Individual-level data, as used in this study, are drawn from the Survey of Health, Ageing and Retirement in Europe (SHARE). SHARE is a multidisciplinary, longitudinal survey on ageing which focuses on individuals aged 50+ and their spouses.Footnote 2 The survey contains both regular and retrospective waves (SHARELIFE). The regular rounds collect information on an individual’s current situation such as health, employment, social network/relations, accommodation, economic situation/assets, behavioural risks and expectations. In addition, two survey rounds add retrospective information on multiple dimensions of the respondent’s past (health, healthcare, accommodation, career, household situation and performance at school during childhood, number of children, childbearing for women, emotional experiences in early life, relationship with parents, adverse childhood experiences, etc.).
What makes SHARE data particularly suited for the purposes of our analysis is the ability to link the information on the current situation of respondents to retrospective childhood/adulthood data. The information on lifestyles such as smoking behaviour over the lifespan, alcohol abuse and obesity in adulthood, as well as personal characteristics (age, gender, and education) are taken from regular waves, while the retrospective childhood conditions, the respondent’s household situation and recently released data on the quality of the parent-child relationship and early-life emotional experiences are drawn from SHARELIFE. The final sample includes all respondents participating in at least one regular SHARE wave (between Waves 4 to 6) and in the SHARELIFE interview of Wave 7. Individuals who entered the survey before Wave 4 are excluded because of the lack of information about their adverse early life experiences. We obtain a data set covering18 European countries (Austria, Germany, Sweden, Spain, Italy, France, Denmark, Greece, Switzerland, Belgium, Czech Republic, Poland, Luxembourg, Hungary, Portugal, Slovenia, Estonia, and Croatia) and Israel.
In order to take the prevailing culture into account, we follow the results of the GLOBE study (Global Leadership and Organisational Behaviour Effectiveness 2004, and 2007) and Mensah and Chen’s (2013) subsequent extension (Global Clustering of Countries by Culture),Footnote 3 and group the countries in our sample into four clusters, namely: Nordic countries (Sweden, Denmark, and Estonia), Germanic countries (Austria, Germany, Switzerland, Belgium, and Luxembourg), Latin countries (Italy, Spain, Portugal, France, and Israel), and Eastern countries (Croatia, Greece, Poland, the Czech Republic, Hungary, and Slovenia). We chose to refer to this type of culture-based clustering since a number of studies state that cultural features have an important influence on parent-child interactions (Coltrane, 2004, Morman & Floyd, 2002, Saracho & Spodek, 2008). After correcting for missing values, the final sample includes 26.877 observations, which are split up as follows: 5024 in Nordic countries, 8033 in Germanic countries, 7031 in Latin countries, and 6789 in Eastern European countries.
Outcome variables
In our analysis, we explore the relationship between adverse childhood conditions and a set of unhealthy behaviours, such as smoking, drinking, excess weight and obesity over the lifespan.
In evaluating smoking behaviour, we use information elicited from regular SHARE waves, considering two variables. In order to evaluate the impact that ACE may have on the probability of starting to smoke, a dummy is used to indicate whether the respondent has ever smoked on a daily basis at any time. For individuals who say they are current smokers or have smoked on a daily basis, we consider a variable that records the number of years of smoking. About 44% of respondents in our sample say they smoked on a daily basis at some stage in their life. The percentage of men is nearly 57%, while for women it is about 34%. If we focus on the persistence of smoking in terms of the number of years as a smoker, men tend to smoke for longer periods (on average 27 years) than women (23 years). Looking at macro-regions, the highest percentage of individuals reporting smoking on a daily basis is in Nordic countries (48%), followed by Germanic (46%), Eastern European (42%), and Latin countries (41%). These outcomes are unconditional and may depend on age and cohort, still the differences are quite remarkable: the econometric analysis below is an attempt to unravel the role of the different variables.
Regarding alcohol abuse, a dummy variable is used starting from the intensity and the frequency with which respondents drink alcoholic beverages in adulthood. Specifically, we consider the following question (available in the regular SHARE waves): “In the last three months, how often did you have six or more units of alcoholic beverages on one occasion? 1. Daily or almost daily; 2. Five or six days a week; 3. Three or four days a week; 4. Once or twice a week; 5. Once or twice a month; 6. Less than once a month; 7. Not at all in the last 3 months”. The heavy drinking dummy has value 1 if respondents declare a consumption of six or more drinks on the same occasion (i) daily or almost daily; (ii) five or six days a week; (iii) three or four days a week; (iv) once or twice a week, and 0 otherwise. About 12% of the respondents in our sample can be considered heavy drinkers according to the above definition. This proportion differs between men and women: rates of self-reported heavy drinking are about 18.4% for men and 6.3% for women. In terms of regional disparities, Germanic countries have the largest percentage of heavy drinkers (about 17%), while Latin countries have the lowest (6%).
