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

2. Alternative Estimates of the Incidence of Child Poverty in Canada and the United States

Previous

Contents

Next

Tables 1a and 1b illustrate 5 alternative sets of poverty lines for Canada and the US in 1994. For Canada, we first report the Low-Income Cutoffs (LICO's) for 1994 (1992 base), which vary with number of persons ( no distinction is made between adults and children) and size of area of residence. We next report 3 sets of poverty lines calculated as one-half median Canadian equivalent before-tax income⁶, but with median equivalent income calculated using three alternative equivalence scales. Each set of scales is widely used in the literature; as mentioned earlier, the 50% of median equivalent income approach to establishing a poverty line is the most widely used in the economics poverty literature. The first equivalence scale is that recommended by the Organization for Economic Cooperation and Development (OECD) (1982), and commonly in use in European countries. In this case, if a single adult is assigned an equivalence scale = 1.0; then each additional adult is assumed to add 0.7 to the scale; each additional child is assumed to add 0.5 to the scale. The 40/30' scale which is used for the Statistics Canada 'Low-Income Measures' (LIM's) follows the same procedure, but each additional adult is assumed to add 0.4 to the scale; each additional child is assumed to add 0.3. Finally, the 'Luxembourg Income Study' (LIS) equivalence scale is simply calculated as the square root of the number of people living in the household. As with the LICO's, no distinction is made between adults and children. The LIS scale is very popular in academic papers (see, for example, Crossley and Curtis, 2000; Osberg, 2000). Finally, we report the US official poverty lines converted to Canadian dollars.⁷

For the US, we report the official US poverty lines for 1994. We then repeat the exercise of calculating three alternative sets of relative poverty lines constructed using '40/30', OECD '70/50' and the LIS scales. However, in this case, the poverty lines are constructed as 50% of US median equivalent before-tax income. Finally, we also convert one set of Canadian relative poverty lines (the OECD 70/50) into 1994 US dollars using purchasing power parities for private final consumption (Statistics Canada, CANSIM D23283).⁸

For Canada, if we compare poverty lines for a family consisting of 2 adults and 2 children, alternatives include $27,100 (OECD), $25,983 (40/30), $26,523 (LIS) and from $31,071 (LICO, large urban area) to $21,472 (LICO, rural area). The US official poverty line is noticeably lower at $19,024. Differences across poverty lines are larger as family size gets larger. For example, for a 6-person household consisting of two adults and 4 children, poverty lines include $37,137 (OECD), $33,778 (40/30); $32,484 (LIS) and from $38,393 (LICO, large urban area) to $26, 533 (LICO, rural area).

For the US, there are also noticeable differences among poverty lines. First, the official US lines are much lower than any of the relative (50% of median income) lines --$15,029 for a family of 2 adults and 2 children versus $21,315 using the OECD 70:50 scale or $20,571 using the LIS scale, for example. If we convert the Canadian 50% of median income poverty line (OECD scale) into US funds, the line for a family of 4 is $21, 408, which is very similar to the OECD relative poverty line for the US. Again, differences among poverty lines are larger for larger family sizes as differences in the implicit 'economies of scale' become more important. In either country, then, it seems quite clear that some families with incomes in the range of 'poverty' could be classified as poor by one measure and 'not poor' by another.

However, our first concern in this section is to see whether estimates of poverty, using any particular poverty line, are the same if we use different data sets (i.e., the SCF versus the NLSCY in Canada; the CPS versus the NLSY in the United States). The point is that, arguably, the SCF contains the better estimates of family incomes,⁹ but the NLSCY is the only data set with rich information about the well-being of children. Thus, to study the links between child well-being and child poverty, we must use the NLSCY. However, if income is less well-measured in the NLSCY, this may be one reason for the weak links found between child health and child poverty in the Canadian literature thus far. The same arguments can be made about the CPS versus the NLSY for the US.¹⁰

Table 2 reports upon the incidence¹¹ of poverty among children aged 0 to 11 years in 1994 in Canada and the US, using the alternative poverty lines and data sets. We focus upon children aged 0 to 11 as this is the age range available in the NLSCY. The same calculation procedure is used for each data set in each country: 1) households are excluded if they do not contain any children aged 0 to 11 years or if they do not have positive income before tax; 2) children are counted as poor if they live in households with income less than the appropriate poverty line; 3) the incidence of poverty among children is computed by assigning each child aged 0 to 11 his or her appropriate sample weight.¹²

For both Canadian data sets and for the US CPS, we repeat all of these calculations for two samples of children: 1) all children aged 0 to 11 years; 2) all children aged 0 to 11 years whose pmk/mother is aged 29 to 37 years. Because the NLSY children's data is a supplement to the basic NLSY, all mothers are aged 29 to 37 years in 1994. Thus, the second set of calculations with the other 3 data sets allows for more accurate comparability across data sets, though the restricted mother age samples obviously give a less accurate picture of the incidence of poverty for children aged 0 to 11 years. One reason for presenting estimates for both samples, where available, is to investigate how important/limiting the age restriction on mothers is in the NLSY data, which has been used in several very influential studies of the link between poverty/income and children's well-being (e.g., Blau, 1999; Korenman, et al., 1995; Mayer, 1997).

