272

0.000 0.000 0.000 0.000 0.000

Note: 13,006 observations. The last two columns report the statistic and the p-value of a Mann-Whitney test on the comparison of the distribution in the two sub-groups of observations with and without the experience of the traumatic event. The null hypothesis is that the distribution is identical; the alternative hypothesis is that it is different. Table 2. Average Life Satisfaction by Traumatic Event

4. Results This section is organized in four sub-sections, where each sub-section tests one of the hypotheses listed in Sub-section 3.1. Sub-section 4.1 discusses the negative correlation between life satisfaction and lifetime traumatic events (Hypothesis 1); Sub-section 4.2 checks whether the correlation is persistent over time (Hypothesis 2); Sub-section 4.3 studies whether domain-specific life satisfaction is correlated with some lifetime traumatic events (Hypothesis 3); finally, Sub-section 4.4 verifies if the correlation differs by gender (Hypothesis 4). The regression output is reported in Tables 3-6. Appendix Tables S1-S4 display average marginal effects for the probability to reach maximum life satisfaction (i.e., a value of 5). Table S5 shows a robustness check on Table 3, using an extended set of lifetime traumatic events, including five more events that, as we explained in the previous section, may suffer from endogeneity; our key results are preserved.

18

A recent critique in Bond and Lang (2019) argues that common subjective measures of well-being are analyzed in the wrong way because standard methods focus on mean statistics, and thus treat the variables as cardinal even if they are ordinal. This prevents one from properly comparing two groups of respondents, unless strong and unrealistic conditions on the distribution are satisfied,15 because the distance between two values is arbitrary and carries no information. In fact, it is possible to take simple monotonic transformations of the well-being measure, which preserve the original ranking, and still get different conclusions from methods assuming cardinality. This problem also holds for our measures of life satisfaction, which are based on discrete ordinal scales. Chen et al. (2019) suggest to address the issue by looking at median rankings rather than mean rankings: unlike the mean, the median preserves rankings because it is invariant to monotonic transformations.16 Ordered probit and logit regressions are preferable candidate models for analyzing discrete ordinal variables. In the context of (ordered) probit and logit regressions, moreover, means and medians coincide due to the symmetry of the normal distribution. Chen et al. (2019) then suggest to consider this type of models, and interpret their output in terms of median changes. We take the point of Chen et al. (2019) and perform our analyses through (random-effect) ordered probit regressions. All models include Mundlak corrections to capture individual-specific effects, and year and region fixed effects to capture potential variations across time and space.17 A potential issue, following Lantz et al.’s (2005) argument, is that we cannot exclude that traumatic life events are more likely to hit individuals with lower socio-economic status. To partially control for this concern, our models include socio-economic variables in the specification. In 15

When the statistic of interest is the mean, it must be that the distribution of one group first-order stochastically dominates the distribution of the other.

16

For this reason in Table 2 we compared groups in terms of their medians.

17

Year and region fixed effects are measured through a set of dummy variables. For year fixed effects we consider three dummies equal to one if the year of data collection is 2008, 2010 or 2012 respectively (reference category: 2006). For region fixed effects we consider eight dummies equal to one if the respondent lives in one of the Census-Bureau divisions (New England, Mid-Atlantic, East-North Central, West-North Central, South Atlantic, East-South Central, West-South Central, Mountain). The reference category is Pacific.

19

what follows, we generally adopt the convention to comment only on coefficients that are significant at least at the 5% level.

4.1. Life Satisfaction and Lifetime Traumatic Events Table 3 reports our benchmark coefficient estimates, where the dependent variable is the index of general life satisfaction, measured in the 1-5 discrete range. Column (1) runs a first random-effect ordered probit regression over the socio-demographic variables; all the thresholds are significant apart from the first one, to indicate that the five possible levels of life satisfaction generally follow different equations. We see that median life satisfaction is higher with women, immigrates, married individuals, high school and college graduates, as well as individuals living in richer households. This evidence is in line with prior studies (see, e.g., Clark and Oswald, 1994; Blanchflower and Oswald, 2004, and Graham and Chattopadhyay, 2013). Column (2) adds to the model the variables on personality and health. As to personality, we learn that median life satisfaction is higher with conscientiousness and extraversion, and lower with neuroticism, cynical hostility, anxiety and anger (in and out). Our findings are broadly consistent with available empirical evidence on the links between personality traits and subjective well-being (see, e.g., Hayes and Joseph, 2003, and Weiss et al., 2008). As to health, we find that self-assessed good health, the MS and ADL scores are always significantly positive, while the IADL score is not significant.18 The coefficients thus indicate that, the better the respondent in general health, mobility, strength and activities of daily living, the higher her perception on life satisfaction. Column (3) adds to the specification the traumatic events variables. We see that three out of the five extremely adverse life events exhibit significantly negative correlation with life satisfaction: physical attack/assault, physical abuse when younger than 18, and lifethreatening illness/accident of the partner/child. In terms of size, the first two variables seem much more relevant (see Appendix Table S1): the median probability to reach maximum life 18

This variable, however, shows small variability in the sample, as most of the respondents report no functional limitations on the instrumental activities of daily living (see Table 1).

20

satisfaction (i.e., a value of 5) is 3.30% lower after the occurrence of a physical attack/assault, and 2.60% lower after a physical abuse, while it is 0.90% lower with the experience of an illness/accident hitting the spouse or a child. These probability changes are broadly in line with some other variables on personality (e.g., a 3.66% decrease with one standard deviation increase in anxiety)19, health (e.g., a 4.70% increase with good health conditions) and socio-demographics (e.g., a 5.70% increase with being married). Even though we unexpectedly find that child death is not significantly associated with life satisfaction, on the whole our evidence provides support to Hypothesis 1. It is also interesting to mention that, with the exclusion of the health-related variables from the specification, the coefficient on personal illness/accident also becomes significantly negative.20 4.2. Timing of the Traumatic Events One of the five traumatic events regards a period of life where the respondent was in her formative years, i.e., before age 18; the remaining four traumatic events may have occured in any phase of life. We exploit this information and, in Column (4) of Table 3, we replace each traumatic event dummy with a set of three dummy variables, meant to capture the occurrence of the traumatic event in a given time window: within 10 years, between 11 and 30 years, over 30 years ago. The distinction is made in such a way to have a roughly similar number of events in the three time windows, and could loosely correspond to a distinction between short, medium and long term. Unfortunately we cannot consider a more accurate distinction, as the distribution of the events in our dataset is quite sparse.

