Abstract
In this paper, we ask whether there are Canada/U.S. differences in the extent to which children who were rich versus poor during their early years have developed differences in outcomes by the time they reach adolescence or early adulthood. Using comparable longitudinal data for each country, separate analyses are first conducted for rich compared to poor children living in Canada and rich compared to poor children living in the United States. We then pool data sets to test whether any rich/poor child outcome gaps that have emerged are greater (or smaller) in Canada compared to the U.S. Our data source for Canada is the Statistics Canada National Longitudinal Survey of Children and Youth and for the U.S. we use the National Longitudinal Survey of Youth 79, Child-Young Adult supplement. Key findings include: 1) rich child/poor child outcome gaps are evident for all outcomes in both countries; 2) larger gaps between rich and poor children are evident in the U.S. for math scores and high school completion.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
In terms of health status, this is both because poor children experience more negative shocks and because they are less able to buffer the consequences of shocks so that children can recover completely. Cunha and Heckman (2009) propose a model in which earlier development of child capacities (e.g., cognitive skills, non-cognitive skills and health) enhance the productivity of later investments; moreover, higher capacity in one dimension is argued to complement the capacity to grow in another (e.g., a healthy child can learn more easily).
A significant body of research documents differences in the extent of intergenerational mobility across countries (see, for example, Black and Devereux 2011 for an overview). Thus, we know that the end points differ, but know less about how/why less mobility takes place in the U.S. than in Canada or, especially, in Scandinavian countries, for example.
A number of studies emphasize that ‘permanent’ income has larger associations with child outcomes (e.g., Phipps and Lethbridge 2006)
Social assistance programmes are a provincial responsibility; the definition of being ‘in need’ thus varies across provinces.
Probit estimates of the probability of attending private school confirm that the rich-child/poor-child difference in private school attendance is statistically larger in the U.S. than in Canada. Bradbury et al. 2012 make the same point.
To the extent that any measures are not exactly comparable across countries, the research strategy of comparing rich/poor child outcome differences (rather than outcome levels) should be helpful.
Thus, we also avoid a potential cross-country comparability problem with ‘college’ which in Canada often means pursuing a two-year technical diploma at a ‘community college’ rather than a university degree.
All analyses have also been conducted using: i) the normalized math score, i.e., math score demeaned and divided by the standard deviation; ii) the quintile position of the child’s math score in own country. Results are extremely robust to these alternatives.
After-tax income is not available in all cycles of the NLSCY.
U.S. mothers report income in paper-and-pencil interviews (PAPI) before 1993 and in Computer-assisted personal interviews (CAPI) beginning in 1993 (Bureau of Labor Statistics (2008).
Canadian CPI is taken from CANSIM Table 3260002; US CPI is taken from CANSIM Table 3870007; PPP is taken from CANSIM Table 3800058.
All models have been estimated using both ordinary least squares and probit analyses. Conclusions are robust to choice of estimation technique.
We tried to use a ‘teen-age mother’ indicator but had insufficient numbers for Canada. We also estimated a quadratic in age (since children of older mothers may have additional health concerns, for example) but found the linear specification in mother age to be the best fit.
We also tried to include an indicator that the mother was attending school at the original survey date, but sample size did not allow us to use this indicator. Mother’s health status was similarly a problem with the Canadian data.
We are unable to control for ‘non-white’ status in the Canadian data given small sample size; we have run all U.S. models including this variable. Results are reported in Appendix 3.
Since the magnitude of estimated associations is not directly evident from probit coefficients, we report ‘average marginal effects.’ That is, using estimated coefficients and his/her own personal characteristics for all but the explanatory variable of interest, we calculate the percentage point change in the probability of, in this case, aspiring to high education,’ for each child as we change a particular explanatory variable (e.g., from ‘bottom quintile in period 1’ = 0 to ‘bottom quintile in period 1’ =1). The average marginal effect is then computed over all children.
Note that sample size falls due to non-response to some of the risk factor variables. In the interests of space, we do not report all estimated coefficients, but focus only on what happens to the estimated size and significance of ‘bottom quintile in first cycle’ after risk factors have been included.
