Does Policy Affect Outcomes for Young Children? An Analysis with International Microdata - August 1999

7. National and Regional Differences in Policy, Macroeconomic Conditions and Social Context

Previous

Contents

Next

7.1 Basic Specification

In an attempt to test the hypothesis that policy can affect the well-being of children even after we have controlled for the influence of family-level characteristics, we pool microdata on outcomes and standard socioeconomic indicators for children from Canada, the U.S. and Norway and add to this information very general regional/national level policy indicators. We have also compiled and include in the analysis indicators of macroeconomic conditions and of social context. Tables 7 and 8 report means for supplementary information on policy, aggregated to the national and to the regional level, respectively, while Tables 9 and 10 record supplementary information on macroeconomic conditions and 'context.'²⁴

TABLE 9: Means - Macro/Context Variables (National Level)

	Unemployment rate	GDP per capita	Percent who believe social inequality is due to laziness	Percent of immigrants
Canada	10.4	22409	31.0	19.1
Norway	4.9	23165	10.8	4.8
United States	6.1	29655	37.5	10.6
Note: Dollar values are in 1994 Canadian dollars, using purchasing power parity conversion factors.
PPP Sources: OECD, 1998. National Accounts. Main Aggregates. Volume 1. 1960-1996.
OECD, 1990. Purchasing Power Parities and Real Expenditures. EKS Results. Volume 1.

The most general indicator of policy we use is 'social spending per capita,' reported in 1994 Canadian dollars for all countries (social spending includes health, but not national defence).²⁵ Norway spends by far the most, at $1,682 per person (1994 Canadian dollars, converted via PPP's) while Canada spends $939 (56 percent of what Norway spends per person) and the U.S. spends only $653 (39 percent of the Norwegian equivalent).²⁶ However, even within Canada, there is significant variation in social spending, from a high of $1,182 per capita in Quebec, to a low of $604 per person in PEI (PEI spending is thus only 51 percent of Quebec's spending per person).

TABLE 10: Means - Macro/Context Variables (Regional Level)

	Unemployment rate	GDP per Capita	Percent who believe social inequality is due to laziness	Percent of immigrants
Newfoundland	20.4	14647	32.3	1.7
Prince Edward Island	17.2	15872	62.5	3.6
Nova Scotia	13.3	16982	27.2	5.9
New Brunswick	12.5	17650	20.0	3.4
Quebec	12.2	20173	24.7	11.3
Ontario	9.6	23795	35.0	27.4
Manitoba	9.2	19514	31.3	14.8
Saskatchewan	7.0	20018	45.4	7.3
Alberta	8.6	27812	32.8	17.4
British Columbia	9.4	23674	25.6	24.4
New England	5.9	31198	30.1	11.0
Mid-Atlantic	6.7	30951	37.9	14.6
East-North Central	4.3	26030	39.4	3.2
West-North Central	5.5	26399	33.3	5.2
South Atlantic	5.7	25863	40.7	8.4
East-South Central	5.6	22530	43.0	1.4
West-South Central	6.5	25982	38.7	10.1
Mountain	5.3	23727	40.0	8.7
Pacific	8.0	30016	31.7	22.8
Note: Dollar values are in 1994 Canadian dollars, using purchasing power parity conversion factors.
PPP Sources: OECD, 1998. National Accounts. Main Aggregates. Volume 1. 1960-1996.
OECD, 1990. Purchasing Power Parities and Real Expenditures. EKS Results. Volume 1.

Our most basic indicator of macroeconomic conditions is the unemployment rate which is much the highest in Canada (10.4 percent) and much the lowest in Norway (4.9 percent). Here again, there is significant variation across regions, within countries, especially for Canada. For example, Saskatchewan's unemployment rate (7.0 percent) is only 34 percent of Newfoundland's unemployment rate (20.4 percent).

Finally, our most basic indicator of social context is the percentage of heads of household who are immigrants. If it is easier to solve social problems in a country/region with an extremely homogeneous population, then variation across the countries studied here (and across regions within the countries) will be an important contextual factor in explaining children's well-being. Notice that this is different from controlling for whether or not the child is from a family with this or that ethnic background. The point is that if society is more ethnically mixed, it may have an impact on social attitudes/social policy. While we do not have data allowing U.S. to control for ethnicity, we were able to obtain information about immigrant status of head (using the Luxembourg Income Study). Immigrant status and ethnicity are obviously not the same thing, though there is presumably a correlation between the concepts. From Tables 9 and 10 it is clear that Norway has far fewer immigrants (4.9 percent) than the U.S. (10.6 percent), or especially Canada (19.1). On the other hand, some regions within Canada or the U.S. have lower rates of immigration than Norway (e.g., Newfoundland, PEI and New Brunswick).

Table 11 reports the results of a regression analysis which adds these three variables to the basic specification discussed earlier (with regional dummy variables removed).²⁷ The first point to notice is that higher social spending per capita is associated with better outcomes for children in 4 of 6 cases. The exceptions are restless/overly active behaviour and disobedience at school which appear to increase as social spending increases. Notice that since social spending per capita includes spending on social transfers which is also included in family gross income, a micro-level variable, the finding of a statistically significant effect of the average level of social spending in the region on children's well-being means that we are 'double counting' this aspect of government spending.²⁸ If anything, this should reduce the role we find for either variable, yet we still find support for the hypothesis that higher levels of social spending are associated with better child outcomes.

