Experiencing Life Events and Childhood Subjective Wellbeing: A Longitudinal Analysis of Growing Up in Scotland

Methodology

Data overview

Data from Sweeps 7 to 10, inclusive, of the GUS study was used for this analysis, as the children themselves are asked about their own life satisfaction from Sweep 7 (aged 7-8) onwards. The sample consisted of 4,069 respondents, 50.5% of which were boys and 49.5% were girls. The sample included those who were present in the data in at least two sweeps, with some being present at two, three and four of the four sweeps captured by this analysis. This decision not to restrict the analysis to only those present in all four sweeps was made to ensure that enough participants were present in the data to provide a meaningful analysis. Restricting the analysis to only those present at all four sweeps would have resulted in a sample size which was too small to draw significant results. Additionally, the most recent data available was used in order to capture the most up-to-date picture of children's wellbeing in Scotland, informing the decision to use the latest sweeps of GUS. Information about when GUS sweeps took place can be found in the data documentation for each sweep ^[2]and on the study website^[3].

Table 1 provides an overview of the sample, showing the number of participants which were present in each sweep along with the age of the participants when each sweep of data was collected. The table also displays the gender composition at each of the four sweeps, showing that the sample was split evenly between boys and girls.

Life Events Variables

Parental Separation

Parental separation was measured through an indicator of a change in family type at some point across the four sweeps. This variable was created from a question to the main carer about whether the family type was 'lone parent' or 'couple family'. Those who previously belonged to a 'couple family' and were later reported to belong to a 'lone parent family' were categorised as becoming a lone parent family.Those who previously belonged to a 'lone parent family' and were later reported to belong to a 'couple family' were categorised as becoming a couple family. Those who remained a lone parent family throughout and those who remained a couple family throughout were both categorised as having no change in family type.

Table 2 shows the frequency distribution for the 'family type change variable'. The majority of the sample (86.1%) experienced no change in family type at any point across the four sweeps, however 320 participants (9%) experienced becoming a lone parent family indicating that they experienced parental separation. A further 175 participants (5%) experienced becoming a couple family.

Bereavement

Main carers were asked about whether their child had experienced the loss of a parent, sibling or grandparent since the previous sweep. Due to the low number of participants who experienced the loss of a parent or sibling, these indicators were combined to form a 'loss of a parent or sibling' variable. As the loss of a grandparent is typically more likely to occur between the ages of 7 and 15 than that of a parent or sibling, and grandparents are often less likely to be co-resident, being bereaved of a grandparent was treated separately.

Table 3.1 shows the frequency distribution for the variable indicating the loss of a grandparent. Almost 50% of participants had experienced the loss of a grandparent at some point across the four sweeps, indicating that being bereaved of a grandparent was the most common of the life events explored in this project.

Table 3.2 shows the frequency distribution of the variable indicating the loss of a parent or sibling. Consistent with what was expected, the majority of participants (96.8%) did not experience the loss of a parent or sibling. A total of 129 participants (3.2%) experienced the loss of a parent or sibling at some point across the four sweeps.

Accident or illness within the family

Main carers were also asked whether their child had experienced an accident or illness within the family. An accident or illness involving a parent or of a sibling were combined to form an indicator of an 'accident or illness of a family member. Due to small numbers experiencing each, these were combined into one variable 'family accident or illness'.

Table 4 shows the frequency distribution for this variable, indicating that 17.6% of participants experienced a family accident or illness at some point in the four sweeps. The majority of the sample (82.4%) did not experience any family accident or illness.

Participants were also asked about other other life events including experiencing having a parent in prison, family experiences of crime and drug taking or alcoholism in the family, all of which were not included in this analysis either due to a small number of participants experiencing them or a lack of consistency in their measurement over time.

Outcome Variables

Subjective wellbeing

The outcome variable used for this project was an indicator of subjective wellbeing, measured by an indicator of life satisfaction. The measure of life satisfaction was created from four quesitons asked ofthe GUS participants about their satisfaction with their lives:

• How often do you wish your life was different?
• How often do you feel that your life is just right?
• How often do you feel you have what you want in life?
• How often do you feel you have a good life?

