1. Introduction
Reducing child poverty is currently very high on the national policy agenda in Canada. For example, the issue received enormous attention in December of 1999 when media and child poverty activists pointed out that the incidence of child poverty has actually increased in Canada since all parties agreed to the elimination of child poverty by the year 2000. There seems to be a popular belief that living in poverty is associated with worse outcomes for children, though this opinion is somewhat in conflict with current academic literature on the subject. In both Canada and the US, there have been studies of the association between child well-being and family poverty and/or income status which find relationships that are small in magnitude or sometimes even insignificant (for example, see Blau, 1999; Curtis et al., 2001; Dooley, et al., 1998a, b, Korenman, et al., 1995; Mayer, 1997). If such findings are correct, the policy implication is that the attention paid to reducing child poverty may be misdirected and resources might better be channelled to children in ways other than through income transfers. This would be a strong conclusion and one which clearly warrants much careful study before it is taken.
In this paper, we examine whether the finding of small or insignificant associations between poverty and child well-being is robust to how we choose to measure poverty.1 There is a large literature in the economics of poverty measurement which emphasizes how sensitive estimates of poverty can be to seemingly esoteric measurement choices such as the 'equivalence scale'2 which is embodied in the poverty lines (see for example, Buhmann, et al., 1988; Phipps, 1993; Ruggles, 1990;). It may thus also be true that poverty measurement choices affect estimates of the association between child outcomes and poverty status.
In Canada, the most common approach to measuring 'low-income' is through the Statistics Canada Low-Income Cutoffs (LICO's). The LICO's are derived using an 'Engel methodology' which judges households to be living with low-income if they devote 20 percentage points more than the average Canadian family to the purchase of necessities. While Statistics Canada is very careful to note that the LICO's measure 'low income' and not 'poverty,' public discussion very much treats the LICO's as official poverty lines.
Thus, the LICO's have a high level of credibility in Canadian policy discussions, and these are the poverty lines used in the Canadian studies of the link between poverty and child health carried out to date (Curtis, et al., 2001; Dooley, et al., 1998a,b). It should nonetheless be noted that there is no consensus even within Canada that this is the 'best' or 'only' way to measure poverty. For example, a Federal/Provincial/Territorial Working Group on Social Development Research and Information has been established by HRDC to develop a 'market-basket measure' (MBM) of poverty (Hatfield, 2002). Moreover, if we move outside the Canadian context, then there are many alternatives to the LICO in use (in fact, no other country of which we are aware uses precisely the LICO approach).3 For example, official US poverty lines were derived by multiplying 'minimum adequate food budgets' for families with different composition by a factor of three. The European Community standard is to define poverty as 50% of median income, and this is the approach most commonly used in the academic literature (see, for example, Ruggles, 1990). The general point is that while the LICO's have enormous credibility within Canada and are extremely important for that reason, there is nothing otherwise which suggests that the LICO's are somehow uniquely suited to identifying poverty in Canada rather than an approach which is more widely in use elsewhere (such as a 50% of median income approach).4
In what follows, we do three things. First, we illustrate how estimates of child poverty can depend upon both choice of data set and choice of poverty line. Thus, we compare incidence estimates obtained using the National Longitudinal Survey of Children and Youth (NLSCY), which is the data set containing information about children's well-being (see Dooley, et al., 1998), with estimates obtained using the Survey of Consumer Finance (SCF), the data source typically employed for measuring income distributions or low-income status in Canada (see Statistics Canada, 1997 a or, b). We do this using five alternative poverty lines. We also show that the sensitivity of poverty estimates to choice of both data set and choice of poverty measure is not unique to Canada by making a similar comparison using equivalent US data sets (the Current Population Survey (CPS) and the National Longitudinal Survey of Youth -- Mother Child Survey (NLSY)).5 Second, given the differences in the level estimates of poverty, we show using simple multivariate analysis, how the characteristics associated with a higher probability that a child lives in poverty vary between the NLSCY and SCF and across alternative poverty lines and we again repeat this analysis using the two US data sets. Third, we show how much choice of poverty line matters to conclusions about the importance of the association between child outcomes and child poverty. Throughout, we emphasize that poverty measurement choices affect not only our understanding of the links between poverty and child well-being within countries, but also the conclusions we will draw when making comparisons across countries.
- 1In a companion paper, also funded by HRDC, we examine whether current income and/or poverty status is the best measure of the economic resources available to children. See Curtis and Phipps, 2000.
- 2An equivalence scale indicates, for example, how much more income two people require than one. Thus, for example, if the poverty line for one person is $10,000 and the equivalence scale for two people is 1.5, then the poverty line for the two-person household would be $15,000.
- 3Phipps and Garner, 1993, used comparable Canadian and US data to estimate 'LICO's' for Canada and the US. The key finding of that research was that when exactly the same methodology for deriving equivalence scales was employed, no statistical or practical differences were discernible. Thus, it does not seem to be the case that there are inherent differences across countries which must be reflected in different equivalence, but rather that different countries have simply chosen different approaches.
- 4Statistics Canada also reports Low-Income Measures (LIM's) which use a 50% of median income approach, though these receive less attention.
- 5See Blau, 1999 or Mayer, 1997. The SCF and the CPS are the data sets which are included in the Luxembourg Income Study and which thus form the basis of a substantial body of literature comparing child poverty across countries. See Bradbury and Jantti, 1999 for example.
What's New