TABLE 11: Probit Analysis of the Probability of Alternative Child Outcomes Including Basic Macro, Policy and Context Variables

	Asthma	Injury	Limited in activity	Anxiety/Fear	Restless/Overly active	Disobedient at school
	Ages 4-11	Ages 0-11	Ages 0-11	Ages 4-11	Ages 4-11	Ages 6-11
Intercept	-0.84* (0.17)	-1.35* (0.07)	-1.66* (0.09)	-0.19* (0.07)	-0.23* (0.07)	-0.73* (0.10)
Dummy=1 if mother smokes daily	0.09* (0.03)	0.10* (0.02)	0.09* (0.03)	0.01 (0.02)	0.23* (0.02)	0.16* (0.03)
Dummy=1 if mother smokes occasionally	-0.03 (0.06)	0.03 (0.05)	-0.08 (0.07)	-0.02 (0.04)	-0.01 (0.04)	0.02 (0.06)
Dummy=1 if poor family	-0.10** (0.04)	-0.11* (0.03)	0.01 (0.04)	0.09* (0.03)	0.05 (0.03)	0.09** (0.04)
Dummy=1 if mother was less than 25 at child's birth	0.06** (0.03)	0.07* (0.03)	0.06*** (0.03)	0.06* (0.02)	0.04 (0.02)	0.07** (0.03)
Dummy=1 if child is aged 8 - 11	0.03 (0.03)	0.14* (0.02)	0.25* (0.03)	0.09* (0.02)	-0.22* (0.02)	0.18* (0.03)
Dummy=1 if child has one sibling	-0.09** (0.04)	0.08* (0.03)	-0.12* (0.04)	-0.14* (0.03)	-0.17* (0.03)	-0.15* (0.04)
Dummy=1 if child has two siblings	-0.16* (0.04)	0.07** (0.03)	-0.01 (0.04)	-0.24* (0.03)	-0.26* (0.03)	-0.19* (0.05)
Dummy=1 if child has three or more siblings	-0.35* (0.06)	0.12* (0.04)	-0.09 (0.06)	-0.41* (0.04)	-0.34* (0.04)	-0.26* (0.05)
Dummy=1 if lone mother	0.18* (0.04)	0.13* (0.03)	0.15* (0.04)	0.12* (0.03)	0.15* (0.03)	0.27* (0.04)
Dummy=1 if female child	-0.24* (0.03)	-0.17* (0.02)	-0.17* (0.03)	0.02 (0.02)	-0.36* (0.02)	-0.61* (0.03)
Equivalent income	7.47E-7 (1.06E-6)	1.14E-6 (7.83E-7)	-2.67E-6** (1.30E-6)	-3.86E-6* (8.15E-7)	-5.43E-6* (8.00E-7)	-2.36E-6** (1.19E-6)
Unemployment	0.02* (0.01)	0.01 (0.00)	0.00 (0.01)	0.04* (0.00)	0.02* (0.01)	-0.04* (0.01)
Social spending per capita	-3.2E-4* (0.9E-4)	-1.0E-4** (4.6E-5)	-2.4E-4* (0.7E-4)	-5.6E-4* (4.5E-5)	5.27E-4* (8.4E-5)	3.7E-4* (1.2E-4)
% of immigrants as heads of households	-0.00 (0.00)	0.00 (0.00)	4.48E-3* (1.66E-3)	5.57E-3* (1.11E-3)	7.71E-3* (1.10E-3)	0.00 (0.00)
Note: Standard errors are in parentheses.
* indicates significance at the 99% level
** indicates significance at the 95% level
*** indicates significance at the 90% level

The general macroeconomic climate, as proxied by the unemployment rate, might also be expected to influence child well-being insofar as it increases economic insecurity for families. And, we do find that higher rates of unemployment are associated with higher levels of anxiety among young children, in a higher incidence of asthma and in more restless/overly active behaviour (3 of 6 outcomes in this specification). Notice that the regression is not measuring whether or not the child's family experienced unemployment. As argued above, a higher rate of unemployment in the region, at any given level of family income, could increase the stress experienced by families who have not yet experienced any unemployment insofar as it increases their economic insecurity and thus have negative implications for child well-being. [Experience of accidents/injury²⁹ or activity limitations are not significantly affected by the regional unemployment rate in this specification. Disobedience at school, surprisingly, falls as unemployment increases.]

Finally, we find that children who live in regions/countries with a higher percentage of immigrants are more likely to be anxious/fearful, to experience activity limitations and to be restless/overly active.

7.2 Unpacking Social Spending — Health, Education and Social Transfers

It is possible to disaggregate 'social spending' somewhat into policy components which may be particularly relevant for children's well-being.³⁰ Tables 7 and 8 report the average level of social transfers received by families with children aged 0 to 11,³¹ student/teacher ratios, physicians per 100,000 population and the percentage of health care expenditures which are public.

First, average social transfers per child (averaging in those who receive and those who do not) are much higher in Norway ($7,162) than in Canada ($6,133) or the U.S. ($3,482). Yet, the difference in levels of spending on social transfers is smaller than the difference in levels of social spending over-all. For example, average per child social transfers in Canada are 86 percent of the Norwegian level, while average per capita social spending in Canada is only 55 percent of the Norwegian level. Average per child social transfers in the U.S. are 49 percent of the Norwegian equivalent while overall U.S. social spending per capita is only 39 percent of the Norwegian value.

