About the National Longitudinal Survey of Children and Youth

View the most recent version.

Survey objectives
Survey design
The present sample
Weighting and variance estimation
Analysis of non-response

Survey objectives

The National Longitudinal Survey of Children and Youth (NLSCY) is a long-term study of Canadian children that follows their development from birth to early adulthood. The NLSCY began in 1994 and is conducted jointly by Statistics Canada and Human Resources and Social Development Canada. The survey is designed to collect information about factors influencing the social, emotional, behavioural, physical and cognitive development of children, and to monitor the impact of these factors on their development over time. The survey covers a broad range of topics including health, physical development, learning, behaviour, and social environment (family, friends, schools and communities).

Survey design

The NLSCY sample frame in the first cycle of the survey (1994-1995) was based on the Labour Force Survey (LFS), a monthly survey of households in Canada conducted by Statistics Canada. Households with children aged 0 to 11 in the ten provinces were selected from the LFS in 1994 to participate in the NLSCY. Of 26,000 eligible households, approximately 23,000 responded.

The first cycle of the NLSCY was conducted between November 1994 and June 1995. The longitudinal cohort from that cycle has been monitored every two years since. In addition, new panels of children aged 0 to 1 have been added to the survey at each data collection. These children are also followed every two years until the ages of 4 and 5, after which time they are dropped from the sample.

For children under 16 years of age, most of the information in the survey is provided by the person most knowledgeable about the child (known as the PMK in the NLSCY), usually the mother. She provides information about herself, the household and family, and the child. In addition, children aged 10 and over provide some information about themselves on a self-completed questionnaire. Direct measures of the child's abilities may also be taken, depending on the child's age. In some cycles of the NLSCY, school teachers and principals also completed a survey, again depending on the age and school status of the child, and on whether permission is given by the parent.

The present sample

Four cycles of NLSCY data were used in this analysis. The collection periods were as follows: Cycle 3 (October 1998 to June 1999), Cycle 4 (September 2000 to May 2001), Cycle 5 (September 2002 to June 2003), and Cycle 6 (September 2004 to June 2005).

Children were identified for inclusion in this study from each cycle at the age of 0 to 1 if they met the following criteria:

1. they responded in all applicable cycles
2. their biological mother was the PMK at each cycle
3. they were first-born children

Children were classified as first-born if their year of birth matched the year for which the mother reported having her first child.

Table A.1 presents the sample selection breakdown for the children in this study.

The 3,382 children retained for the present study represent about 556,000 children in the Canadian population.

This study was interested in outcomes measured during early childhood. Due to the NLSCY longitudinal sample design, early childhood was limited to ages 0 to 5 and data were collected at starting at ages 0 to 1, two years later at ages 2 to 3, and once again at ages 4 to 5.

For this study, children were pooled together from four cycles of the NLSCY. As a result, not all children had data at all time points. The availability of data depended upon the cycle in which the child was introduced to the NLSCY. As can be seen in Table A.2, while all of the children in this study have data for characteristics measured at ages 0 to 1, only those children who were introduced in Cycles 3 and 4 have data at ages 4 to 5.

As a result, there were 3,382 children with data at ages 0 to 1, 2,365 children with data two years later at ages 2 to 3, and 1,705 children with data two years after that at ages 4 to 5.

Weighting and variance estimation

To calculate all point estimates, the appropriate longitudinal survey weight was used (Table A.3), along with the corresponding bootstrap weights for variance estimation and significance tests. Children introduced in Cycle 6 were the exception; the cross-sectional survey and bootstrap weights were used for this cohort. All weights were taken from the last cycle to which the child responded.

Each of these survey weights allows one to make inferences about the population of Canadian children represented by the sample that entered the survey at ages 0 to 1.

Analysis of non-response

A number of children were dropped from this study because they did not meet the sample selection criteria (Table A.1). In effect, these children became non-respondents. Given that non-response has the potential to bias results if non-respondents have significantly different characteristics from respondents, it was important to assess the possible impact on the analysis.

As mentioned in the previous section, 3,094 eligible 0 to 1-year-olds were lost to attrition and a further 3,211 were excluded because their biological mother was not the PMK at all cycles. Of this first group, 625 were first-born children whose biological mother was the PMK during the first interview. Of the second group, 396 were first-born children whose biological mother was the PMK during the first interview. For the following analysis of non-response, these two subgroups were compared to the 3,382 children retained in the study. Comparisons were made with respect to a number of characteristics that were assessed at ages 0 to 1 and were discussed in the main analysis.

For comparison purposes, each child was assigned his/her cross-sectional weight from their first cycle of interview, and the corresponding bootstrap weights were applied for variance estimation and significance tests (Table A.4). Note that for respondents in the final study sample, this weighting strategy differs from that used in the main analysis.

Socio-demographic characteristics. Children lost to attrition (Group 1) differed significantly from children in the study sample (Group 3) on several socio-demographic characteristics (Table A.5). A greater proportion of Group 1 children had a teenaged or young mother, had a mother with a high school diploma or less, lived in a low-income household, or had a mother who was a single parent.

There were almost no differences between children whose biological mother had not been the PMK over time (Group 2) and children in the final sample (Group 3). The exception was that a lower proportion of Group 2 children had a mother who was a single parent (Table A.5).

Prenatal and perinatal characteristics. Compared to children in the study sample, a lower proportion of Group 1 children had a mother who suffered from gestational diabetes, while a greater proportion had a mother who smoked throughout pregnancy (Table A.5). There were no differences between Group 2 children and those in the study sample.

Neither Group 1 nor Group 2 children differed from children in the study sample with respect to perinatal characteristics.

Health characteristics. The only significant difference between Group 1 and Group 3 children was the mother's higher average depressive symptoms score among the former group (Table A.5). In addition, a smaller proportion of Group 2 children reported that the child was in excellent or very good health compared to the children in the final study sample.

The non-response analysis suggested that excluding children from the study sample because the PMK at each interview was not the biological mother did not result in any significant bias. Children excluded from the study for this reason did not differ significantly from the final sample on any of the examined characteristics except for single-parent status and the child's general health measure.

There were, however, several significant differences between the children lost to attrition and children in the study sample. For the most part, all of these differences reflect the fact that children who drop out of the NLSCY over time in general are more likely to have lower socio-economic status (SES) than those who remain. There are two main reasons why possible bias resulting from the loss of these children is not a serious concern to the main study. First, the longitudinal weights provided with the data have been adjusted to address this type of non-response. Second, the children born to older mothers were, on average, of relatively high SES and therefore less likely to drop out of the survey. Bias may have been more of a concern if the focus of this paper had been on children of teenaged mothers rather than those of older mothers. Given their low SES, the former had an increased likelihood of being under-represented in the selected sample.