U.S. results are unchanged when ‘non-white’ is included as a risk factor, though ‘non-white’ is associated with lower math scores (see Appendix 3).
Bailey and Dynarski (2011) find that the rich/poor gap in both high school completion and post-secondary achievement has increased over time in the U.S.
If we use U.S. quintile cut-points for Canada when selecting Canadian children whose family income 10 years ago would have put them in either the bottom or top of the U.S. income distribution, we lose some observations, in particular because fewer than 20 % of Canadian children had family incomes high enough to place them with the top 20 % of U.S. incomes. It is also true that 21.4 % of Canadian children had family equivalent incomes less than the cut point for the bottom U.S. quintile.
Again, although we use different samples for different adolescent/young adult outcomes, for the sake of brevity, we simply present results about ‘income stickiness’ for this one broad age group.
Again, please notice that our research focuses on the period between 1994 and 2008. To the extent that policies/institutions have changed since that time, these findings may be less relevant.
References
Almond, D., & Currie, J. (2011a). Human capital development before age five. Handbook of Labor Economics: Elsevier.
Almond, D., & Currie, J. (2011b). Killing me softly: the fetal origins hypothesis. Journal of Economic Perspectives, 25(3), 153–172.
Bailey, M. J., & Dynarski, S. M. (2011). Gains and gaps: Inequality in U.S. College entry and completion. National Bureau of Economic Research. Working Paper 17633.
Black, S., & Devereux, P. J. (2011). Recent developments in intergenerational mobility. In Handbook of Labor Economics. 4:5, Ashenfelter, Orley and Card, David (ed). Elsevier.
Bradbury, B., Corak, M., Waldfogel, J., & Washbrook, E. (2012). Inequality in early childhood outcomes. In J. Ermisch, M. Jantti, & T. Smeeding (Eds.), From parents to children: The intergenerational transmission of advantage (pp. 87–119). New York: Russell Sage Foundation.
Brooks-Gunn, Jeanne & Duncan, Greg J. (1997). The effects of poverty on children. The Future of Children. 7:2 Children and Poverty, pp. 55–71.
Bureau of Labor Statistics. (2006). Data user guide, National longitudinal survey of youth 1979 children & young adults.
Bureau of Labor Statistics. (2008). NLSY79 user’s guide: A guide to the 1979–2006 national longitudinal survey of youth data.
Bureau of Labor Statistics. (2009). NLSY79 child & young adult data users guide: A guide to the 1986–2006 child data 1994–2006 young adult data.
Burton, Peter. Phipps, Shelley & Zhang, Lihui. (2012). The Prince and the Pauper: Movement of children up and down the Canadian income distribution. http://myweb.dal.ca/phipps/Prince%20and%20Pauper%20April%202012.pdf.
Case, A., Lubotsky, D., & Paxson, C. (2002). Economic status and health in childhood: The origins of the gradient. American Economic Review, 92, 1308–1334.
Corak, M. (2006). Do poor children become poor adults? Lessons for public policy from a cross country comparison of generational earnings mobility. Research on Economic Inequality, 13, 143–188.
Corak, Miles. Curtis, Lori & Phipps, Shelley. (2011). Economic mobility, Family background, and the well-being of children in the United Sates and Canada, Chapter 3 in Persistence, Privilege and Parenting: The Comparative Study of Intergenerational Mobility. Robert Erikson, Markus Jantti, and Timothy Smeeding (Editors), Russell Sage Foundation, pp. 73–108.
Cunha, F., & Heckman, J. (2009). The economics and psychology of inequality and human development. Journal of the European Economics Association, 7(2–3), 320–364.
Currie, J. (2011). Inequality at birth: some causes and consequences. American Economic Review, 101(3), 1–22.
Currie, J. (2009). Health, wealthy, and wise: socioeconomic status, poor health in childhood, and human capital development. Journal of Economic Literature, 47(1), 87–122.
Currie, J., & Stabile, M. (2003). Socioeconomic status and child health: why is the relationship stronger for older children? American Economic Review, 93, 1813–1823.