As these questions were asked to the children directly, they are self-reported items. Responses to these questions were ordered, with the categories ranging from 'never' to 'always'. As these response options were ordered in a structured way, an average score on all four items was generated to create an 'overall life satisfaction' score. To capture changes in this variable over time, the overall satisfaction score was used to create a binary variable indicating 'low' and 'high' life satisfaction. Participants who initially had a high life satisfaction score and who had a low score in a subsequent sweep were categorised as having 'deteriorating life satisfaction'. Participants who initially had a low life satisfaction score who later had a high score were categorised as having 'improving life satisfaction'. Those who continued to have low satisfaction throughout and those who continued to have high satisfaction throughout were categorised as 'staying constant' as they did not experience a change in life satisfaction over the sweeps.

Table 5 shows the frequency distribution for the change in life satisfaction variable. Over half of the participants (52%) experienced no change in life satisfaction over time. A total of 953 participants (32.1%) experienced a deterioration in life satisfaction and 473 (15.9%) experienced an improvement.

Body Mass Index – Sensitivity outcome

Body Mass Index (BMI) is a score that adjusts a person's weight for their height. Individuals are placed into bands to show where they stand in relation to the rest of the population, in particular whether they have high or low BMI. Life adversities are a known risk factor for emotional overeating as well as restrained eating in children (Thomas et al. 2020). Only those height and weight measurements considered by the interviewer to be reliable were used to calculate the BMI scores available in the GUS data. Across the four sweeps, those who were either overweight/obese or a healthy weight initially and became underweight were categorised as 'becoming underweight'. Those who were either underweight or a healthy weight and then became overweight/obese were categorised as 'becoming overweight'. Those who remained underweight, remained overweight or remained a healthy weight were categorised as experiencing no change in BMI over time.

Table 6 shows the frequency distribution for the BMI change variable. The majority of the sample (83.1%) did not experience becoming either overweight/obese or underweight at any point in the four sweeps. A total of 83 participants (2.5%) experienced becoming underweight and 489 participants (14.5%) experienced becoming overweight/obese indicating that becoming overweight was more likely than becoming underweight.

Physical health and disability – Sensitivity outcome

Experiencing changes in physical health was explored using a measure which captured the presence of a physical health condition or disability. This variable comes from a question to the main carer asking if the child has developed a new illness or disability since the previous sweep. These included: visual impairments, hearing difficulties, mobility issues, learning difficulties, stamina or breathing difficulties, social and behavioral issues or any other impairments. Participants who were reported to have one of these at an earlier sweep and were later reported to have no illness were categorised as 'physical health improving'. Participants who were initially reported to have no conditions and were reported to have one or more at a subsequent sweep were categorised as 'physical health deteriorating'. Participants who had no condition throughout the sweeps and who had a condition continually throughout the sweeps were categorised as their physical health remained constant. This variable is referred to as 'physical health' for the remainder of the report.

Table 7 shows the frequencies for the change in physical health variable. Most participants (71.9%) did not experience a change in physical health. Only 8.9% of participants experienced an improvement in physical health whereas 19.2% experienced a deterioration, indicating that they developed or experienced an illness or disability at some point across the four sweeps.

Mental health – Sensitivity outcome

Changes in mental health were measured by changes in strengths and difficulties questionnaire (SDQ) scores^[4]. The SDQ is asked of the children and captures five aspects of mental health: emotion, conduct, hyperactivity, peer problems and prosocial behavior. Previous literature has established clinical cut-off scores, with scores above 17 traditionally being considered to be 'abnormal' (Bryant et al. 2020). Therefore this predetermined cut-off was used for this variable to differentiate between those with 'normal' and 'abnormal' functioning. Participants who initially had SDQ scores considered to be 'normal' who later had an 'abnormal' score were categorised as having deteriorating mental health. Participants who initially had SDQ scores considered to be 'abnormal' who later had a 'normal' score were categorised as having improving mental health. Those who had 'abnormal' scores throughout and those who had 'normal' scores throughout were both categorised as experiencing no change in mental health.