19

Computed as the average marginal effect on Table S1 times one standard deviation in anxiety, -0.200*0.183.

20

This result is not surprising as personal illness/accident is closely related to physical conditions. Indeed, subjects in the sample reporting a personal illness/accident show poorer health conditions (Wilcoxon ranksum test: 17.01; p-value <0.01), MS (t-test: 15.24; p-value <0.01), ADL (t-test: 10.58; p-value <0.01) and IADL (t-test: 5.63; p-value <0.01) scores than subjects reporting no personal illness/accident. We read this result as indicating that the correlation we detected between life satisfaction and personal illness/accident is entirely attributable to the occurrence of functional limitations, while the correlations involving physical attack/assault, illness/accident occurred to others and physical abuse are nearly unaffected by the healthrelated variables.

21

(1)

(2)

Child death

(3) -0.034 (0.041)

Within 10 years

(4) -0.056 (0.062) 0.017 (0.070) -0.051 (0.062)

Within 11-30 years Over 30 years Physical attack/assault

-0.207*** (0.056)

Within 10 years

-0.297*** (0.096) -0.199* (0.103) -0.151* (0.078)

Within 11-30 years Over 30 years Illness/Accident

-0.028 (0.031)

Within 10 years

-0.056 (0.039) 0.010 (0.052) -0.002 (0.056)

Within 11-30 years Over 30 years Illness/Accident (others)

-0.059** (0.029)

Within 10 years Within 11-30 years Over 30 years Physical abuse (<18) Openness

-0.044 (0.097) 0.508*** (0.121) 1.149*** (0.102) 0.092 (0.112) -0.616*** (0.096) -0.367*** (0.067) -1.287*** (0.085) -0.432*** (0.063) -0.375*** (0.087)

Conscientiousness Extraversion Agreeableness Neuroticism Cynical hostility Anxiety Anger-in Anger-out

-0.056 (0.036) -0.038 (0.045) -0.112* (0.065) -0.166*** -0.170*** (0.053) (0.053) -0.003 -0.006 (0.097) (0.097) 0.502*** 0.503*** (0.121) (0.121) 1.145*** 1.147*** (0.102) (0.102) 0.097 0.096 (0.112) (0.112) -0.625*** -0.625*** (0.095) (0.096) -0.359*** -0.359*** (0.067) (0.067) -1.259*** -1.255*** (0.085) (0.085) -0.415*** -0.417*** (0.063) (0.063) -0.350*** -0.352*** (0.087) (0.087) (Continues in the next page)

Table 3. Benchmark Analysis

22

(1) Good health MS score ADL score IADL score Age/10 Woman Non-white Immigrate Married High school College Employee Self-employed Ln(1+income) Ln(1+fin. wealth) Ln(1+real wealth) Home Constant Threshold 1 Threshold 2 Threshold 3 Threshold 4 Region effects Year effects Mundlak effects Log-likelihood Households Observations

0.043 (0.259) 0.168*** (0.035) -0.039 (0.050) 0.138* (0.072) 0.345*** (0.080) 0.085* (0.045) 0.342*** (0.057) 0.007 (0.052) 0.014 (0.084) -0.009 (0.017) 0.015*** (0.005) 0.008 (0.008) 0.031 (0.085) 1.144*** (0.048) 0.139 (0.291) 1.203*** (0.291) 2.769*** (0.293) 4.589*** (0.295) YES YES YES -15,894.320 6,503 13,006

(2) 0.303*** (0.030) 0.422*** (0.078) 0.305** (0.124) 0.078 (0.149) 0.120 (0.259) 0.087*** (0.033) -0.075* (0.045) 0.166*** (0.064) 0.365*** (0.080) -0.016 (0.040) 0.215*** (0.052) -0.028 (0.052) -0.038 (0.084) -0.011 (0.017) 0.013** (0.005) 0.006 (0.008) 0.011 (0.085) 0.774*** (0.038) -0.285 (0.311) 0.793** (0.311) 2.381*** (0.312) 4.209*** (0.315) YES YES YES -15,051.063 6,503 13,006

(3) 0.299*** (0.030) 0.399*** (0.078) 0.299** (0.124) 0.065 (0.149) 0.118 (0.259) 0.091*** (0.033) -0.080* (0.045) 0.157** (0.064) 0.359*** (0.080) -0.013 (0.040) 0.215*** (0.052) -0.032 (0.052) -0.040 (0.084) -0.010 (0.017) 0.013** (0.005) 0.006 (0.008) 0.007 (0.085) 0.767*** (0.038) -0.382 (0.312) 0.699** (0.312) 2.288*** (0.313) 4.115*** (0.315) YES YES YES -15,032.946 6,503 13,006

(4) 0.298*** (0.030) 0.397*** (0.078) 0.293** (0.124) 0.067 (0.149) 0.121 (0.260) 0.091*** (0.033) -0.081* (0.045) 0.158** (0.064) 0.359*** (0.080) -0.013 (0.040) 0.217*** (0.052) -0.033 (0.052) -0.043 (0.084) -0.010 (0.017) 0.013** (0.005) 0.006 (0.008) 0.008 (0.085) 0.767*** (0.038) -0.381 (0.313) 0.700** (0.312) 2.290*** (0.313) 4.117*** (0.316) YES YES YES -15,030.363 6,503 13,006

Note: Standard errors in parentheses; *** p<0.001, ** p<0.01, * p<0.05. Table 3. (Continued) Here we confirm a significant correlation for the same traumatic events as in Column (3). It is interesting to note that, while for illness/accidents occurred to others the correlation holds only in the long run (over 30 years), for physical attack/assault it is persistent over time. Looking at the coefficients for this traumatic event, it seems that the correlation is

