Childhood psychological distress and youth unemployment: Evidence from two British cohort studies

(2015) - Mark Egan, Michael Daly, Liam Delaney

Journal: Social Science & Medicine
Link:: https://linkinghub.elsevier.com/retrieve/pii/S0277953614007461
DOI:: 10.1016/j.socscimed.2014.11.023
Links::
Tags:: #paper #NCDS #Unemployment #LYSPE
Cite Key:: [@eganChildhoodPsychologicalDistress2015]

Abstract

The effect of childhood mental health on later unemployment has not yet been established. In this article we assess whether childhood psychological distress places young people at high risk of subsequent unemployment and whether the presence of economic recession strengthens this relationship. This study was based on 19,217 individuals drawn from two nationally-representative British prospective cohort studies; the Longitudinal Study of Young People in England (LSYPE) and the National Child Development Study (NCDS). Both cohorts contain rich contemporaneous information detailing the participants' early life socioeconomic background, household characteristics, and physical health. In adjusted analyses in the LSYPE sample (N ¼ 10,232) those who reported high levels of distress at age 14 were 2 percentage points more likely than those with low distress to be unemployed between ages 16 and 21. In adjusted analyses of the NCDS sample (N ¼ 8985) children rated as having high distress levels by their teachers at age 7 and 11 were 3 percentage points more likely than those with low distress to be unemployed between ages 16 and 23. Our examination of the 1980 UK recession in the NCDS cohort found the difference in average unemployment level between those with high versus low distress rose from 2.6 pct points in the pre-recession period to 3.9 points in the post-recession period. These findings point to a previously neglected contribution of childhood mental health to youth unemployment, which may be particularly pronounced during times of economic recession. Our findings also suggest a further economic benefit to enhancing the provision of mental health services early in life.

Notes

“n adjusted analyses of the NCDS sample (N ¼ 8985) children rated as having high distress levels by their teachers at age 7 and 11 were 3 percentage points more likely than those with low distress to be unemployed between ages 16 and 23.” (Egan et al., 2015, p. 11)

“examination of the 1980 UK recession in the NCDS cohort found the difference in average unemployment level between those with high versus low distress rose from 2.6 pct points in the pre-recession period to 3.9 points in the post-recession period.” (Egan et al., 2015, p. 11)

“Understanding how childhood mental health can shape economic outcomes over the lifespan is a key question that cuts across health, education, and employment policy. Depression and mental health problems in childhood have been shown to detrimentally affect family income (Goodman et al., 2011; Smith and Smith, 2010), labor supply (Goodman et al., 2011), educational attainment (Fletcher, 2008; Cornaglia et al., 2012), and earnings in adulthood (Fletcher, 2013).” (Egan et al., 2015, p. 11)

“Standard errors were clustered by individual to account for repeated observations of the same person and we estimated marginal effects after the Probit regression to calculate percentage point changes in unemployment probability (Long and Freese, 2014).” (Egan et al., 2015, p. 12) SUPER important. Need to remember to cluster standard errors if using the employment status varaible.

“A negative binomial model is appropriate for over-dispersed count data: in both Study 1 and 2 there is significant clustering at zero months of accumulated unemployment and the mean number of unemployed months is much lower than the variance (see Sturman, 1999 for the merits of using the negative binomial model versus other models when analyzing count data).” (Egan et al., 2015, p. 12)

“We used employment data from wave 4, which elicited responses from 68 per cent of all-time participants at age 23. After matching childhood data from birth, age 7, and age 11 waves with employment data gathered at age 23 and deleting observations which did not contain data for the main covariates, we constructed a sample of 597,858 monthly unemployment status observations for 8985 cohort members.” (Egan et al., 2015, p. 13)

“Hawkes and Plewis (2006) show that those who left the survey did not differ significantly from the rest of the sample on the basis of observable socioeconomic characteristics.” (Egan et al., 2015, p. 13)

“At ages 7 and 11 the cohort members were assigned a ‘depression’ score by their teacher based on questions from the depression subscale of the British Social Adjustment Guide” (Egan et al., 2015, p. 14)

“We constructed a measure of childhood distress by averaging age 7 and 11 scores on this measure” (Egan et al., 2015, p. 14)

“When the cohort members were aged 24 they were asked to recall their monthly employment history from age 16 to 23, spanning the period June 1974 to February 1982” (Egan et al., 2015, p. 14)

“used these data to generate two outcome variables: (i) a variable tracking monthly employment status for the 1974e82 period where being in fulltime employment is coded as 0 and being unemployed is coded as 1 and (ii) a continuous variable for summed total months of youth unemployment.” (Egan et al., 2015, p. 14)

“Given that lower childhood cognitive ability is associated with higher adult distress (Gale et al., 2009) and unemployment (Caspi et al., 1998), this may reduce potential confounding of the main effect of distress” (Egan et al., 2015, p. 14)

“Lastly, we specified a Probit model (Model 5) to conceptualize the interaction of distress and the onset of the 1980s UK recession, dated as starting in January 1980 (Jenkins, 2010). We created a binary variable for the recession where 0 ¼ June 1974eDecember 1979 (participants aged 16e21) and 1 ¼ January 1980eFebruary 1982 (participants aged 21e23). We then interacted this variable with the distress measure to determine whether the more distressed were more likely to become unemployed after the recession began” (Egan et al., 2015, p. 14) Fascinating idea. Certainly something to replicate in any models looking at labour market stability/earnings.

“Five months at the beginning and end of the data-range are omitted due to small sample sizes” (Egan et al., 2015, p. 15)

“Secondly, while we use a well-validated measure of childhood distress in Study 1, such a measure was not available in Study 2.” (Egan et al., 2015, p. 16)

“Lastly, although we found that the highly distressed were more likely to become unemployed after the 1980s recession, the generalizability of this finding is unclear e this assocation may be specific to the time period, the country or the nature of that particular recession.” (Egan et al., 2015, p. 16)