The first striking point to take from the top panel of Table 2 is that the different data sources yield rather different estimates of the incidence of poverty. Regardless of the poverty line used, the NLSCY provides higher estimates of the incidence of poverty for children (0 to 11) in Canada than the SCF (3.7 percentage points higher in the case of the LICO's; 4.7 percentage points higher using a 50% of median equivalent income definition of poverty and the OECD equivalence scale). To put this in perspective, there is a 6.4 percentage point gap between the highest (20.9 in 1996) and lowest incidence of child poverty in Canada (14.5% in 1989) over the 1980 to 1997 period (National Council of Welfare, 1999, p 11).

The bottom panel of Table 2 provides a similar comparison for the US. Interestingly, in this case, the children's data set (the Mother-Child Survey of the NLSY) yields lower estimates of the incidence of child poverty than the Current Population Survey (CPS, as available through the Luxembourg Income Study). The consequence is that while the incidence of child poverty for all children 0 to 11 years is estimated to be about 10 percentage points higher in the US than in Canada (depending upon the poverty measure used), when SCF and CPS data are used, this gap is essentially eliminated. This point is true whether we compare the relative incidence of poverty across the countries (e.g., two of the 50% of median equivalent income definitions, using country-specific median equivalent income) or whether we make an 'absolute' comparison of the incidence of poverty by using, for example, the official US poverty lines for both the US and Canada. To give some specific numbers, consider a comparison of the incidence of poverty using the two 50% of median equivalent income poverty lines, with equivalent income calculated using a 'LIS' scale. If we use SCF and CPS data, the incidence of poverty in Canada is 21.2% while the incidence of poverty is estimated to be 30.2% for the US. If we use the same poverty lines, but the NLSCY and NLSY data sets, the incidence of poverty is estimated to be 24.0% in Canada and 24.9% in the United States.

Of course, as noted earlier, all mothers in the NLSY data set are aged 29 to 37 years, and it seems, based on the calculations for the other 3 data sets, that this age restriction results in considerably lower estimates of poverty (about 3 percentage points for either Canadian data sets; 4 to 5 percentage points using the CPS). However, the same qualitative points made above remain valid when we focus on the mother age-restricted samples for all the data sets: 1) the NLSCY produces higher estimates of the incidence of poverty than the SCF; 2) the NLSY produces lower estimates of the incidence of poverty than the CPS (though the two sets of US estimates are much closer once we make the mother age restriction).

Our understanding of the incidence of child poverty within countries can also be affected by choice of poverty line. For example, for Canada, there is a 2 percentage point gap between the LICO and the 50% of median income approach with an OECD equivalence scale, if NLSCY data are used (24.5 versus 26.5%). Note, on the other hand, that when SCF data are used, the incidence estimates are very close for all but the 'official US poverty lines.' The official US lines yield incidence estimates most different from the others -- using NLSCY data, child poverty is estimated to be only 14.4% versus 24.5% with the LICO's.

Differences across poverty lines in estimates of the incidence of poverty are similarly apparent for the US. For example, using the NLSY data and the official US poverty lines, the estimated incidence of poverty is 17.5% while with the OECD 50% of median income approach, the estimated incidence of poverty is 26.4%.

 

Table 1a 1994 Canadian Before Tax Poverty Lines

persons	Low-Income Cutoffs					½ the Median Equivalent Income
	Size of Area of Residence				Rural	OECD		40/30		LIS	US Official Poverty Lines (in Canadian $)
	Urban Areas
	500,000+	100,000 - 499,999	30,000 - 99,999	Less then 30,000		1 adult	2 adults	1 adult	2 adults		1 child	2 children
1	16,511	14,162	14,063	13,086	11,410	10,037	-	12,992	-	13,262	-	-
2	20,639	17,702	17,579	16,357	14,263	15,056	17,063	16,889	18,188	18,755	12,930	-
3	25,668	22,016	21,863	20,343	17,739	20,074	22,081	20,786	22,086	22,970	15,100	15,114
4	31,071	26,650	26,465	24,626	21,472	25,093	27,100	24,684	25,983	26,523	19,666	19,024
5	34,731	29,791	29,583	27,527	24,003	30,111	32,118	28,581	29,880	29,654	23,673	22,948
6	38,393	32,931	32,702	30,428	26,533	35,130	37,137	32,479	33,778	32,484	26,946	26,390
7	42,054	36,072	35,820	33,329	29,064	40,148	42,155	36,376	37,675	35,087	31,073	30,409
LICO's obtained from Statistics Canada 13-207 (base, 1992).