TABLE 7: Means - Policy Variables (National Level)

	Social spending per capita	Average social transfers per child*	Students per teachers	Physicians per 100,000	Percentage of health care publically funded	Percentage of children in top quintile receiving social transfers*
Canada	939	6133	16.2	186.7	72.1	67.9
Norway	1682	7162	15.0	309.0	96.6	99.5
United States	653	3482	17.3	225.0	44.6	14.5
Note: Dollar values are in 1994 Canadian dollars, using purchasing power parity conversion factors.
PPP Sources: OECD, 1998. National Accounts. Main Aggregates. Volume 1. 1960-1996. OECD, 1990. Purchasing Power Parities and Real Expenditures. EKS Results. Volume 1.
* Source: Author's calculations using the Luxembourg Income Study

There is also substantial variation within countries in average levels of social transfers (see Table 8). For example, average per child transfers received in Alberta ($4,012) are only 63 percent of what is received in Quebec ($6,332) or 47 percent of what is received in PEI ($8,470). Also, once we look at averages for regions, it is clear that transfer levels for some provinces are higher than for Norway (e.g., Newfoundland and PEI). And, average transfers for some provinces are lower than for some U.S. states (in fact, average transfers received in Alberta are lower than in any U.S. state). Of course, much of the difference in transfer receipts reflects differences in economic conditions in these regions.³²

TABLE 8: Means - Policy Variables (Regional Level)

	Social spending per capita	Average social transfers per child*	Students per teachers	Physicians per 100,000	Percentage of health care publically funded	Percentage of children in top quintile receiving social transfers*
Newfoundland	991	8337	14.7	167.5	76.2	73
Prince Edward Island	604	8994	17.4	131.2	69.5	82.5
Nova Scotia	1059	6281	17.4	189.8	71.5	63.7
New Brunswick	892	7045	17.3	141.6	71.0	66.4
Quebec	1182	6987	14.7	205.3	73.2	90.5
Ontario	812	6156	15.9	186.9	70.0	60.2
Manitoba	1021	5814	15.2	174.8	75.3	74.6
Saskatchewan	814	5338	17.4	152.4	75.2	75.5
Alberta	888	4337	18.5	166.9	71.6	45.9
British Columbia	911	5854	17.3	194.9	74.0	46.9
New England	n/a	3007	14.6	269	44.6	11.3
Mid-Atlantic	n/a	4109	15.5	284	44.6	13.7
East-North Central	n/a	2762	15.9	188	44.6	18.5
West-North Central	n/a	4083	17.6	200	44.6	10.8
South Atlantic	n/a	3109	16.9	264	44.6	11.5
East-South Central	n/a	3479	17.7	173	44.6	15
West-South Central	n/a	3966	16.0	173	44.6	17.2
Mountain	n/a	2768	19.1	172	44.6	13.5
Pacific	n/a	4540	22.9	211	44.6	19.4
Note: Dollar values are in 1994 Canadian dollars, using purchasing power parity conversion factors.
PPP Sources: OECD, 1998. National Accounts. Main Aggregates. Volume 1. 1960-1996. OECD, 1990. Purchasing Power Parities and Real Expenditures. EKS Results. Volume 1.
* Source: Author's calculations using the Luxembourg Income Study

Thus, there is substantial variation across countries and regions in terms of spending on social transfers. However, this is less than the variation observed in over-all social spending. Other forms of government expenditure must therefore partially account for observed differences. One possibility which is important for children is that countries differ in terms of investment in education (though this would presumably only affect school-aged children).³³ Tables 7 and 8 report student to teacher ratios (for elementary and secondary schools). Norway has the smallest number of students to teachers (15) while the U.S. has the largest number (17.3). And, there is variation across regions within Canada and the U.S. However, there is much less variation in these ratios than is apparent in terms, for example, of social transfers.

Investment in health care is another form of social spending which might be expected to influence the well-being of children. Tables 7 and 8 report physicians per 100,000 population. In terms of this measure, Norway yet again dominates Canada or the U.S., with 309 physicians per 100,000 population. The U.S. ranks second according to this measure (225) and Canada ranks third (187). However, there is, once again, significant variation across regions (e.g., from 205 in Quebec to 152 in Saskatchewan or from 172 in Mountain to 284 in Mid-Atlantic), so that some Canadian provinces have more physicians per 100,000 population than some U.S. regions. However, no Canadian provinces or U.S. region can match the number of physicians per 100,000 population of Norway.

Physicians per 100,000 population can be viewed as a measure of the over-all level of spending on health care. But, it might also matter how dollars are spent. Another dimension of health care policy which differs among these 3 countries is the extent to which health care is publicly provided. Tables 7 and 8 indicate that while 96.6 percent of healthcare is public in Norway, only 44.6 percent is public in the U.S.³⁴ Canada is, as usual, in between these extremes at 72.1 percent and public provision differs somewhat across provinces (e.g., from 75 percent in Manitoba or Saskatchewan to 70 percent in Ontario).

