3. Overview of Empirical Strategy
As noted in the introduction, this project examines three inter-related issues: whether past levels of resources, contexts and opportunity structures carry long-term consequences for subsequent child attainments and behaviours; whether shocks, or transition events, alter the path of these outcomes in a positive or negative fashion; and what role is played, and can be played, by policy interventions in improving these attainments and behaviours. In this section, we provide some additional detail about the data set we employ — Cycles 1, 2 and 3 of the National Longitudinal Survey of Children and Youth (NLSCY). We then explain how, by exploiting two features of the NLSCY — its longitudinal nature, and the existence of cohorts of children for whom different types of information are available — the research questions noted above are addressed.
The National Longitudinal Survey of Children and Youth (NLSCY) is an ongoing survey of Canadian children designed to help analyze child development and well-being. There are three cycles of data available to date with interviews in 1994, 1996 and 1998 with the expectation of continued biennial interviews until the children reach the age of 25. In addition to the longitudinal file, cross-sectional data are available for each survey year yielding nationally representative results when the sampling weights are applied. In 1994, the sample was of children 0-11 years of age , in 1996 0-13 and in 1998 0-15. Additional children are added to the cross-sectional file each survey year in order to maintain both nationally and provincially representative samples for each year.
The sample for the NLSCY was originally drawn from the Labour Force Survey (a monthly survey by Statistics Canada used to produce labour force information). The survey uses a multistage probability sample where each province is an independent sample. Through stratification, cities, small urban areas and rural areas are broken down into clusters of dwellings from which households are surveyed. From the LFS, households containing children could be selected for the NLSCY. Note that the LFS excludes those living in institutions and on Indian Reserves. In cycle one, 22,831 children were interviewed which included about 5,000 children from households of those in the National Population Health Survey. These children were dropped from the second cycle due to budget constraints. In cycle one, up to four children per household were interviewed but by cycle two, only two children per household were interviewed (for those households with more than two children, those interviewed were randomly selected). Again, this was due to budget constraints. However there was a large increase in the number of children 0-5 year olds interviewed leading to a total sample of 20,025 children in cycle two. For cycle three, no new siblings of children already in the survey were interviewed (as they were in cycle two) but new children selected from the Labour Force Survey households and birth registries increased the sample size to 31,194 0-15 year olds.
Much of the information which we use for our analysis was collected during a personal interview with the 'person' most knowledgeable about the child (pmk), generally the child's mother. In fact, for over 90% of children in the survey, the mother is the pmk. We also use results of tests administered to the child.
In carrying out our research, we focus on four cohorts. Note that for each cohort, the NLSCY has detailed "historical" data (i.e. information on families, neighbourhoods and schools) collected in the first cycle, as well as indicators of attainments as of 1994:
- Children aged 0-2 during cycle 1, the "baby cohort". For these children, there is detailed "attainment" information, for example, on birth-weight, chronic health conditions, and nursing history while infants. The 'youngest' children are aged 4-6 when cycle 3 was fielded and thus will taken the Peabody Picture and Vocabulary Test (PPVT), an indicator of school readiness;
- Children aged 4-5, whose attainments as of 1994 are captured by the PPVT fielded during cycle 1, the "pre-schooler cohort". They are now aged 8-9 when cycle 3 was fielded and will have completed indicator tests on math and reading ability.
- Children aged 7 and 9 in 1994, who completed the math indicator test during cycle 1, the "middle cohort" (there are problems with the 1994 math test for children aged 8 and 10 and hence we exclude these children). Children in our middle cohort are aged 11 and 13 when cycle 3 was fielded and will have completed further indicator tests on math and reading ability in 1996 and 1998.
- Children aged 11 in 1994 and 15 in 1998, the 'oldest cohort'. These children wrote the math indicator test in 1994 and completed subsequent tests (math and reading) in 1996 and 1998.
By using these four cohorts, we can answer:
- To what extent is past history "destiny"? That is: if in the youngest cohort, conditions of advantage and disadvantage at birth are strongly associated with levels of early school readiness; if, in the pre-schooler cohort, such readiness is associated with subsequent early school success; if in the "middle and older" cohorts early school success is associated with desirable subsequent attainments; and if we assume that these associations are stable over time, this analysis can tell us to what extent the level and distribution of early advantage and disadvantage plays out in attainments observed in early adolescence.
- What role is played by shocks such as family income loss, changes in household structure (resulting from divorce, re-marriage or other changes to household living arrangements) or significant changes in the child's health? To what extent are the implications for Children's attainments of shocks asymmetric? Do negative shocks have adverse effects, but positive shocks have limited, if any advantageous effects?
What's New
