Poverty and Child Well-Being in Canada and the United States:Does it Matter How We Measure Poverty? - September 2000

5. Conclusion

Previous

Contents

Next

This paper examines the robustness of our conclusions about the association between child poverty and child well-being to alternative issues of measurement. Specifically, we focus upon the sensitivity of results to choice of data set, choice of sample and choice of poverty line. Throughout the paper, we pay attention, as well, to how Canada/US comparisons are affected by these choices. Our conclusions include the following:

1. Canadian National Longitudinal Survey of Children and Youth (NLSCY) and Survey of Consumer Finance (SCF) data yield rather different estimates of the incidence of child poverty; the same is true for the US Current Population Survey (CPS) and National Longitudinal Survey of Youth (NLSY) surveys. To the extent that the income estimates of the NLSCY and/or NLSY may be less reliable, some caution should be exercised in the interpretation of findings about the link between income/poverty and child health at this stage. Perhaps future waves of NLSCY data can improve upon the collection of income information. It is also worth noting that while SCF and CPS data indicate about a 10 percentage point gap in the incidence of child poverty between Canada and the US, this gap essentially disappears if we use NLSCY and NLSY data since the Canadian child outcomes data set produces higher estimates of child poverty than the SCF while the US child outcomes data set produces lower estimates than the CPS.
2. Given a particular data set, alternative common procedures for measuring poverty (e.g., the Canadian Low-Income Cutoffs (LICO) versus the US official poverty lines) yield different estimates of the incidence of child poverty.
3. Correlates of child poverty differ somewhat for different data sources (i.e., NLSCY and SCF; NLSY and CPS), sample selections, and poverty lines. Specifically, child age does not have a significant relationship with poverty status using SCF data, controlling for mother's age, but being 8 to 11 years old is associated with a lower probability of being poor using NLSCY data. However, this relationship disappears if we restrict the sample to children whose mothers are aged 29 to 37 years in order to match a key sample restriction of the US NLSY data set. One important message to take from this example is that the NLSY data set is not entirely representative of all US children of a particular age, and results obtained using this data set can be sensitive to this limitation.
4. Finally, however, we find that while poverty is almost always significantly associated with worse outcomes for children regardless of how it is measured for the four outcomes we study here (body mass index, trouble concentrating, hyperactivity, Peabody picture vocabulary score). Thus, our conclusions about the link between child poverty status and child outcomes appear to be less sensitive to measurement issues than our conclusions about the extent of poverty. There are, however, exceptions to this generally reassuring conclusion. For example, if we measure poverty with the US official poverty lines (which are very low by comparison with all other alternatives considered), we find no relationship between poverty and body mass index or trouble concentrating for the US; between poverty and hyperactivity for Canada.