We measure adult excess weight and obesity using information on the body mass index (BMI) elicited in the regular waves of SHARE. BMI is calculated as body weight in kilograms divided by the square of body height in metres (kg/m2). In line with the medical indications that BMI-for-age should only be used for children and teenagers (below the age of 20) and with the recent literature (e.g., Feigl et al., 2019; Otang-Mbeng et al., 2017; Devaux & Sassi, 2015), we use a single threshold for defining the overweight and one cut-off for obesity for the individuals in our sample (https://www.cdc.gov/healthyweight/assessing/bmi/adult_bmi/).Footnote 4 In line with the World Health Organisation definition of excess weight and obesity, we consider that an individual is obese with a BMI of 30 or more while a person is overweight with a BMI of 25 or more. In order to evaluate the impact that ACE may have on the probability to be overweight or obese later in life, we first use a dummy with value 1 when the respondent has a BMI of 25 or more, and 0 otherwise. The overweight and obese account for 65.68% of our sample. More men than women are overweight or obese (70.3% versus 62.19%). These percentages reflect recent European statistics,Footnote 5 confirming a high prevalence of overweight people and obesity especially in the adult population. Second, we focus on the most severe form of being overweight, i.e. obesity, by creating a dummy variable that has value 1 for an individual with a BMI of 30 or more and 0 otherwise. About 24.5% of the overall sample risks obesity, and this percentage slightly differs between genders (men 23.47%; women 25.41%). There are significant differences in the percentages of obese people between macro-regions; interestingly, Eastern Europe has the highest percentage of obese people (27%), followed by about 23% in Nordic countries, 21% in Germanic countries, and about 18% in Latin countries.
Adverse childhood experiences
Our key explanatory variables are related to adverse early-life experiences. SHARELIFE asks respondents to provide information on exposure to child neglect and childhood physical abuse, from mother, father or a third party. In relation to physical abuse from the mother or father, the questionnaire asks:
-
1.
How often did your mother/your father push, grab, shove, throw something at you, slap or hit you? 1. Often 2. Sometimes 3. Rarely 4. Never
In addition, the survey also collects data on child physical abuse by third parties:
-
2.
How often did anybody else physically harm you in any way? 1. Often 2. Sometimes 3. Rarely 4. Never.
Albeit different from the items used in epidemiological research, we consider an additional indicator for child neglect derived from the following question:
-
3.
How much did your mother/your father (or the woman/man that raised you) understand your problems and worries? 1. A lot 2. Some 3. A little 4. Not at all
Finally, we also include among the explanatory variables the self-reported quality of the relationship with each parent:
-
4.
How would you rate the relationship with your mother/your father (or the woman/man that raised you)? 1. Excellent 2. Very good 3. Good 4. Fair 5. Poor.
The literature in the field distinguishes between various subtypes of neglect based on the dimension in which parents prove to be inadequate. As regards emotional neglect, Straus et al. (1997), for instance, associate emotional neglect with parental failure to provide “affection, companionship, and support”. They point out that this form of early adverse experience may have important social and psychological implications, which may even be more damaging than some types of psychologically “abusive” attention (such as hostile and verbally abusive parents). However, there is a large heterogeneity in measuring child neglect in surveys (see Stoltenborgh et al. (2013) for a comprehensive overview). With data from the Recruitment Assessment Program study, Young et al. (2006) use one item to assess emotional neglect (“You felt loved”), while Straus et al. (1997) use a short version of “The Neglect Scale” and approximate emotional neglect from the scores to the following statement: “did not help me when I had problems”. Finally, the refined CDC-Kaiser Permanent ACE Study asks a series of questions in the sphere of emotional neglect, such as “You knew there was someone to take care of you and protect you?”, “Your family was a source of strength and support?”, “People in your family looked out for each other?”, “There was someone in your family who helped you feel important or special?” and “You felt loved?”. All these measures have been validated as good instruments to measure emotional neglect. Although the above-reported questions on emotional neglect in SHARE (lack of understanding or poor relations) slightly differ from the questions in other ACE-specific studies, we believe that, from a conceptual point of view, they are informative and aligned with the existing proxies for the presence or absence of emotional neglect.
In order to obtain a set of adverse childhood experience variables, we first recode the answers into dichotomous variables, where a value of 1 indicates that the individual was exposed to a negative experience in early life. We consider that an individual experienced physical abuse from either the mother or the father if she/he answers ‘1. Often’ or ‘2. Sometimes’ to question 1. We treat question 2 in the same manner to capture physical harm from another person. A situation of ‘child neglect’ is shown by answers ‘3. A little’ or ‘4. Not at all’ to question 3. The relationship with the mother/father in childhood is rated 1, i.e., problematic/negative, if the respondent answers ‘4. Fair’ or ‘5. Poor’ to the last question.
We then create a set of dummy indicators with value 1 if respondents have experienced (i) physical harm from father/mother/other parties, (ii) child neglect from either the mother or the father, (iii) a poor relationship with either parent, and 0 otherwise.