Table 12 reports probit results for the 6 child outcomes with social spending per capita replaced by the 4 somewhat disaggregated policy variables (social transfers per child, student to teacher ratios, physicians per 100,000 population and percent of health care publicly funded). Before discussing results for these variables, notice that results for unemployment have changed slightly. Higher rates of unemployment are now associated with more accidents, more activity limitation, more fear/anxiety, and more restless/overly active behaviour (4 of 6 outcomes). (Asthma is no longer significantly associated with unemployment, the unexpected result for disobedience at school has disappeared.³⁵)

TABLE 12: Probit Analysis of the Probability of Alternative Child Outcomes Including Additional Policy, Macro and Context Variables

	Asthma	Injury	Limited in         activity	Anxiety/Fear	Restless/Overly active	Disobedient at school
	Age 4-11	Ages 0-11	Ages 0-11	Ages 4-11	Ages 4-11	Ages 4-11
Intercept	-0.87 (0.60)	-1.86* (0.18)	-1.91* (0.27)	0.85* (0.17)	0.29 (0.23)	0.24 (0.41)
Dummy=1 if mother smokes daily	0.09* (0.03)	0.10* (0.02)	0.09* (0.03)	0.01 (0.02)	0.23* (0.02)	0.16* (0.03)
Dummy=1 if mother smokes occasionally	-0.03 (0.06)	0.03 (0.05)	-0.08 (0.07)	-0.02* (0.04)	-0.01 (0.04)	0.02 (0.06)
Dummy=1 if poor family	-0.10** (0.04)	-0.11* (0.03)	0.02 (0.04)	0.09* (0.03)	0.06*** (0.03)	0.08** (0.04)
Dummy=1 if mother was less than 25 at child's birth	0.06** (0.03)	0.07* (0.03)	0.05 (0.03)	0.07* (0.02)	0.04*** (0.02)	0.06** (0.03)
Dummy=1 if child is aged 8 - 11	0.03 (0.03)	0.14* (0.02)	0.25* (0.03)	0.09* (0.02)	-0.22* (0.02)	0.18* (0.03)
Dummy=1 if child has one sibling	-0.09** (0.04)	0.07* (0.03)	-0.12* (0.04)	-0.14* (0.03)	-0.18* (0.03)	-0.15* (0.04)
Dummy=1 if child has two siblings	-0.16* (0.04)	0.07** (0.03)	-0.00 (0.04)	-0.24* (0.03)	-0.27* (0.03)	-0.20* (0.05)
Dummy=1 if child has three or more siblings	-0.35* (0.06)	0.11* (0.04)	-0.09*** (0.06)	-0.41* (0.04)	-0.35* (0.04)	-0.27* (0.05)
Dummy=1 if lone mother	0.18* (0.04)	0.12* (0.03)	0.16* (0.04)	0.12* (0.03)	0.18* (0.03)	0.27* (0.04)
Dummy=1 if female child	-0.24* (0.03)	-0.17* (0.02)	-0.17* (0.03)	0.02 (0.02)	-0.36* (0.02)	-0.61* (0.03)
Equivalent income	7.16E-7 (1.08E-6)	1.01E-6 (7.88E-7)	-2.46E-6*** (1.30E-6)	-3.68 E-6* (8.18 E-7)	-4.82E-6* (8.00E-7)	-2.52E-6** (1.19E-6)
Unemployment	0.008 (0.01)	0.02** (0.01)	0.04* (0.01)	0.03* (0.01)	0.02*** (0.01)	0.00 (0.01)
Average social transfers	5.4E-5 (3.8E-5)	-3.0E-5** (2.2E-5)	-1.7E-4* (3.0E-5)	-4.0E-5*** (2.0E-5)	-5.0E-5** (2.4E-5)	-0.00 (0.00)
Student/Teacher ratio	0.05** (0.02)	0.02* (0.01)	-0.00 (0.01)	-0.02* (0.01)	-0.02** (0.01)	-0.03** (0.01)
Physicians per 100,000	0.001 (0.001)	5.84E-4 (3.57E-4)	0.00 (0.00)	-0.003* (0.000)	-1.8E-3* (4.4E-4)	-0.00 (0.00)
% of public expenditures on health care	-0.02* (0.01)	-0.00 (0.00)	9.41E-3* (2.1E-3)	-0.01* (0.00)	0.02* (1.75E-3)	-0.00 (0.00)
% of immigrants as heads of households	-0.005 *** (0.003)	0.002 (0.001)	7.72E-3* (1.79E-3)	0.01* (0.00)	0.00 (0.00)	0.00 (0.00)
Note: Standard errors are in parentheses.
* significant at the 99% level
** significant at the 95% level
*** significant at the 90% level

What of the social policy variables? Higher average social transfers per child are associated with fewer accidents, less activity limitation, less fear/anxiety and less restless/overly active behaviour³⁶ (4 of 6 outcomes). A higher student to teacher ratio is associated with more accidents and more asthma (though there is no relationship for the 6 to 11 year old sample). Unexpectedly, higher student to teacher ratios are associated with less anxiety/fear, less restless behaviour and less disobedience at school (though the disobedience at school result disappears for the 6 to 11 year old sample). These results for student/teacher ratios are very mixed. Perhaps there is a better measure of 'social investment' in schooling. It also seems likely that cognitive measures would be more strongly influenced by investments in education, but none were available on a comparable basis for all 3 countries.

More physicians per 100,000 population are associated only with less anxiety/fear and less restless behaviour. Percentage of health care publicly funded is statistically significant more often (for 4 of 6 outcomes) — a higher proportion of public funding is associated with less asthma, fewer activity limitations, less anxiety/fear and, surprisingly, more restless behaviour. In general, these signs meet with prior expectations. A general point is that the percentage of healthcare funding which is private is a more important determinant of child well-being than the over-all level of healthcare quality, at least as proxied by the physicians per 100,000 variable.