Danziger, S. K., & Danziger, S. (2010). Child poverty and antipoverty policies in the United States: Lessons from research and cross-national policies. In S. B. Kamerman, S. Phipps, & A. Ben-Arieh (Eds.), From child welfare to child well-being: An international perspective on knowledge in the service of policy making. New York: Springer Press, pp. 255–274.
Ding, W., & Lehrer, S. F. (2010). Estimating Treatment Effects from Contaminated Multiperiod Education Experiments: The Dynamic Impacts of Class Size Reductions. The Review of Economics and Statistics, 92(1), 31–42.
Esping-Andersen, G. (1990). The three worlds of welfare capitalism. Princeton: Princeton University Press.
Fortin, N., Oreopoulos, P., & Phipps, S. (2012). Leaving boys behind: Gender disparities in high academic achievement. http://faculty.arts.ubc.ca/nfortin/LeavingBoysBehind.pdf.
Frenette, M. (2005). Is post-secondary access more equitable in Canada or the United States? Statistics Canada. Analytical Studies Branch Research Paper Series. Catalogue no. 11F0019MIE – No. 244.
Human Resources and Social Development Canada. (2010). National Child Benefit Progress Report, 2007. Available at: http://www.nationalchildbenefit.ca/eng/07/sp_119_11_07_eng.pdf (accessed October, 22, 2012)
Luxembourg Income Study (LIS) Key Figures, http://www.lisproject.org/key-figures/key-figures.htm ({October 4, 2012})
Phipps, S., & Lethbridge, L. (2006). Income and the outcomes of children. Statistics Canada. Analytical Studies Branch Research Paper Series. Catalogue No. 11F0019MIE – No. 281.
Picot, G., Zyblock, M., & Pyper, W. (1999) Why do children move into and out of low income? Changing labour market conditions or marriage or divorce? Statistics Canada, Analytical Studies Branch Working Paper.
Smeeding, T. M., Erikson, R., & Jantii, M. (Eds.). (2011). Persistence, privilege, and parenting: The comparative study of intergenerational mobility. New York: Russell Sage.
Statistics Canada. (2005). Microdata user guide, national longitudinal survey of children and youth, cycle 6.
Statistics Canada. (2008). Microdata user guide, national longitudinal survey of children and youth, cycle 8. Available at: http://www23.statcan.gc.ca/imdb-bmdi/document/4450_D4_T9_V8-eng.pdf
Waldfogel, J., & Washbrook, E. (2011). Income-related gaps in school readiness in the United States and the United Kingdom. Chapter 6. In T. M. Smeeding, R. Erikson, & M. Jantii (Eds.), Persistence, privilege, and parenting: The comparative study of intergenerational mobility (pp. 175–207). New York: Russell Sage, pp. 175–208.
Wooldridge, J. M. (2002). Inverse probability weighted m-estimators for sample selection, attrition, and stratification. Portuguese Economic Journal, 1(2), 117–139.
Wooldridge, J. M. (2001). Econometric analysis of cross section and panel data. Cambridge, Massachusetts: The MIT Press.
Author information
Authors and Affiliations
Corresponding author
Appendices
Appendix 1
Table 13 shows the detailed process of arriving at our baseline regression (as in Tables 2, 3, 4, 5, 6 and 7) samples. Starting from the samples of children in appropriate age ranges 10 years ago, we follow five steps to construct our baseline regression samples. First, we use the lowest common denominator approach to make the samples from both countries comparable to each other. Second, we drop those observations that were in the survey in the first but not in the sixth cycle. Third, we drop those who do not have valid responses to our dependent variables. Fourth, those without valid responses to our independent variables are dropped. Finally, we keep only those in incomes quintiles 1 or 5 given our primary interest in the top–bottom disparity. As seen in Table 13, in the Canadian case, most observations are dropped for the purpose of making the samples comparable to the US. Attrition is a large cause for loss of observations for both countries. In addition, non-responses to the dependent variables and to the independent variables lead to some further reductions in sample sizes. The last step drops about 60 % of the sample, which is as expected.
Given the significant number of observations dropped due to attrition and non-responses, a natural concern is that this may bias the statistical results obtained from those children that remain in the sample if such drops are not at random (Wooldridge 2001). That is, the elimination of observations depends on either observables or un-observables.