23

weaker as time goes by. However, a Chi-squared test accepts the null hypothesis of equality of the three coefficients for physical attack/assault (statistic: 1.51; p-value: 0.47). We cannot run the same exercise for the physical abuse dummy. Nevertheless, since this traumatic event occurred before age 18, we can view also this effect as persistent because it is still relevant for the respondent. Regarding Hypothesis 2, we can then conclude that it is only partly supported by our data: the correlation between life satisfaction and a traumatic event decays over time for an illness or accident, while it remains persistent for a physical attack/assault and for being physically abused (an event that for sure occurred more than three decades before). 4.3. Specific Domains of Life Satisfaction So far we considered only the standard notion of general life satisfaction. However, prior work has shown that more specific measures of life satisfaction, referring to particular aspects of life, also depend on objectively measurable variables and are significantly associated with general life satisfaction (see, e.g., Van Praag et al., 2003). Therefore, we sought to see whether, once we take into account specific measures of life satisfaction related to different domains of life, they turn out to exhibit negative correlation with our traumatic events, as we expected, or not. To this aim, Table 4 replicates the analysis of Column (3) in Table 3 using six domain-specific measures of life satisfaction. The new measures refer to satisfaction for the house, for leisure activities, for family life, for the city, for the financial situation and for personal health. The dependent variables are all ordinal indicators in the 1-5 scale; for this reason we keep on treating random-effect ordered probit models. Only the variable on satisfaction about personal finances was collected as early as in year 2006. This implies that, for all the other variables, we have one observation for some respondents and two observations for the others. To avoid over-representation of some respondents, we run weighted regressions where respondents with two observations receive a weight of one and respondents with one observation receive a weight of two. This way, all individuals count the same in the analysis.

24

Life Satisfaction Child death Physical attack/assault Illness/Accident Illness/Accident (others) Physical abuse (<18) Openness Conscientiousness Extraversion Agreeableness Neuroticism Cynical hostility Anxiety Anger-in Anger-out Good health MS score ADL score IADL score Constant Threshold 1 Threshold 2 Threshold 3 Threshold 4 Socio-demographics Region effects Year effects Mundlak effects Log-likelihood Households Observations

(1) House 0.017 (0.043) -0.218*** (0.061) -0.077** (0.034) -0.151*** (0.033) -0.158*** (0.056) -0.303*** (0.106) 0.964*** (0.131) 0.833*** (0.110) 0.778*** (0.121) -0.504*** (0.107) -0.313*** (0.074) -0.974*** (0.094) -0.215*** (0.069) -0.357*** (0.095) 0.189*** (0.033) -0.067 (0.085) 0.245* (0.130) -0.390** (0.155) 1.034*** (0.072) -0.690** (0.328) 0.316 (0.327) 1.772*** (0.328) 3.420*** (0.333) YES YES YES YES -13,657.138 10,082 12,990

(2) Leisure -0.113** (0.045) -0.339*** (0.064) -0.034 (0.035) -0.063* (0.034) -0.041 (0.058) -0.262** (0.109) 0.500*** (0.135) 1.051*** (0.114) 0.882*** (0.125) -0.226** (0.110) -0.561*** (0.076) -0.947*** (0.097) 0.020 (0.072) -0.591*** (0.098) 0.123*** (0.034) 0.043 (0.088) -0.038 (0.136) -0.014 (0.159) 1.215*** (0.079) -0.587* (0.342) 0.528 (0.339) 2.031*** (0.341) 3.755*** (0.346) YES YES YES YES -13,960.670 10,071 12,974

(3) Family life -0.025 (0.041) -0.239*** (0.059) -0.038 (0.032) -0.143*** (0.031) -0.134** (0.053) 0.231** (0.099) 0.722*** (0.123) 1.770*** (0.107) 0.423*** (0.114) -0.768*** (0.101) -0.364*** (0.070) -1.552*** (0.091) -0.454*** (0.065) -0.392*** (0.090) 0.306*** (0.031) 0.190** (0.080) 0.617*** (0.124) 0.398*** (0.146) 0.857*** (0.061) 0.568* (0.308) 1.926*** (0.309) 3.783*** (0.314) 5.498*** (0.321) YES YES YES YES -13,955.613 10,049 12,958

(4) City -0.006 (0.044) -0.319*** (0.063) 0.031 (0.034) -0.185*** (0.033) -0.335*** (0.056) -0.209* (0.107) 0.523*** (0.132) 1.142*** (0.111) 1.026*** (0.122) -0.760*** (0.108) -0.522*** (0.075) -1.505*** (0.096) -0.682*** (0.070) -0.505*** (0.096) 0.220*** (0.034) -0.090 (0.085) 0.271** (0.132) -0.136 (0.156) 1.124*** (0.076) -1.193*** (0.331) 0.125 (0.328) 1.718*** (0.330) 3.500*** (0.335) YES YES YES YES -13,961.219 10,068 12,966

(5) Finances 0.024 (0.027) -0.147*** (0.041) -0.055** (0.022) -0.049** (0.021) 0.009 (0.037) -0.064 (0.067) 0.169* (0.087) 0.304*** (0.071) 0.026 (0.079) -0.056 (0.070) -0.039 (0.048) -0.469*** (0.062) -0.136*** (0.045) -0.106* (0.061) 0.096*** (0.022) 0.004 (0.056) 0.101 (0.093) 0.033 (0.115) 0.000 (0.000) 0.084 (0.210) 0.933*** (0.210) 1.974*** (0.210) 2.776*** (0.210) YES YES YES YES -18,396.211 6,542 13,004

(6) Health 0.029 (0.039) -0.223*** (0.058) -0.366*** (0.031) -0.095*** (0.030) -0.185*** (0.051) 0.274*** (0.095) 0.529*** (0.119) 1.238*** (0.100) 0.009 (0.109) -0.299*** (0.097) -0.035 (0.066) -1.793*** (0.089) -0.143** (0.062) -0.285*** (0.087) 1.227*** (0.034) 1.723*** (0.082) 0.845*** (0.123) -0.201 (0.143) 0.730*** (0.056) 1.386*** (0.297) 2.830*** (0.300) 4.818*** (0.307) 6.655*** (0.316) YES YES YES YES -14,142.036 10,069 12,980