7.3 Removing the Effects of Unemployment from Average Social Transfer Receipt

A concern raised in the preceding section is that average social transfer levels will be higher in regions experiencing difficult economic times. Thus, for example, although a descriptive survey of programmes available in Norway gives the impression that a much more generous system of transfers is available in that country than in Canada, the average level of social transfers received by children in Newfoundland is actually higher than the average level of transfers received in Norway. Given that the unemployment rate in Newfoundland is 20.4 while the unemployment rate in Norway is 4.9 percent, there is almost certainly a link and it may be leading us to conclude that transfers are less important this is really the case.³⁷ The dilemma is to find a measure of the generosity of the social transfer structure which is distinct from current use of the system.

We have two approaches to solving this problem. The first involves using microdata from the Luxembourg Income Study to predict social transfer receipt for children, conditional upon family unemployment status. The second involves the substitution of a proxy measure — societal attitudes toward those who live in need derived from the World Values Study.

Table 13 reports OLS estimates of the level of social transfers received in each of the 3 countries, where the estimating sample includes all children, regardless of whether or not they received a transfer. This is consistent with the approach used to calculate average expenditures on social transfers per child in the previous section. Again, it is a blend of the probability of receipt and the level of receipt in the event of a positive transfer.³⁸ Control variables for this estimation include child's age, mother's age, family market income, a dummy indicating lone mother status, a dummy indicating that either the head or spouse of head experienced unemployment during the survey year³⁹ and regional dummies (for Canada and the U.S.).

Notice that the intercept term for the Canadian equation is $3,972 (Cdn) versus $5,544 (Cdn) for Norway. The U.S. intercept is not statistically different from zero. Children living in lone-mother families in Canada receive $2,577 more than others, $3,609 more than others in Norway and $1,939 more in the U.S. Transfers increase more quickly with additional siblings in Norway than in Canada or the U.S. Having someone in the household who is unemployed increases social transfers by a larger amount in Canada than in the other countries. (Unfortunately, we are unable to control for duration of unemployment, so this result is probably reflecting longer durations in Canada, where unemployment rates are much higher.)

While interesting in themselves, these regressions are used here only in an attempt to 'purge' unemployment effects from average social transfers. To do this, we predict, for each region, what the level of social transfers would be for a child living with two parents,⁴⁰ and all other characteristics at the mean level for Ontario (i.e., child's age, mother's age, market income, number of siblings and household unemployment). The thought experiment conducted is: suppose we took a 'representative child,' with the same family characteristics, including probability of unemployment, how much more or less would he/she receive in the form of social transfers if living in Texas or California or Norway than he/she currently receives in Ontario? Across regions within Canada, the difference in predicted transfers will be entirely due to the intercept shift estimated for the benefit receipt equation. Across countries, predicted social transfers will differ both as a result of different intercept terms and different coefficients associated with any particular characteristic.

TABLE 13: Ordinary Least Squares Level of Social Transfers

	Canada	Norway	United States
Variable	All Children	All Children	All Children	Variable
Intercept	3972.15* (249.34)	5544.94* (469.02)	-148.92 (298.40)	Intercept
Child's age	-234.47* (14.25)	26.21 (25.53)	-55.78* (18.12)	Child's age
Mother's age	128.98* (7.31)	21.66 (15.42)	181.53* (6.37)	Mother's age
Market income	-0.08* (0.00)	-0.06* (0.00)	-0.05* (0.00)	Market income
Dummy=1 if lone mother	2577.09* (138.01)	3609.90* (232.95)	1939.65* (133.08)	Dummy=1 if lone mother
Number of siblings	1579.69* (43.78)	1907.52* (81.35)	1149.56* (43.32)	Number of siblings
Dummy=1 if a person in the household is unemployed	3811.44* (102.88)	3149.52* (185.46)	1025.45* (141.62)	Dummy=1 if a person in the household is unemployed
Newfoundland	367.52 (330.83)	-	-6.81 (355.53)	New England
Prince Edward Island	155.39 (624.78)	-	205.43 (218.84)	East North Central
Nova Scotia	-1142.56* (263.79)	-	-1615.55* (291.53)	West North Central
New Brunswick	-644.63** (292.96)	-	-1519.37* (220.75)	South Atlantic
Quebec	74.35 (115.59)	-	-1493.88* (281.16)	East South Central
Manitoba	-1667.80* (235.76)	-	-1216.61* (230.11)	West South Central
Saskatchewan	-2152.32* (236.42)	-	-2348.02* (292.72)	Mountain
Alberta	-2577.38* (157.85)	-	-54.05 (212.40)	Pacific
British Columbia	-783.63* (149.19)	-	-
Adjusted R2	0.3338	0.3572	0.1483	Adjusted R2
Number of observations	16976	3838	28096	Number of observations

'Purged' social transfers means are less correlated with regional unemployment (+0.58) than are actual social transfer means (+0.70). However, as a comparison of Tables 12 and 14 indicates, there is relatively little impact on the final probit regression results if we replace actual transfers with 'purged' transfers. Results are certainly not 'improved.' We lose statistical significance for anxiety/fear, social transfers become statistically significant for disobedience at school (though with an unexpected negative sign), there is a sign change in the accidents/injuries equation (though this is not true for the unweighted regressions). Certainly, when we use estimated equations to predict 'purged' transfers, rather than using actual transfers received, information is lost (adjusted r-squared from the estimated equations are not particularly high). Perhaps this loss of information explains the relatively poor results obtained with this procedure.

