Growing Up in Ireland Cohort ‘98 at Age 25: Childhood Influences on Education and Finances

Purpose of Survey

Growing Up in Ireland (GUI) is the national longitudinal study of children and young people in Ireland. The study is a joint project by the Central Statistics Office (CSO) and the Department of Children, Disability and Equality (DCDE). The CSO is responsible for the GUI survey itself: designing and building the survey; collecting, processing, and analysing the data; and facilitating data access to researchers and policy makers. DCDE as the departmental sponsor has responsibility for the wider elements of the GUI study: engaging with policy and scientific stakeholders to understand data requirements; consulting with children/young people; identifying research needs, data priorities and policy objectives, and promoting the use of GUI data for research and policy development. Working closely together ensures seamless integration of the complimenting responsibilities. 

Established in 2006, the study originally followed two groups of Irish children: Cohort '98 who joined the study when they were 9-years-old, and Cohort '08 who joined the study when they were 9-months-old. The two cohorts have been surveyed at regular intervals since. Cohort '08 are aged 17/18 in the wave of data collection currently ongoing. Cohort '98 were age 25 at their last wave of data collection (Wave 5) and the results of that exercise are the subject of this report. Only those who took part in Wave 5 of the study were included in the analyses presented in this release. This release analyses data from Wave 1 and Wave 5 of the study. 

For further information on the sample frame and design see Background Notes from the Cohort ‘98 Main Results release. 

Reference Period

Data collection for Cohort '98 Wave 1 took place between August 2007 and May 2008. Data collection for Cohort '98 Wave 5 took place between April 2023 and April 2024.

Regression analysis

Logistic regression analyses were carried out using the survey package in R. Each variable was tested in a logistic regression analysis independently without confounders in the model. These were then combined into one model containing all confounders (see list of variables below) and a stepwise approach was taken whereby the least significant variable in the combined model was removed in the next model and the regression was rerun. The models were then tested using Akaike Information Criterion which compares the relative quality of different statistical models for a given set of data to find the model that best explains the data without being overly complex, balancing goodness of fit with model simplicity.  

The marginal effects for each predictor variable in the best fitting model were then calculated, accounting for all other variables in the model. The Wave 5 weights were included in these models (see Background Notes from the Cohort ‘98 main results release for more information). 

Below is the full list of variables that were examined against highest education level of the respondent at age 25 prior to the stepwise process being applied. The variables for inclusion as confounders in the model were selected to reduce the risk of overestimation of the effect of the predictor variables.  

The variables used in the final model can be found in Table A1 of the Appendix. 

• Did Primary Caregiver (PCG) use the public library for the Study Child at Wave 1 
• Did the Study Child have a TV in their bedroom at Wave 1  
• How far the PCG expected the Study Child to go in their education asked at Wave 1 
• Sex of the respondent 
• Highest level of education of the PCG at Wave 1 
• Household social class at Wave 1: Social class of PCG and Secondary Caregiver (SCG) is derived from their occupation. During the Wave 1 survey, both caregivers, where relevant, were asked to provide details on their occupation; from current or, where the respondent was economically inactive at the time of interview, previous employment outside the home. On this basis it is possible to generate a social class classification for both the PCG and SCG.  The household’s social class is then taken as the highest social class category of both partners in the household (as relevant). The household class variable used was categorised as follows:  
  • Professional managers  
  • Other non-manual/skilled manual  
  • Semi-skilled/unskilled manual  

Those with no-valid social class were not included in the models with social class. 

• Household experiencing enforced deprivation at Wave 1. Enforced deprivation was derived using the SILC criteria 
• The PCG family’s ability to make ends meet when the PCG was 16 as reported by PCG at Wave 1  
• The household’s ability to make ends meet at Wave 1  
• Household structure at Wave 1 (lone parent household and non-lone parent household) 
• Housing tenure at Wave 1 defined as owner occupied or other tenure status. Other tenure status included: 
  • Being purchased from a Local Authority 
  • Rented from a Private Landlord/Voluntary Body/Local Authority 
  • Living with and paying rent to parents 
  • Occupied free of rent with parents/from job 
• Household equivalised income quintile at Wave 1  
• How many children’s books did the Study Child have access to at home at Wave 1 (according to the PCG) 
• Age of the PCG at Wave 1 

Below is the full list of variables that were examined against the respondent’s ability to make loan repayments and their ability to make ends meet at age 25 prior to the stepwise process being applied. The variables used in the final model can be found in Table A2 and Table A3 of the Appendix. 

• Household equivalised income quintile at Wave 1  
• Sex of the respondent 
• Highest level of education of the PCG at Wave 1 
• Household social class at Wave 1 (see description above) 
• Household experiencing enforced deprivation at Wave 1. Enforced deprivation was derived using the SILC criteria 
• The PCG’s family’s ability to make ends meet when the PCG was 16 as reported by PCG at Wave 1 
• The household’s ability to make ends meet at Wave 1  
• Household structure at Wave 1 (lone parent household or two parent household) 
• Housing tenure at Wave 1 (see description above) 
• Age of the PCG at Wave 1 

FAQ

What are regression models?

A regression model is a tool that uses information to make predictions. In this release, we use logistic regression models. These models use the data provided to predict whether something is more or less likely to happen.

What do these regression results mean?

In this release, we report our regression results as average marginal effects. An average marginal effect is a type of probability statistic that tells us how many more or fewer people out of 100 would end up with a certain outcome when one thing changes, on average.

For example, in this release we found that the parent using the public library for the child at age 9 increases the probability of the respondent having a degree at age 25 by 7%, on average in the sample, relative to respondents with parents who did not use the public library for their child.  This means that out of 100 similar respondents, about seven more respondents would end up with a degree at the age of 25 if they used the library at age 9 compared with those who did not use the library at age 9, on average.

Why do we refer to the average?

We say on average as this number will not be exactly the same for every group or every person. Take the example in the previous question and answer, some groups might see more than a 7% increase and some groups might see less than that, but 7% is the typical result overall.

What do these results not mean?

These results do not mean that a certain outcome is guaranteed based on a certain set of factors. If we again refer to the example regarding library use, it does not mean that everyone who used the library at age 9 will get a degree and vice versa.

The results do not mean that the factors examined in this study are the only factors that influence education and financial well-being at age 25. There are likely many other factors that influence these outcomes.