Modelling of Education Outcomes: How Do Childhood and Socioeconomic Factors Link to Later Educational Achievement?

• In Event: Effects of Social Disparities on Early Child Development: Results Based on European Birth Cohort Studies

Abstract

Trying to understand the mechanisms and potential ‘policy levers’ that are underneath educational and life trajectories is a priority area for education research.

Through the analysis of longitudinal data a number of socio-economic and intergenerational factors, such as parents education level and income, have been identified and accepted as significant predictors of educational and life outcomes (Hartas, 2012, Reilly et al., 2010, De Coulon, Augustin & Vignoles, 2008). We see this replicated in administrative datasets, where governments and researchers etc. monitor changes in educational attainment and explore inequalities in educational trajectories (Department for Education England, 2019, Education Endowment Foundation, 2018).

While key predictors and measures have been identified they can be too broad and open to debate. For example, the 'big word gap' shown by Hart and Risley (1995), showing the differences in the verbal input provided by parents to children up to 36 months of age has been key in establishing parental income as a predictor for child attainment. However, there has been concern expressed about the construct of the word gap itself (Raz & Beatty, 2018) and about the over extrapolation of data from the Hart and Risley study, which was based on only 42 children rather than population data. In comparison data from the UK's Millennium Cohort Study (MCS), while there are differences in the language performance of young children based on socio-economic factors, there is tremendous variability within groups (Law et al., 2013, Wareham et al., in preparation). Addressing these limitations and further exploration of the variation can ensure that better interventions and more informed policy can be developed.

The MCS is a UK representative longitudinal birth cohort study that began in 2000 with nearly 19,000 families. Permission has been given by just over 9,000 of these families for their data to be linked to wider UK administrative data. Data is held by the UK data service, and given the sensitivity of this data, access is granted through special request. This is currently in process and while we await approval are developing and refining analysis models using proxy education outcomes measures that readily available in the MCS data (i.e. vocabulary test at age 14).

This linked data provides a new and unique opportunity to bring together detailed data on cohort members family background (i.e. home learning environment and income patterns) and their educational attainment over time. Using analytical techniques such as latent class analysis and sequence modelling we will be presenting results and policy implications at part of the session.

Authors

• Helen Wareham, Newcastle University

• James Law, Newcastle University