Note: Weighted regression, with weight equal to one for respondents with two observations and weight equal to two for respondents with one observation. Standard errors in parentheses; *** p<0.001, ** p<0.01, * p<0.05. Table 4. Specific Domains

25

From a broad picture, the traumatic events displaying stable correlation with each measure of life satisfaction are physical attack/assault and illness/accident to a partner/child, although with different sizes. In particular the median probability to reach maximum life satisfaction (i.e., a value of 5; see Appendix Table S2) falls after a physical attack/assault between 2.90% (life satisfaction on health) and 11.1% (life satisfaction on leisure) – the latter effect is very large compared to the corresponding effect of general life satisfaction (-3.3%) and seems to indicate that a traumatic event such as physical attack/assault significantly impacts individuals’ life satisfaction with regard to the quality of their leisure. Each of the five traumatic events, moreover, has at least three significant correlations. We highlight two results. First, the measures of life satisfaction involving the financial situation and leisure activities are the only ones to be uncorrelated with traumatic events arisen before age 18. Second, the death of a child – that, so far, turned out to be surprisingly irrelevant in our analysis – is negatively correlated with life satisfaction on leisure. In this regard, it is plausible that passing through an important bereavement such as the loss of a child lowers the quality of one’s leisure, likely due to memories of him or her and thoughts about how leisure would be more meaningful and rewarding today if the child were alive. On the whole, then, our findings on domain-specific life satisfaction provide partial support to Hypothesis 3, with the correlation between life satisfaction and lifetime traumatic event depending on the domain under consideration. 4.4. Gender Dimension In this sub-section we test our hypothesis that women are more affected than men by traumatic events involving a close relative (child or partner), rather than themselves, using the output in Tables 5 and 6, which replicate our previous analysis respectively for men and women. Specifically, we repeat the regression of Column (3) in Table 3 and the six regressions of Table 4 in the two sub-samples of men and women. This approach is more general than incorporating interactions between gender and life events in the same model, because it also allows for possible gender differences in all the explanatory variables. As for Table 4, we rescale the observations to give the same weight to each respondent.

26

From the comparison between the two groups we notice a number of interesting findings. First, general life satisfaction of men (Column 1 of Table 5) is virtually unrelated to the occurrence of a lifetime traumatic event. The only exception regards having suffered for a physical attack/assault, in which case the coefficient is significantly negative (-0.162) and in line with the one found on women (Column 1 of Table 6).21 Second, from Column (1) of Table 6 we find that the same three events we detected in Sub-section 4.1 remain negatively correlated with life satisfaction also in the sub-sample of women, and with similar coefficients and effects. Further, we surprisingly keep on detecting no gender differences with regard to a traumatic event such as passing through the death of a child. The only exception is the negative correlation with life satisfaction on leisure activities that we observe only among females. Going into the details of specific domains of life satisfaction, we learn some regularities. The most evident one is the always significantly negative effect of a physical attack/assault on life satisfaction: it happens in all domains and for both genders. In general, we observe on females fewer significant effects for illnesses/accidents (2 instead of 3 out of 7), and more significant effects for illness/accident of a partner/spouse (5 instead of 4) and physical abuse (5 instead of 3). Females display more correlation than males on general life satisfaction, as well as on domains related to house and health conditions, and fewer correlations on the financial domain. The difference between males and females is not strong, though.

21

The difference on this coefficient between the two sub-samples is not significantly different from zero (Chi squared test: 0.01; p-value: 0.94).

27

Life Satisfaction Child death Physical attack/assault Illness/Accident Illness/Accident (others) Physical abuse (<18) Openness Conscientiousness Extraversion Agreeableness Neuroticism Cynical hostility Anxiety Anger-in Anger-out Good health MS score ADL score IADL score Constant Threshold 1 Threshold 2 Threshold 3 Threshold 4 Socio-demographics Region effects Year effects Mundlak effects Log-likelihood Households Observations

(1) General 0.017 (0.066) -0.162* (0.088) -0.033 (0.045) -0.053 (0.047) -0.111 (0.085) -0.146 (0.149) 0.574*** (0.179) 1.247*** (0.153) 0.150 (0.159) -0.636*** (0.148) -0.339*** (0.101) -1.277*** (0.134) -0.296*** (0.096) -0.209* (0.125) 0.318*** (0.045) 0.441*** (0.126) 0.456** (0.210) 0.290 (0.248) 0.631*** (0.053) 0.270 (0.476) 1.368*** (0.476) 2.959*** (0.478) 4.753*** (0.482) YES YES YES YES -6,044.855 2,650 5,300

(2) House -0.049 (0.070) -0.376*** (0.098) -0.076 (0.050) -0.066 (0.053) -0.160* (0.090) -0.576*** (0.164) 1.162*** (0.196) 0.784*** (0.167) 1.239*** (0.176) -0.793*** (0.169) -0.356*** (0.114) -1.124*** (0.149) -0.052 (0.107) -0.368*** (0.139) 0.165*** (0.051) -0.366*** (0.137) 0.522** (0.224) -0.290 (0.252) 0.896*** (0.109) -0.159 (0.511) 0.816 (0.509) 2.266*** (0.513) 4.000*** (0.522) YES YES YES YES -5,340.900 4,157 5,296

(3) Leisure -0.118 (0.074) -0.488*** (0.105) -0.035 (0.052) -0.128** (0.056) 0.068 (0.096) -0.444** (0.173) 0.575*** (0.207) 1.204*** (0.178) 1.115*** (0.185) -0.396** (0.178) -0.520*** (0.120) -1.064*** (0.158) 0.100 (0.114) -0.625*** (0.148) 0.110** (0.053) -0.016 (0.145) 0.047 (0.237) 0.198 (0.266) 1.222*** (0.129) 0.538 (0.547) 1.706*** (0.546) 3.218*** (0.552) 5.014*** (0.563) YES YES YES YES -5,682.650 4,159 5,292