Table 8 shows the frequency distribution for the change in mental health variable. Most participants (79.4%) experienced no change in their mental health over time, whilst 9.6% experienced a deterioration and 10.9% experienced an improvement in mental health.

Additional Covariates

Gender

There is a lack of conclusive evidence regarding gender differences in subjective wellbeing, as some research studies have found that boys and girls show different patterns whereas others have indicated no differences (Chen et al. 2020). Studies which have found differences in subjective wellbeing between boys and girls have found that different aspects of wellbeing matter depending on gender. For example, self-awareness and self-esteem have been found to be more important for girls' wellbeing whereas achievements and feeling successful were more important for boys (Kaye-Tzadok et al. 2017). However, results are clearer in terms of mental health outcomes as large number of studies show that girls start to manifest higher rates of depression than boys between the ages of 13 and 15 years old (Ge et al. 2001).

Household socio-economic position

Two indicators of household socio-economic position were included in the analysis: household income and household education. Previous research has evidenced that household income has a positive causal effect on children's health outcomes, including cognitive and behavioural development (Lindeboom et al. 2009). Household income is also shown to indirectly influence children's outcomes through other aspects which are important for their development, such as maternal mental health, parenting and the home environment (Cooper and Stewart 2021). Household income is measured in GUS by a variable with four categories, '£7999 or less', '£8K-£14,999', '£15K-£28,999' and '£29K+'. This was used to create a binary measure of household income, indicating those who had an annual household income of over £29,000 compared with those who had less than this.

Parental education can also affect children's health as higher levels of education are associated with positive parenting styles and an increased ability to make informed health decisions (Lindeboom et al. 2009). Higher levels of education can also lead to higher earnings and these resources could be used to invest in health and to act as a buffer for the potential negative impact of adversities (Case et al. 2002). A measure of highest household education was also constructed, consisting of the categories 'degree', 'higher', and 'standard grade'.

Number of adults present in the household

A measure of the number of additional adults - other than the main carer respondent - present in the household was also included in the analysis as an indicator of the potential support available to both the young person and their main carer. Having more than one adult in the household could help to mitigate adverse effects following a significant life event. Social support is used to refer to the extent to which someone has access to, or perceives they have access to, resources provided by social network connections. A substantial body of research supports the idea that social support plays an influential role in the relation between stressful life events and children's outcomes (Jackson and Warren 2000). In particular, it is often regarded as an important protective factor for positive mental health outcomes at all ages including during both childhood and adolescence (Bauer et al. 2021).

Multinomial logistic regression modelling

Multinomial logistic regression is an extension of binary logistic regression which allows for more than two categories of the outcome variable (Starkweather and Moske 2011). This regression modelling approach allowed the two categories 'life satisfaction deteriorates' and 'life satisfaction improves' to be compared with the third category 'life satisfaction remains constant', capturing changes in subjective wellbeing in both directions. To ease interpretation of the model results, the reference category was set to 'life satisfaction remains constant', allowing this to act as the baseline to which a deterioration and an improvement could be compared. Directly comparing the outcome remaining constant with it improving and deteriorating allows us to capture not only the presence of a change in health but also the direction of the change. The categorisation of the outcome variable follows a similar method to that of Rajmil et al. (2009), who explored changes in children's mental health outcomes using strength and difficulty questionnaire (SDQ) scores. The authors measured a change in SDQ scores over time by respondents membership in one of three categories; improves, remains stable and worsens (Ramjil et al. 2009).

Sensitivity Analysis

It is acknowledged that children's wellbeing is highly subjective as it requires their own perception and evaluation which could be subject to reporting bias (Camerini and Schulz 2018). To further support the findings on children's subjective wellbeing, this project applied a sensitivity analysis by also looking at different outcome measures to explore the impact of life events on indicators of both physical and mental health to incorporate more objective measures of children's outcomes. Furthermore, modelling the health outcomes separately allows for an exploration of whether experiencing significant life events impacts upon different aspects of health in meaningfully different ways.