Growing Up in Scotland: patterns of maternal employment and barriers to paid work

2 Methods

2.1 Introduction

This chapter provides details on the methods used in the production of the report, including details about the data and the analytical approach. First, it provides details about the Growing Up in Scotland ( GUS) data, including detailed notes about the sample selected for the analysis presented here. Next, it outlines the analytical approach taken, including notes on how to interpret the findings.

2.2 Data and sample

The analysis presented in this report uses data from the two GUS birth cohorts. More specifically it uses data collected at the three age points where comparable data is available, namely when the cohort children were aged 10 months, 3 years and 5 years ^[9] . For Birth Cohort 1 ( BC1), this means the analysis draws on data collected in 2005/06, 2007/08 and 2009/10. For Birth Cohort 2 ( BC2) the corresponding data were collected in 2011, 2013 and 2015 (Table 2‑1).

Table 2‑1: Sample overview

The main data collection on GUS takes place through annual or biennial ‘sweeps’ of face-to-face survey interviews with the cohort child’s main carer. In the vast majority of cases this is the child’s mother. At each sweep, information previously collected about the mother’s employment is checked and, where necessary, updated. If the mother has a resident partner their employment details are collected too. In addition to this, a range of information about the household is obtained including household income and the mother’s educational attainment.

This means that GUS data contains information about mothers’ employment as well as about a range of individual and household circumstances collected across a number of different time points.

Given the focus of the research, cases where information about the mother's employment status was missing at one or more relevant sweeps were excluded from the analysis, and all analyses were restricted to cases where the main respondent was the cohort child’s mother ^[10] .

Analysis which used information about the mother’s employment during pregnancy was further restricted to cases where the respondent was the child’s biological mother. Furthermore, some analyses use data only from the youngest cohort ( BC2). Finally, for the longitudinal analyses only cases where the cohort child’s mother was the main respondent at all relevant sweeps were included.

These restrictions mean that base sizes vary across and within chapters. Clear descriptions of the groups included in the analysis are provided in the text and base sizes are provided for all tables and charts.

2.3 Analytical approach and interpreting the findings

The report makes comparisons between mothers in the two cohorts, as well as comparisons between mothers within each cohort at different time points: at the time the cohort child was aged 10 months, 3 years and 5 years. It also looks at mothers’ employment status and certain employment trajectories according to a number of socio-economic and demographic characteristics. Details of key variables are provided at the beginning of the relevant chapters. Details of the remaining variables are available in Appendix A ^[11] .

As already noted, the GUS sample design means that the data can be used to produce estimates about all mothers of children of a certain age living in Scotland. For example, based on GUS data we can estimate the proportion of mothers of 5 year old children living in Scotland in 2009/10 and in 2015 who were in paid work.

A substantial part of the analysis presented in this report consists of bivariate analyses comparing differences in outcomes or experiences for mothers according to their status measured using a single variable – for example, employment status or household income. Unless otherwise stated, only differences which were statistically significant at the 95% level or above are commented on in the text.

Not all families who initially took part in GUS did so for all of the subsequent sweeps. There are a number of reasons why respondents drop out from longitudinal surveys and such attrition is not random. Therefore, the data were weighted using specifically designed weights which adjust for non-response and sample selection. Different weights were applied for cross-sectional and longitudinal analyses. All results have been calculated using weighted data and all comparisons take into account the complex clustered and stratified sample structures.

2.3.1 Multivariable analysis

Many of the factors we are interested in are related to each other as well as being related to maternal employment. For example, younger mothers are more likely to have lower educational qualifications, to be lone parents, and to live in areas with high levels of deprivation. Simple analysis may identify a relationship between maternal age and maternal employment – for example, younger mothers are more likely to give up work after having a child. However, this relationship may be occurring because of the underlying association between maternal age and education. Thus, it may actually be the lower education levels among younger mothers which are driving the association with giving up work rather than the fact that they are younger in age.

To avoid this difficulty, multivariable regression analysis was used. This analysis allows the examination of the relationships between an outcome variable (e.g. whether a mother gave up work after having a child) and multiple explanatory variables (e.g. the mother’s age and education level, household income, whether she lived with a partner, etc.) whilst controlling for the interrelationships between each of the explanatory variables. This means it is possible to identify an independent relationship between any single explanatory variable and the outcome variable. In this report, this means, for example, that we can identify characteristics which are independently associated with being in the position of seeking work whilst having a young child, and characteristics independently associated with giving up work after having a child. Note, though, that the identification of associations between one or more explanatory variables and an outcome variable does not necessarily imply that the explanatory variable(s) causes the outcome.

The multivariable analysis undertaken for this report uses logistic regression models. Full results of the models are included in the Technical Annex along with notes on how to interpret them.

Note that the statistical analysis and approach used in this report represents one of many available techniques capable of exploring this data. Other analytical approaches may produce different results from those reported here.