In our baseline regressions, we use the longitudinal weights from the sixth cycle supplied by Statistics Canada in the NLSCY masterfiles, which are intended to preserve the representativeness of the original longitudinal children, given sample attrition (Statistics Canada 2005).
To further confirm that these baseline regression results are robust to the sample reductions, we use the inverse probability weighted (IPW) M-estimator approach suggested by Wooldridge (2002) and illustrated by Ding and Lehrer (2010). The IPW approach produces consistent estimators provided that attrition is based on observables rather than un-observables, that is, attrition probability is independent of the dependent variable. Following this approach, we first estimate the probability of a child staying in the sample using probit regressions, where staying in the sample means that the child is present in both the first and the last cycle and has valid responses to the dependent variables. The independent variables include those used in our baseline regressions, with the bottom income quintile dummy replaced by the log of household equivalent income to retain more information on family economic resources. To increase the precision of our estimated probabilities of staying in the sample, we also include four dummy variables indicating whether the child is present in the second, third, fourth, and fifth cycle, respectively, and we make use of children in all income quintiles, not just those in the bottom and top quintiles 10 years ago. The second step entails using the inverse of the estimated probabilities of staying in the sample obtained from the first step to reweight our baseline regressions. The IPW probit regression results are presented in Tables 14, 15, 16, 17, 18 and 19. As explained in Wooldridge (2002), here the standard errors are overly large rendering more conservative inferences, i.e., we are less likely to reject the null hypotheses. A comparison between results in Appendix Tables 14, 15, 16, 17, 18 and 19 and those in Tables 2, 3, 4, 5, 6 and 7 reveal no dramatic differences in the estimates, suggesting non-random attrition bias, at least in terms of observables, is not a significant concern.
Appendix 2. Details of Sample Construction for Outcome Variables
1.1 Equivalent Household Income
Canadian sample: 10–17 years old in 2004 (0–7 years old in 1994)
U.S. sample: a pooled sample of four cohorts: i) 16–17 years old in 200 (6–7 years old in 1990); ii) 16–17 years old in 2002 (6–7 years old in 1992); iii) 10–17 years old in 2004 (0–7 years old in 1994); and, iv) 10–11 years old in 2006 (0–1 years old in 1996).
1.2 Educational Aspirations
Canadian sample: a pooled sample of two cohorts: i) 12–15 years olds in 2004 (2–5 years old in 1994); and, ii) 12–13 years old in 2006 (2–3 years old in 1996)
U.S. sample: a pooled sample of four cohorts: i) 14–15 years old in 2000 (4–5 years old in 1990); ii) 14–15 years old in 2002 (4–5 years old in 1992); iii) 12–15 years old in 2004 (2–5 years old in 1994); and, iv) 12–13 years old in 2006 (2–3 years old in 1996).
1.3 Math Score
Canadian sample: a pooled sample of two cohorts: i) 12–14 years olds in 2004 (2–4 years old in 1994); and, ii) 12–13 years old in 2006 (2–3 years old in 1996)
U.S. sample: a pooled sample of four cohorts: i) 13–14 years old in 2000 (3–4 years old in 1990); ii) 13–14 years old in 2002 (3–4 years old in 1992); iii) 12–14 years old in 2004 (2–4 years old in 1994); and, iv) 12–13 years old in 2006 (2–3 years old in 1996).
1.4 Education Attainment
Canadian sample: a pooled sample of two cohorts: i) 20–21 years old in 2004 (10–11 years old in 1994); and, ii) 19–21 years old in 2006 (9–11 years old in 1996).
U.S. sample: a pooled sample of two cohorts: i) 20–21 years old in 2004 (10–11 years old in 1994); and, ii) 19–21 years old in 2006 (9–11 years old in 1996).
Appendix 3
Table 20
Rights and permissions
About this article
Cite this article
Burton, P., Phipps, S. & Zhang, L. From Parent to Child: Emerging Inequality in Outcomes for Children in Canada and the U.S.. Child Ind Res 6, 363–400 (2013). https://doi.org/10.1007/s12187-012-9175-1
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s12187-012-9175-1