 

Table 1b 1994 United States Before Tax Poverty Lines

persons	US Official Lines number of kids <18			OECD		40/30		LIS	½ Median Canadian Income OECD Scale (US dollars)
	1	2	3	1 adult	2 adults	1 adult	2 adults		1 adult	2 adults
1	-	-	-	7,895	-	10,106	-	10,286	7,929	-
2	10,215	-	-	11,842	13,421	13,137	14,148	14,546	11,894	13,479
3	11,929	11,940	-	15,789	17,368	16,169	17,179	17,815	15,858	17,444
4	15,536	15,029	15,081	19,736	21,315	19,200	20,211	20,571	19,823	21,408
5	18,702	18,129	17,686	23,684	25,262	22,232	23,243	22,999	23,787	25,373
6	21,287	20,848	20,427	27,631	29,210	25,264	26,274	25,194	27,752	29,337
7	24,548	24,023	23,657	31,578	33,157	28,295	29,306	27,213	37,716	33,302
Source: US Official Lines: http://www.census.gov/hhes/poverty/threshld/thresh94.htm

 

Table 2 Poverty Among Children Aged 0-11 Years Alternative Poverty Lines and Data Sets Canada and the United States - 1994

Canada			Poverty Line = ½ the Median Equivalent Income
	LICO1		40/30 Scale		OECD2 Scale		LIS3 Scale		US Official Poverty Lines (in Can. $)
	SCF4	NLSCY5	SCF	NLSCY	SCF	NLSCY	SCF	NLSCY	SCF	NLSCY
% Poor
All	20.8%	24.5%	21.2%	24.3%	21.8%	26.5%	21.2%	24.0%	11.2%	14.4%
PMK 29-37	17.6%	21.5%	18.4%	21.3%	19.1%	24.0%	18.3%	20.7%	9.7%	12.3%
United States	½ Median Canadian  Income OECD Scale (US$)		40/30 Scale		OECD2 Scale		LIS3 Scale		US Official  Poverty Lines
	CPS6	NLSY7	CPS	NLSY	CPS	NLSY	CPS	NLSY	CPS	NLSY
% Poor
All	33.0%	-	30.7%	-	32.8%	-	30.2%	-	20.7%	-
PMK 29-37	28.8%	26.4%	26.6%	24.9%	28.7%	26.4%	26.0%	24.9%	17.3%	17.5%
1. Low-Income Cut-off. 2. Organization of Economic Cooperation and Development. 3. Luxembourg Income Study. 4. Survey of Consumer Finance. 5. National Longitudinal Survey of Children and Youth. 6. Current Population Survey. 7. National Longitudinal Survey of Youth.

• ⁶'Equivalent income' is family income divided by the appropriate household equivalence scale. The idea is to adjust family income for the fact that more than one individual depends upon this income, but to acknowledge the fact that there are economies of scale available to people who live together. Note that neither of the child outcome data sets provide after-tax income data, hence we have no option but to focus on before-tax income.
• ⁷US official poverty lines are available at http://www.census.gov/hhes/poverty/threshld/thresh94.html. Conversion to Canadian funds is made using purchasing power parities for private final consumption expenditures (CANSIM D23283).
• ⁸It is not obvious how to apply LICO's in the US context given the variation in poverty lines by size of area of residence. Would the same population thresholds (e.g., population greater than 500,000) define a 'large urban' centre for the US?
• ⁹The SCF asks detailed questions about individual components of income and the survey is conducted near tax time so that individuals will have the necessary information available to them. The NLSCY asks only two questions about personal and household income. The survey is not timed to coincide with tax time.
• ¹⁰Curtis, et al., 2001 employ the Ontario Child Health Study (OCHS) rather than the NLSCY. Family income is not particularly well-measured in this survey either.
• ¹¹ We also calculated depth of poverty for each measure, but focus here only upon incidence.
• ¹²In the NLSCY, each child is a separate observation and may have a different weight to produce appropriate age distributions of children, though the data contain siblings (i.e., children from the same family may have different sample weights). For the SCF, we create an observation for each child aged 0 to 11 years and assign each child the income of his or her household and the sample weight of the household.