Overview

The data for this study are taken from the synthetic file released for cycle three of the National Longitudinal Survey of Children and Youth (NLSCY). The data provided represent only a subset of the data available. The provided data represent children aged 4, 5 or 6, living in one of 24 major metropolitan areas. 1,016 records are provided. In addition to studying the relationship between child outcomes and determinants, you will learn about hierarchical methods and small area statistics.

Appendices

• Appendix A: Derivation of Socio-economic Status (CINHD08)
• Appendix B: Score on the Peabody Picture Vocabulary Test
• Appendix C: Component variables of CSFHS6

Introduction

Neighbourhood factors such as poverty and residential instability have been identified as being important in explaining neighbourhood problems such as delinquency and crime encountered in many poor urban neighbourhoods (Sampson, 1992; Sampson & Groves, 1989; Sampson & Morenoff, 1997). Neighbourhood conditions of poverty and instability impede the establishment of formal and informal institutions of neighbourhood organization which are believed to maintain and foster strong community relations as well as public order within a community. For example, neighbourhood safety and cohesion or a sense of trust and belonging are seen to strengthen the community and have positive effects on its members. Often these factors are spatially based so that poverty conditions co-occur in similar areas (Massey, 1990; 1996; Massey & Denton, 1993). The geographic or spatial associations may be due in part to housing policies, housing affordability, as well as to conditions of ethnic and economic segregation (Wilson, 1987). For example, public housing is often found in predominantly low socio-economic neighbourhoods leading to areas of isolated and concentrated poverty as well as other separate areas of concentrated affluence. These differences as well as the conditions of neighbourhoods children reside in may be important for child health and well-being. When discussing the associations of neighbourhood characteristics with child outcomes it is important to note that both risk and protective factors occur at multiple levels, individual, family, and neighbourhood and it is not just a single protective or risk factor but the accumulation of factors that result in negative or positive child and family outcomes.
 

The emerging literature on the effects of neighbourhood factors on children and youth has focused on structural characteristics of the neighbourhood such as income/socio-economic conditions and residential instability yet most of the literature is based on studies conducted in the United States. Most studies have focused on outcomes in early childhood or late adolescence (see Leventhal & Brooks-Gunn, 2002 for review). Some consistent findings have been reported. For example, neighbourhood effects for socio-economic factors are more common than effects of residential instability across all child outcomes, and neighbourhood effects are generally small (explaining 5-10% of the variability in outcomes). As would be expected, family level factors tend to be more strongly associated with individual child outcomes than neighbourhood level factors but neighbourhood effects are consistently reported even after controlling for family level factors, for outcomes of children, youth, and adolescents.

Data Description

National Longitudinal Survey of Children and Youth

The National Longitudinal Survey of Children and Youth (NLSCY) is a long-term survey designed to measure child development and well-being. The first cycle of the survey was conducted by Statistics Canada in 1994-1995 on behalf of Human Resources Development Canada. The requirement for the NLSCY design was to select a representative sample of children in Canada and to follow and monitor these children over time into adulthood. All of the information for the household collection was collected in a face-to-face or telephone interview using computer-assisted interviewing (CAI). Questions were asked to the respondent in the home or by telephone and directly entered into a computer by the interviewer.
 

Before the NLSCY was undertaken there were few statistical studies describing a broad range of characteristics of children in Canada. Measures of health, well-being and life opportunities are needed, however, if governments and researchers hope to learn more about the ongoing life conditions of Canadian children and youth, and their developmental experiences. Longitudinal data are central to discovering developmental changes occurring in children over time, and studying the impacts of the social environment of the child and various family-related factors.
 

The primary objective of the NLSCY is to develop a national database on the characteristics and life experiences of children and youth in Canada as they grow from infancy to adulthood. The more specific objectives of the NLSCY are:

• To determine the prevalence of various biological, social and economic characteristics and risk factors of children and youth in Canada,
• To monitor the impact of such risk factors, life events and protective factors on the development of these children,
• To provide this information to policy and program officials for use in developing effective policies and strategies to help young people live healthy, active and rewarding lives.

Underlying these objectives is the need to:

• Fill an existing information gap regarding the characteristics and experiences of children in Canada, particularly in their early years,
• Focus on all aspects of the child in a holistic manner (i.e., the child, his/her family, school, and community),
• Provide national, and as far as possible, provincial-level data,
• Explore subject areas that are amenable to policy intervention and which affect a significant segment of the population.

Background: Survey Weights

Suppose we have a finite population P, of size N=100 individuals. We are interested in estimating a total, mean or other variable of interest from this population. In a simple random sample s of size n=20 (in a simple random sample each individual has the same probability to be selected in the sample) we observe y1, y2, …, y20. How can we estimate the population total Y of y1 to y100? Since the population size is N=100, each individual in the sample represents 5 individuals in the population and is assigned a sampling weight of 5. If wi is the sampling weight of individual I in the sample, in this example, wi=5 for I=1,…,20, the estimator of the total Υ is: 
 

 

In the previous example all the individuals had the same sampling weight. In surveys it is common to select a sample with unequal probability of selection and hence unequal weights. In the data set for this case study, each individual has an associated sampling weight, but they are not all equal, since the survey was not a simple random sample. Using these in the analysis will help the results reflect the survey population, not just the survey sample.