Our second attempt to find a measure of social transfer generosity which does not depend upon current economic conditions, involves the use of an attitudinal proxy. While it is not necessarily true that social attitudes influence the social transfer system which is in place, it seems reasonable that there would be a connection between the two. The World Values Study asks respondents in many countries the question: "Why do you think people live in need?" Table 14 also reports means by country and by region of the percentage of people who answered this questions 'because they are lazy.' It seems likely that regions in which a relatively high percentage of the population believe that people live in need because they are lazy, rather than because, for example, of social injustice (another possible response), would be less willing to support generous social support systems.

It is clear from Table 9 that there are striking differences across countries and regions in the percentage of the population who believe that economic need is the results of laziness.⁴¹ For example, 31 percent of Canadian respondents, 37.5 percent of U.S. respondents, but only 11 percent of Norwegian respondents answered that economic need is the result of laziness. Within countries, there is also substantial variation: 35 percent of people in Ontario believe laziness is the problem versus only 25 percent in Quebec or BC.

TABLE 14: Probit Analysis of the Probability of Alternative Child Outcomes Including 'Purged' Transfers for Couples

	Asthma	Injury	Limited in        activity	Anxiety/Fear	Restless/Overly active	Disobedient at school
	Ages 4-11	Ages 0-11	Ages 0-11	Ages 4-11	Ages 4-11	Ages 4-11
Intercept	-0.81 (0.61)	-1.73* (0.19)	-1.53* (0.28)	0.89* (0.18)	0.67** (0.27)	-0.55 (0.37)
Dummy=1 if mother smokes daily	0.09* (0.03)	0.10* (0.02)	0.09* (0.03)	0.01 (0.02)	0.23* (0.02)	0.16* (0.03)
Dummy=1 if mother smokes occasionally	-0.03 (0.06)	0.03 (0.05)	-0.08 (0.07)	-0.02 (0.04)	-0.01 (0.04)	0.02 (0.06)
Dummy=1 if poor family	-0.10** (0.04)	-0.11* (0.03)	0.01 (0.04)	0.09* (0.03)	0.05*** (0.03)	0.09** (0.04)
Dummy=1 if Mother was less than 25 at child's birth	0.06** (0.03)	0.06* (0.03)	0.05 (0.03)	0.07* (0.02)	0.04 (0.02)	0.07** (0.03)
Dummy=1 if child is aged 8 -11	0.03 (0.03)	0.14* (0.02)	0.25* (0.03)	0.09* (0.02)	-0.22* (0.02)	0.18* (0.03)
Dummy=1 if child has one sibling	-0.09** (0.04)	0.07* (0.03)	-0.12* (0.04)	-0.14* (0.03)	-0.18* (0.03)	-0.15* (0.04)
Dummy=1 if child has two siblings	-0.16* (0.05)	0.07** (0.03)	-0.01 (0.04)	-0.24* (0.03)	-0.27* (0.03)	-0.19* (0.05)
Dummy=1 if child has three or more siblings	-0.35* (0.06)	0.11* (0.04)	-0.09 (0.06)	-0.41* (0.04)	-0.35* (0.04)	-0.26* (0.05)
Dummy=1 if single mother	0.18* (0.04)	0.13* (0.03)	0.16* (0.04)	0.12* (0.03)	0.18* (0.03)	0.27* (0.04)
Dummy=1 if Female child	-0.24* (0.03)	-0.17* (0.02)	-0.17* (0.03)	0.02 (0.02)	-0.36* (0.02)	-0.61* (0.03)
Equivalent income	6.84E-7 (1.08E-6)	1.00E-6 (7.88E-7)	-2.38E-6*** (1.30E-6)	-3.66E-6* (8.18E-7)	4.78E-6* (7.99E-7)	-2.45E-6** (1.19E-6)
Unemployment	0.01 (0.01)	0.02* (0.01)	0.02*** (0.01)	0.02* (0.01)	0.02** (0.01)	-0.03* (0.01)
Purged social transfers	3.7E-5 (3.9E-5)	5.0E-5* (2.0E-5)	-1.3E-4* (2.8E-5)	-6.82E-6 (1.90E-5)	-8.0E-5* (2.3E-5)	5.5E-5*** (3.1E-5)
Student teacher Ratio	0.05** (0.02)	0.01 (0.01)	-0.02 (0.01)	-0.02** (0.01)	-0.04* (0.01)	9.02E-4 (0.01)
Physicians per 100,000	7.18E-4 (1.20E-3)	6.14E-4*** (3.42E-4)	-3.5E-4 (4.9E-4)	-3.34E-3* (3.27E-4)	-2.12E-3* (4.59E-3)	1.9E-5 (6.22E-4)
% of immigrants as heads of households	-5.81E-3*** (3.28E-3)	3.71E-3** (1.50E-3)	0.01* (0.00)	9.92E-3* (1.41E-3)	4.85E-3** (2.00E-3)	-2.55E-3 (2.72E-3)
% of public expenditures on health care	-0.02** (0.01)	9.89E-4 (1.35E-3)	6.35E-3* (1.88E-3)	-7.05E-3* (1.27E-3)	0.02* (0.00)	-1.61E-3 (2.31E-3)
Note: Standard error s are in parenthesis.
* significant at the 99% level
** significant at the 95% level
*** significant at the 90% level

When the 'lazy' variable is substituted for the social transfer variable, Table 15 again indicates that it is not a very helpful proxy.⁴² In this specification, only 2 of the 6 children's outcomes studied are significantly affected: activity limitation and restless/overly active behaviour both increase with an increase in the percentage of the population responding that economic need is the result of laziness. [However, it is interesting to note that if the 'lazy' variable is the only one added to the traditional set of micro-determinants, then it is statistically significant, and has the expected sign, for 5 of the 6 outcomes studied.]