(4) Family life -0.107 (0.066) -0.270*** (0.095) -0.098** (0.047) -0.186*** (0.050) -0.078 (0.085) 0.125 (0.155) 0.851*** (0.185) 1.634*** (0.162) 0.729*** (0.164) -1.027*** (0.161) -0.372*** (0.108) -1.818*** (0.147) -0.188* (0.101) -0.420*** (0.132) 0.294*** (0.048) 0.005 (0.130) 0.733*** (0.213) 0.829*** (0.242) 0.760*** (0.091) 1.671*** (0.487) 3.016*** (0.490) 4.831*** (0.501) 6.627*** (0.516) YES YES YES YES -5,521.745 4,153 5,284

(5) City -0.063 (0.076) -0.198* (0.107) 0.006 (0.053) -0.218*** (0.057) -0.339*** (0.096) -0.320* (0.176) 0.883*** (0.209) 0.976*** (0.181) 1.256*** (0.187) -0.647*** (0.180) -0.533*** (0.122) -1.764*** (0.163) -0.643*** (0.116) -0.419*** (0.149) 0.288*** (0.055) -0.077 (0.147) 0.233 (0.237) 0.340 (0.268) 1.251*** (0.135) 0.014 (0.549) 1.334** (0.548) 2.898*** (0.554) 4.816*** (0.567) YES YES YES YES -5,478.016 4,160 5,292

(6) Finances 0.028 (0.046) -0.206*** (0.066) -0.069** (0.033) -0.072** (0.035) -0.010 (0.062) 0.074 (0.108) 0.229* (0.134) 0.228** (0.111) 0.125 (0.117) -0.112 (0.113) 0.020 (0.076) -0.526*** (0.101) -0.121* (0.072) -0.014 (0.091) 0.084** (0.034) -0.035 (0.094) 0.220 (0.165) 0.176 (0.195) 0.000 (0.000) 0.548 (0.337) 1.448*** (0.337) 2.507*** (0.337) 3.351*** (0.338) YES YES YES YES -7,368.386 2,663 5,300

(7) Health -0.014 (0.062) -0.267*** (0.092) -0.397*** (0.045) -0.059 (0.048) -0.221*** (0.082) 0.226 (0.146) 0.657*** (0.177) 1.275*** (0.153) 0.183 (0.156) -0.503*** (0.152) -0.004 (0.101) -1.911*** (0.141) -0.012 (0.096) -0.363*** (0.127) 1.189*** (0.052) 1.572*** (0.130) 0.841*** (0.210) -0.220 (0.234) 0.633*** (0.083) 1.431*** (0.461) 2.807*** (0.466) 4.672*** (0.476) 6.502*** (0.490) YES YES YES YES -5,805.471 4,155 5,292

Note: Weighted regression, with weight equal to one for respondents with two observations and weight equal to two for respondents with one observation. Standard errors in parentheses; *** p<0.001, ** p<0.01, * p<0.05. Table 5. Analysis for Men

28

Life Satisfaction Child death Physical attack/assault Illness/Accident Illness/Accident (others) Physical abuse (<18) Openness Conscientiousness Extraversion Agreeableness Neuroticism Cynical hostility Anxiety Anger-in Anger-out Good health MS score ADL score IADL score Constant Threshold 1 Threshold 2 Threshold 3 Threshold 4 Socio-demographics Region effects Year effects Mundlak effects Log-likelihood Households Observations

(1) General -0.067 (0.053) -0.233*** (0.073) -0.025 (0.043) -0.064* (0.038) -0.193*** (0.068) 0.073 (0.129) 0.458*** (0.164) 1.071*** (0.137) 0.075 (0.159) -0.618*** (0.125) -0.377*** (0.089) -1.241*** (0.111) -0.502*** (0.083) -0.479*** (0.120) 0.292*** (0.040) 0.376*** (0.099) 0.235 (0.155) -0.044 (0.188) 0.860*** (0.053) -0.876** (0.414) 0.203 (0.414) 1.795*** (0.415) 3.648*** (0.417) YES YES YES YES -8,959.102 7,706 3,853

(2) House 0.052 (0.056) -0.156* (0.080) -0.092** (0.047) -0.195*** (0.041) -0.155** (0.072) -0.152 (0.139) 0.872*** (0.176) 0.872*** (0.146) 0.407** (0.169) -0.285** (0.139) -0.280*** (0.097) -0.875*** (0.121) -0.332*** (0.091) -0.334** (0.131) 0.209*** (0.044) 0.118 (0.109) 0.138 (0.162) -0.504** (0.198) 1.109*** (0.095) -1.090** (0.430) -0.048 (0.427) 1.428*** (0.429) 3.028*** (0.434) YES YES YES YES -8,233.905 7,694 5,925

(3) Leisure -0.105* (0.057) -0.272*** (0.082) -0.039 (0.048) -0.021 (0.042) -0.113 (0.074) -0.131 (0.142) 0.446** (0.180) 0.980*** (0.149) 0.651*** (0.171) -0.099 (0.142) -0.594*** (0.099) -0.842*** (0.124) -0.027 (0.093) -0.576*** (0.133) 0.134*** (0.045) 0.076 (0.111) -0.065 (0.167) -0.153 (0.200) 1.206*** (0.101) -1.461*** (0.441) -0.370 (0.437) 1.137*** (0.438) 2.821*** (0.442) YES YES YES YES -8,238.262 7,682 5,912

(4) Family life 0.025 (0.052) -0.246*** (0.077) 0.001 (0.044) -0.112*** (0.039) -0.164** (0.068) 0.276** (0.131) 0.619*** (0.167) 1.930*** (0.143) 0.161 (0.160) -0.597*** (0.132) -0.380*** (0.092) -1.355*** (0.117) -0.631*** (0.087) -0.381*** (0.124) 0.316*** (0.042) 0.328*** (0.103) 0.628*** (0.155) 0.131 (0.187) 0.936*** (0.083) -0.128 (0.405) 1.274*** (0.405) 3.185*** (0.411) 4.868*** (0.419) YES YES YES YES -8,351.393 7,674 5,896