Background: Geo-Codes

Census Metropolitan Area (CMA)

A very large urban area, together with adjacent urban and rural areas that have a high degree of economic and social integration with that urban area. A CMA is comprised of one or more contiguous census subdivisions (CSD). CMA’s are defined by Statistics Canada.
 

A CMA is delineated around an urban area (called the urbanized core and having a population of at least 100,000, based on the previous census). Census subdivisions are included in the CMA on the basis of decennial place-of-work commuting data. Once an area becomes a CMA, it is retained in the program even if its population subsequently declines.
 

Census Metropolitan Area (CMA) codes are listed in the data documentation.

Background: Linking data files

If you choose to include the CMA level variables, you must first merge this data by CMA onto the NLSCY synthetic file. In addition to this file you are free to add on other macro level variables from other sources.

Macro Level Data: National Longitudinal Survey of Children and Youth (NLSCY)
Data file sheet for Download: Excel

The data for this study were taken from published results and we would like to thank Nancy Ross of McGill University for allowing us to use the data in this case study. The file contains five variables, the combined province-CMA code, the median share of income, the Gini coeficient, the percentage of persons below the poverty line, and the median income for each of the CMAs.
 

Income inequality measures were calculated for households in 53 Canadian and 282 U.S. metropolitan areas with populations greater than 50,000 in 1991 (Canada). Income inequality measures for Canadian metropolitan areas were derived from a specially prepared micro data file of the 2B sample of the 1991 Census of Population. The 2B sample represents information gathered from 20% of Canadian households which includes detailed information regarding income sources and amounts. Income included income for all household members from wages and salaries, net self-employment income, government transfers and investment income. All of the measures were calculated with earned household income over 1,000 dollars.
 

Median Share:
 

A median share is a middle-sensitive measure of income inequality defined as the proportion of total or earned household income belonging to the less well-off 50 percent of households within a geographical area. In order to estimate the median share, the population has first to be ranked from low to high income, second, identify the income category containing the 50th percentile of the population, i.e., the median, and finally, calculate the proportion of total household income earned by the first half of the population.
 

The median income falls within the income category that contains the 50 th percentile of the population ranked from low to high. The median income value can be linearly extrapolated assuming that the distribution of income within the income category is linear.
 

Gini Coefficient:
 

The Gini coefficient is an overall measure quantifying the degree of income inequality of a particular income distribution and can be derived directly from the Lorenz curve. The Lorenz curve represents the cumulative distribution of households (horizontal axis) against the cumulative distribution of income (vertical axis) (Figure 1). In situations of perfect equality, the shares of population and income will be equal and a 45-degree line on the graph represents this perfect equality. For example, in a situation of perfect equality, 10% of population has 10% of income. In reality, the actual cumulative shares of income possessed by the cumulative shares of the population will fall below this line of perfect equality. It is this Lorenz Curve that allows the estimation of the Gini Coefficient, a global income inequality measure explained below.
 

The Gini coefficient is calculated as follows:
 

 

where A is the area surrounded with the line of perfect equality and the Lorenz curve and B is the area below Lorenz curve (see Figure 1). It is clear from the Figure that the Gini coefficient is a middle-sensitive inequality measure since the measure is more sensitive to the middle range of the income distribution. As is true for any measure of proportions, the Gini coefficient lies between 0 and 1, where a Gini coefficient close to 0 indicates a more equal income distribution while a coefficient close to 1 indicates a more unequal income distribution. Using this measure in isolation, however, can be somewhat misleading given that, for example, two Gini coefficients can be equivalent with totally different underlying Lorenz curves. 
 

Figure 1: Lorenz Curve for the State of Alabama, 1990
 

 

In order to facilitate the calculation of the Gini coefficient, the above equation can be re-expressed as Gini=1-2B, since the area under the line of perfect equality (see Figure 1) is A+B=1⁄2, therefore A=1⁄2 -B and the Gini=(1⁄2-B)/1⁄2 . As such, only the area B needs to be calculated to estimate the Gini coefficient.
 

Proportion of persons below the poverty threshold of half the median income
 

This measure is defined as the proportion of persons below half of the median inco me. Persons living under this threshold are considered living in poverty for the purposes of this study.
 

Coefficient of Variation
 

The coefficient of variation (CV) is a summary measure of income dispersion (illustrating also the degree of inequality in the income distribution) and is considered to be a “top-sensitive” income inequality measure. High incomes will result in a greater increase in the CV compared to low or average incomes. The CV is the standard deviation of income divided by the average income and can be written as:
 

 

where
 

is the overall average income, pi is the proportion of the population within income category i and yi is the average income within the income category i.
 

Urban, population 100,000 to 499,999
 

The proportion pi is identical to the rectangle width used for the Gini coefficient calculation. This measure gives more weight to larger deviations and expresses the standard deviation as a proportion to the average income. The larger the CV, the greater the income inequality and the skewness in the income distribution.
 

Median Income
 

The median income of the city is the median income where income included income for all household members from wages and salaries, net self-employment income, government transfers and investment income above 1,000.