To conclude this section, it seems preferable not to replace the measure of actual transfers received with either of the proxies considered here. As argued above, it seems likely that, if anything, the unemployment 'contamination' would mean that we are under-stating the true impact of social transfers on children's well-being. Since we find actual transfers to play a larger role than either of the proxies, there is little point in replacing it.

TABLE 15: Probit Analysis of the Probability of Alternative Child Outcomes Including '% Who Believe Social Inequality is Due to Laziness'

	Asthma	Injury	Limited in activity	Anxiety/Fear	Restless/Overly active	Disobedient at school
	Ages 4-11	Ages 0-11	Ages 0-11	Ages 4-11	Ages 4-11	Ages 4-11
Intercept	-0.81 (0.63)	-1.86* (0.28)	-2.85* (0.38)	0.87* (0.26)	-0.27 (0.28)	-0.02 (0.38)
Dummy=1 if mother smokes daily	0.09* (0.03)	0.10* (0.02)	0.09* (0.03)	0.01 (0.02)	0.23* (0.02)	0.16* (0.03)
Dummy=1 if mother smokes occasionally	-0.03 (0.06)	0.03 (0.05)	-0.08 (0.07)	-0.02 (0.04)	-3.89E-3 (0.04)	0.02 (0.06)
Dummy=1 if poor family	-0.10** (0.04)	-0.11* (0.03)	0.01 (0.04)	0.09* (0.03)	0.05*** (0.03)	0.09** (0.04)
Dummy=1 if mother was less than 25 at child's birth	0.06** (0.03)	0.07* (0.03)	0.05 (0.03)	0.07* (0.02)	0.04*** (0.02)	0.07** (0.03)
Dummy=1 if child is Aged 8 -11	0.03 (0.03)	0.14* (0.02)	0.25* (0.03)	0.09* (0.02)	-0.22* (0.02)	0.18* (0.03)
Dummy=1 if child has one sibling	-0.09** (0.04)	0.08* (0.03)	-0.12* (0.04)	-0.14* (0.03)	-0.18* (0.03)	-0.15* (0.04)
Dummy=1 if child has two siblings	-0.16* (0.04)	0.07** (0.03)	-4.94E-3 (0.04)	-0.24* (0.03)	-0.27* (0.03)	-0.20* (0.05)
Dummy=1 if child has three or more siblings	-0.35* (0.06)	0.11* (0.04)	-0.09 (0.06)	-0.41* (0.04)	-0.35* (0.04)	-0.26* (0.05)
Dummy=1 if single mom	0.18* (0.04)	0.12* (0.03)	0.16* (0.04)	0.12* (0.03)	0.18* (0.03)	0.27* (0.04)
Dummy=1 if Female child	-0.24* (0.03)	-0.17* (0.02)	-0.17* (0.03)	0.02 (0.02)	-0.36* (0.02)	-0.61* (0.03)
Equivalent income	6.47E-7 (1.08E-6)	1.04E-6 (7.88E-7)	-2.3E-6*** (1.29E-6)	-3.65E-6* (8.18E-7)	-4.73E-6* (7.99E-7)	-2.46E-6** (1.19E-6)
Unemployment	0.02** (0.01)	0.01** (0.00)	1.5E-4 (6.62E-3)	0.02* (0.00)	1.63E-3 (5.77E-3)	-0.02** (7.95E-3)
% who believe social inequality is due to laziness	2.28E-3 (3.18E-3)	3.01E-4 (2.26E-3)	0.01* (2.99E-3)	3.6E-5 (2.09E-3)	4.74E-3** (2.09E-3)	-1.94E-3 (2.86E-3)
Student teacher ratio	0.04** (0.02)	0.02* (7.26E-3)	0.02** (0.01)	-0.02* (0.01)	-8.07E-3 (6.99E-3)	-0.02*** (9.44E-3)
Physicians per 100,000	1.71E-3 (1.39E-3)	4.23E-4 (3.95E-4)	9.6E-5 (5.63E-4)	-3.37E-3* (3.77E-4)	-1.29E-3* (4.74E-4)	-4.3E-4 (6.38E-4)
% of immigrants as heads of households	-4.91E-3*** (2.97E-3)	1.53E-3 (1.26E-3)	3.61E-3** (1.70E-3)	9.61E-3* (1.17E-3)	-1.0E-3 (1.36E-3)	1.31E-3 (1.85E-3)
% of public expenditures on health care	-0.02** (0.01)	-1.30E-3 (1.24E-3)	3.92E-3** (1.70E-3)	-7.35E-3* (1.19E-3)	0.02* (1.82E-3)	-1.97E-3 (2.45E-3)
Note: Standard error are in parenthesis.
* significant at the 99% level
** significant at the 95% level
*** significant at the 90% level