(5) City 0.032 (0.054) -0.411*** (0.079) 0.037 (0.046) -0.166*** (0.040) -0.339*** (0.070) -0.155 (0.136) 0.289* (0.172) 1.284*** (0.143) 0.857*** (0.164) -0.816*** (0.137) -0.514*** (0.095) -1.343*** (0.119) -0.701*** (0.090) -0.564*** (0.127) 0.179*** (0.043) -0.048 (0.106) 0.316** (0.159) -0.421** (0.193) 1.050*** (0.091) -1.773*** (0.420) -0.437 (0.415) 1.185*** (0.416) 2.899*** (0.420) YES YES YES YES -8,426.452 7,674 5,908

(6) Finances 0.025 (0.035) -0.113** (0.052) -0.047 (0.030) -0.036 (0.027) 0.025 (0.046) -0.154* (0.087) 0.121 (0.114) 0.355*** (0.093) -0.073 (0.110) -0.017 (0.090) -0.077 (0.062) -0.435*** (0.080) -0.154*** (0.059) -0.190** (0.083) 0.107*** (0.029) 0.024 (0.070) 0.040 (0.113) -0.034 (0.143) 0.000 (0.000) -0.200 (0.270) 0.620** (0.270) 1.655*** (0.270) 2.430*** (0.271) YES YES YES YES -10,994.842 7,704 3,879

(7) Health 0.065 (0.050) -0.212*** (0.075) -0.346*** (0.043) -0.111*** (0.038) -0.151** (0.066) 0.300** (0.125) 0.466*** (0.160) 1.230*** (0.134) -0.213 (0.154) -0.157 (0.126) -0.070 (0.088) -1.698*** (0.116) -0.221*** (0.082) -0.229* (0.119) 1.255*** (0.045) 1.816*** (0.106) 0.914*** (0.154) -0.190 (0.183) 0.785*** (0.076) 1.255*** (0.390) 2.753*** (0.395) 4.834*** (0.405) 6.681*** (0.415) YES YES YES YES -8,287.857 7,688 5,914

Note: Weighted regression, with weight equal to one for respondents with two observations and weight equal to two for respondents with one observation. Standard errors in parentheses; *** p<0.001, ** p<0.01, * p<0.05. Table 6. Analysis for Women

29

On the whole, then, this evidence provides some support to our Hypothesis 4, because it suggests that women are more affected than men by traumatic events – especially when the events involved a close relative rather than them personally. The evidence from Tables 5 and 6, combined with the fact that physical attacks or assaults, illnesses or accidents to others and physical abuses before age 18 are more frequently reported among women,22 suggests that it is extremely important to address an issue such as the connection between life satisfaction and traumatic events by devoting special attention to the gender dimension.

5. Conclusion Our analysis examined the relationship between a set of extremely adverse life events out of individuals’ control and their current levels of subjective well-being. General life satisfaction is lower after the occurrence of three traumatic events such as life-threatening illness or accident that hit a close relative (partner or child) and, in particular, after a serious physical attack or assault that hit the respondent and when the respondent was physically abused by either parent in her childhood or adolescence. We also assess life satisfaction through several domain-specific measures and document that the three traumatic events that turn out to be more strongly associated with general life satisfaction also matter for several aspects of life. The two major events displaying more correlations with domain-specific life satisfaction are physical attack/assault and illness/accident to a partner/child. As far as the timing of the events is concerned, our overall analysis interestingly reveals that two traumatic events such as physical attack/assault and physical abuse before 18 years old persistently contribute to explain general life satisfaction, even after many years. The correlation between life satisfaction and a major event instead decays for illness/accident of a 22

In our sample physical attacks/assaults, illnesses/accidents occurred to others and physical abuses before age 18 are reported respectively by 6.3, 31.4 and 7.8% of the women vs. 5.4, 24.6 and 6.2% of the men. For each event the difference by gender is always highly significant according to a two-sample proportion test (physical attack/assault: -2.146, p-value 0.032. Illness/Accident to others: -8.42, p-value <0.01. Physical abuse before age 18: -3.38, p-value <0.01).

30

close relative. Therefore, on the whole, our findings corroborate the idea that, for some events, individuals do not fully recover from adverse experiences: these events can leave a scar that time cannot heal on general life satisfaction.23 Our analysis also suggests that the connection between life satisfaction and traumatic events is mainly gender-specific, with women being more involved than men. Therefore, we believe that an important implication that can be drawn from our empirical findings is that, since the effects of some events are persistent over time and mostly related to women, society as a whole should be seriously concerned also about the hedonic effects generated by violence against children and women. The major limitations of our study can be summarized as follows. First, even though we examine a set of traumatic events out of individuals’ control and show that our key findings importantly hold once we control for respondents’ socio-economic conditions, personality traits and functional limitations, it is fair to acknowledge that the available data (based on self-assessed measures and with a limited panel dimension) do not lend themselves to a specific analysis of causality, preventing us from shedding light on the channels underlying the links that we detect between some extremely adverse events and life satisfaction. Second, and relatedly, unlike some of the economics and psychology studies we reviewed in Section 2, our dataset prevents us from drawing specific implications with regard to the well-known “hedonic adaptation” issue, as we have no information about individuals’ life satisfaction levels before the traumatic events occurred. Third, our study does not allow us to dig deeper into the specific characteristics of the psychological processes underlying the detected relationships between major life events and life satisfaction. We claim that our work might be extended in several directions. Do traumatic events such as the ones that we analyzed in this study also matter for other relevant variables – in addition to life satisfaction – such as time preferences and social preferences? It is plausible

23

These results differ from the ones illustrated in Misheva’s (2016) aforementioned study, where the detected effects of traumatic events such as being assaulted, being raped and being involved in an accident turn out not to be persistent over time: we speculatively argue that this is mainly due to the fact that she exclusively examines well-being in terms of emotional well-being, rather than by looking at the classic life satisfaction question – as we do by considering both the general life satisfaction measure as well as domain-specific ones.

31

that both one’s orientation towards the future and her social concerns are affected by some traumatic events passed through in her life. Next, future research might seek to understand whether not only extremely adverse personal events but also a set of extremely positive experiences (including events occurred in the distant past) do play a role in shaping one’s current level of general as well as domain-specific life satisfaction, controlling for the key dimensions taken into account in the analysis presented in this study. These extensions are left as interesting avenues for empirical research on the determinants of subjective wellbeing.

References Andreß, H., & Brockel, M. (2007). Income and life satisfaction after marital disruption in Germany. Journal of Marriage and Family, 69, 500-512. Anusic, I., Yap, S.C.Y, & Lucas, R.E. (2014). Testing set-point theory in a Swiss national sample: reaction and adaptation to major life events. Social Indicators Research, 119(3), 1265-1288. Bardo, A.R., & Yamashita, T. (2014). Validity of domain satisfaction across cohorts in the US. Social Indicators Research, 117(2), 367-385. Bartels, M. (2015). Genetics of wellbeing and its components satisfaction with life, happiness, and quality of life: A review and meta-analysis of heritability studies. Behavior Genetics, 45(2), 137-156. Bellis, M.A., Hughes, K., Jones, A., Perkins, C., & McHale, P. (2013). Childhood happiness and violence: a retrospective study of their impacts on adult well-being. BMJ Open, 3:e003427. Bernile, G., Bhagwat, V., & Rau, P.R. (2017). What doesn’t kill you will only make you more risk-loving: early-life disasters and CEO behavior. Journal of Finance, 72(1), 167206. Blanchflower, D.G., & Oswald, A.J. (2004). Well-being over time in Britain and the USA. Journal of Public Economics, 88, 1359-1386.

32

Bond, T.N., & Lang, K. (2019). The sad truth about happiness scales. Journal of Political Economy, 127(4), 1629-1640. Bonsang, E., & Klein, T.J. (2012). Retirement and subjective well-being. Journal of Economic Behavior & Organization, 83, 311-329. Bruni, L., & Stanca, L. (2008). Watching alone: Relational goods, television and happiness. Journal of Economic Behavior & Organization, 65(3-4), 506-528. Bucciol, A., & Zarri, L. (2015). The shadow of the past. Journal of Economic Psychology, 48, 1-16. Bunce, S.C., Larson, R.J., & Peterson, C. (1995). Life after trauma: Personality and daily life experiences of traumatized people. Journal of Personality, 63(2), 165-188. Calvo, R., Arcaya, M., & Baum, C.F. (2015). Happily ever after? Pre-and-post disaster determinants of happiness among survivors of Hurricane Katrina. Journal of Happiness Studies, 16(2), 427-442. Carmil, D., & Breznitz, S. (1991). Personal trauma and world view – Are extremely stressful experiences related to political attitudes, religious beliefs, and future orientation? Journal of Traumatic Stress, 4(3), 393-405. Chen, L.Y., Oparina, E., Powdthavee, N., & Srisuma, S. (2019). Have econometric analyses of happiness data been futile? A Simple truth about happiness scales. IZA Discussion Paper No. 12152. Clark, A.E., & Oswald, A.J. (1994). Unhappiness and unemployment. Economic Journal, 104 (426), 648-659. Clark, A.E., Doyle, O., & Stancanelli, E. The impact of terrorism on well-being. Evidence from the Boston Marathon bombing. Economic Journal, forthcoming. Cohen, M.A. (2008). The effect of crime on life satisfaction. Journal of Legal Studies, 37(52), S325-S353. Cornaglia, F., Feldman, N.E., & Leigh, A. (2014). Crime and mental wellbeing. Journal of Human Resources, 49(1), 110-140. Costa, P.T., & McCrae, R.R. (1992). Revised NEO Personality Inventory (NEO-PI-R) and NEO Five Factor Inventory (NEO-FFI) professional manual. Psychological Assessment Resource: Odessa, FL.

33

Diener, E. (2006). Guidelines for national indicators of subjective well-being and ill-being. Applied Research in Quality of Life, 1, 151-157. Edwards, V.J., Holden, G.W., Felitti, V.J., & Anda, R.F. (2003). Relationship between multiple forms of childhood maltreatment and adult mental health in community respondents: results from the adverse childhood experiences study. American Journal of Psychiatry, 160(8), 1453-1460. Easterlin, R.A., & Sawangfa, O. (2009). Happiness and domain satisfaction: New directions for the economics of happiness, in Dutt A.K., & Radcliff B. (eds.), Happiness, Economics and Politics: Towards a Multi-Disciplinary Approach, Edward Elgar, 70-94. Di Tella, R., & MacCulloch, R. (2005). Partisan social happiness. Review of Economic Studies, 72(2), 367-393. Di Tella, R., & MacCulloch, R. (2006). Some uses of happiness data in economics. Journal of Economic Perspectives, 20(1), 25-46. Eckel, C.C., & Grossman, P.J. (1998). Are women less selfish than men? Evidence from dictator experiments. Economic Journal, 108(448), 726-735. Fonda, N., & Herzog, T. (2004). Documentation of physical functioning measured in the Health and Retirement Study and the Asset and Health Dynamics among the oldest old study: Documentation report. The HRS Health Working Group. Frazier, P., Conlon, A., & Glaser, T. (2001). Positive and negative life changes following sexual assault. Journal of Consulting and Clinical Psychology, 69(6), 1048-1055. Frijters, P., Johnston, D.W., & Shields, M.A. (2011). Life satisfaction dynamics with quarterly life events data. Scandinavian Journal of Economics, 113(1), 190-211. Georgellis, Y., Lange, T., & Tabvuma, V. (2012). The impact of life events on job satisfaction, Journal of Vocational Behavior, 80(2), 464-473. Graham, C., & Chattopadhyay, S. (2013). Gender and well-being around the world. International Journal of Happiness and Development, 1(2), 212-232. Hanslmaier, M. (2013). Crime, fear and subjective well-being: How victimization and street crime affect fear and life satisfaction. European Journal of Criminology, 10(5), 515-533. Havari, E., & Mazzonna, F. (2011). Can we trust older people’s statements on their childhood circumstances? Evidence from Sharelife. SHARE Working Paper Series 05-2011.

34

Hayes, N., & Joseph, S. (2003). Big 5 correlates of three measures of subjective well-being. Personality and Individual Differences, 34, 723-727. Hentschel, S., Eid, M., & Kutscher, T. (2017). The influence of major life events and personality traits on the stability of affective well-being. Journal of Happiness Studies, 18(3), 719-741. Holman, E.A., & Silver, R.C. (1998). Getting ‘stuck’ in the past: Temporal orientation and coping with trauma. Journal of Personality and Social Psychology, 74, 1146-1163. Kahneman, D., Wakker, P.P., & Sarin, R. (1997). Back to Bentham? Explorations of experienced utility. Quarterly Journal of Economics, 112, 375-405. Lantz, P.M., House J.S., Mero R.P., & Williams, D.R. (2005). Stress, life events, and socioeconomic disparities in health: Results from the Americans’ Changing Lives Study. Journal of Health and Social Behavior, 46(3), 274-288. Lubian, D., & Zarri, L. (2011). Happiness and tax morale: An empirical analysis. Journal of Economic Behavior & Organization, 80, 223-243. Lucas, R.E., Clark, A.E., Georgellis, Y., & Diener, E. (2003). Reexamining adaptation and the set-point model of happiness: Reactions to changes in marital status. Journal of Personality and Social Psychology, 84(3), 527-539. Lucas, R.E. (2007a). Adaptation and the set-point model of subjective well-being. Does happiness change after major life events? Current Directions in Psychological Science, 16(2), 75-79. Lucas, R.E. (2007b). Long-term disability is associated with lasting changes in subjective well-being: Evidence from two nationally representative longitudinal studies. Journal of Personality and Social Psychology, 92(4), 717-730. Luechinger, S., & Raschky, P.A. (2009). Valuing flood disasters using the life satisfaction approach. Journal of Public Economics, 93, 620-633. Luhmann, M., & Eid, M. (2009). Does it really feel the same? Changes in life satisfaction following repeated life events. Journal of Personality and Social Psychology, 97(2), 363381. Luhmann, M., Hofmann, W., Eid, M., & Lucas, R.E. (2012). Subjective well-being and adaptation to life events: A meta-analysis. Journal of Personality and Social Psychology,

35

102(3), 592-615. Luttmer, E.F.P. (2005). Neighbors as negatives: Relative earnings and well-being. Quarterly Journal of Economics, 120(3), 963-1002. Mahuteau, S., & Zhu, R. (2016). Crime victimization and subjective well-being: panel evidence from Australia. Health Economics, 25, 1448-1463. Metcalfe, R., Powdthavee, N., & Dolan, P. (2011). Destruction and distress: Using a quasiexperiment to show the effects of the September 11 on mental well-being in the United Kingdom. Economic Journal, 121, F81-F103. Misheva, V. (2016). What determines emotional well-being? The role of adverse experiences: Evidence using twin data. Journal of Happiness Studies, 17, 1921-1937. Moor, N., & de Graaf, P. (2016). Temporary and long-term consequences of bereavement on happiness. Journal of Happiness Studies, 17, 913-936. Mundlak, Y. (1978). On the pooling of time series and cross section data. Econometrica, 46(1), 69-85. Oswald, A.J., & Powdthavee, N. (2008). Does happiness adapt? A longitudinal study of disability with implications for economists and judges. Journal of Public Economics, 92, 1061-1077. Oswald, A.J., Proto, E., & Sgroi, D. (2015). Happiness and productivity. Journal of Labor Economics, 33(4), 789-822. Powdthavee, N. (2009). What happens to people before and after disability? Focusing effects, lead effects, and adaptation in different areas of life, Social Science and Medicine, 69, 1834-1844. Rojas, M. (2006). Life satisfaction and satisfaction in domains of life: is it a simple relationship?, Journal of Happiness Studies, 7, 467-497. Smith, J.P., & Smith, G.C. (2010). Long-term economic costs of psychological problems during childhood. Social Science & Medicine, 71(1), 110-115. Smith, J., Fisher, G., Ryan, L., Clarke, P., House, J., & Weir, D.R. (2013). Psychosocial and lifestyle questionnaire 2006-2010: Documentation report. The HRS Psychosocial Working Group. Stanca, L. (2012). Suffer the little children: Measuring the effects of parenthood on well-

36

being worldwide. Journal of Economic Behavior & Organization, 81(3), 742-750. Staubli, S., Killias, M., Frey, B.S. (2013). Happiness and victimization: An empirical study for Switzerland. European Journal of Criminology, 11(1), 57-72. Tedeschi, R.G., & Calhoun, L.G. (1996). The posttraumatic growth inventory: Measuring the positive legacy of trauma. Journal of Traumatic Stress, 9(3), 455-471. Van Praag, B.M.S., Frijters, P., & Ferrer-i-Carbonell, A. (2003). The anatomy of subjective well-being. Journal of Economic Behavior & Organization, 51, 29-49. Van Praag, B.M.S., Romanov, D., & Ferrer-i-Carbonell, A. (2010). Happiness and financial satisfaction in Israel: Effects of religiosity, ethnicity, and war. Journal of Economic Psychology, 31, 1008-1020. Verme, P. (2009). Happiness, freedom and control. Journal of Economic Behavior & Organization, 71, 146-161. Weiss, A., Bates, T.C., & Luciano, M. (2008). Happiness is a personal(ity) thing. The genetics of personality and well-being in a representative sample. Psychological Science, 19(3), 205-210. Yap, S.C.Y., Anusic, I., Lucas, R.E. (2012). Does personality moderate reaction and adaptation to major life events? Evidence from the British Household Panel Survey. Journal of Research in Personality, 46, 477-488. Yehuda, R. (2002). Post-traumatic stress disorder. New England Journal of Medicine, 346(2), 108-114.

37



We investigate the effects of exposure to traumatic life experiences on subjective well-being.



We employ large-scale survey data from four waves of the US Health and Retirement Study (HRS).



Our findings reveal that both general and domain-specific life satisfaction are negatively related to some extremely adverse events.



We provide evidence that the effects of some traumatic events are persistent over time and mostly related to women.



Surprisingly, the effects of child death are negligible also in the short term.

38