7.4 What about the Structure as Opposed to the Level of Benefits?

Not only are average levels of spending on social transfers very different across the countries, but so are the structures of social transfer systems. For example, in Norway, every child receives social transfers. In the U.S., 55 percent receive transfers while in Canada, 73 percent receive transfers. This difference in structure is further illustrated by comparing social transfer receipt for children with family income in different quintiles of the income distribution (see Table 16). In Norway, basically 100 percent of children in the top quintile receive benefits, whereas in the U.S., fewer than 20 percent do. In Canada, a majority of children still receive some transfers, even in the top quintile of the income distribution. In all 3 countries, virtually all children with family incomes in the bottom income quintile receive transfers but already in the second quintile, the proportion receiving transfers drops in the U.S. In Canada, on the other hand, almost all children receive at least some transfer income until we reach children with family incomes in the top quintile. Over-all, we can characterize the Norwegian social transfer system as the most universal/least targeted and the U.S. system as the least universal/most targeted.

TABLE 16: Percentage of Children Receiving Transfers by Income Quintile

	Quintile 1	Quintile 2	Quintile 3	Quintile 4	Quintile 5
Canada	99.9	99.9	99.9	95.6	66.3
Norway	98.9	100	100	100	99.5
United States	97.7	87.1	50.8	27	14.4
Newfoundland	100	100	100	100	73
Prince Edward Island	100	100	100	100	82.5
Nova Scotia	100	100	100	98.2	63.7
New Brunswick	100	100	100	100	66.4
Quebec	100	100	100	99.9	90.5
Ontario	100	100	100	91.5	60.2
Manitoba	100	100	100	100	74.6
Saskatchewan	99.7	100	100	100	75.5
Alberta	100	99.4	100	97.6	45.9
British Columbia	99.7	100	99.5	94.9	46.9
New England	98.2	64.9	36.4	13.6	11.3
Middle Atlantic	96.7	84	46.8	27.4	13.7
East-North Central	99.6	87.3	47.3	24.5	18.5
West-North Central	97.7	83.7	38.3	27	10.8
South Atlantic	97.4	84.6	49.1	21.2	11.5
East-South Central	98.5	93.2	64.6	28.8	15
West-South Central	99	97.1	67.9	35.5	17.2
Mountain	98.6	82.1	49.6	23.4	13.5
Pacific	98	97	59.7	33.6	19.4
Source: Luxembourg Income Study

Norway, the country with the least targeted/most universal transfer system, has the best outcomes for children. Of course, it is also true that Norway offers the highest level of benefits. Thus, it is not clear which is the more important factor. When entered in a probit specification which includes all the other 'macro/policy/context' variables, the percentage of children in the top quintile of the income distribution who receive transfers is not a particularly important predictor of children's outcomes in comparison with the level of social transfers (see Table 17). Perhaps the most important role of a universal system of transfers is in ensuring continued support for high levels of transfers through good times and bad by avoiding an 'us and them' mentality. A simple regression of the average level of transfers received in each region on the percentage of the top quintile receiving benefits in that region shows that regions with more universal benefits (i.e., higher percentages of the top quintile receiving benefits) have higher average levels of benefits.⁴³

TABLE 17: Probit Analysis of the Probability of Alternative Child Outcomes, Adding Structure of Transfers

	Asthma	Injury	Limited in     activity	Anxiety/Fear	Restless/Overly active	Disobedie nt at school
	Ages 4-11	Ages 0-11	Ages 0-11	Ages 4-11	Ages 4-11	Ages 4-11
Intercept	0.35 (2.34)	-1.84* (0.22)	-1.58* (0.32)	1.05* (0.21)	0.40 (0.30)	0.44 (0.41)
Dummy=1 if mother smokes daily	0.09* (0.03)	0.10* (0.02)	0.10* (0.03)	0.02 (0.02)	0.23* (0.02)	0.16* (0.03)
Dummy=1 if mother smokes occasionally	-0.03 (0.06)	0.03 (0.05)	-0.08 (0.07)	-0.01 (0.04)	-6.84E-3 (0.04)	0.03 (0.06)
Dummy=1 if poor family	-0.10** (0.04)	-0.11* (0.03)	0.02 (0.04)	0.09* (0.03)	0.06*** (0.03)	0.09** (0.04)
Dummy=1 if mother was less than 25 at child's birth	0.06** (0.03)	0.07* (0.03)	0.05 (0.03)	0.07* (0.02)	0.04*** (0.02)	0.07** (0.03)
Dummy=1 if child is aged 8 - 11	0.03 (0.03)	0.14* (0.02)	0.25* (0.03)	0.09* (0.02)	-0.22* (0.02)	0.18* (0.03)
Dummy=1 if child has one sibling	-0.09** (0.04)	0.07* (0.03)	-0.12* (0.04)	-0.14* (0.03)	-0.18* (0.03)	-0.14* (0.04)
Dummy=1 if child has two siblings	-0.16* (0.04)	0.07** (0.03)	-4.72E-3 (0.04)	-0.24* (0.03)	-0.27* (0.03)	-0.19* (0.05)
Dummy=1 if child has three or more siblings	-0.35* (0.06)	0.11* (0.04)	-0.09 (0.06)	-0.41* (0.04)	-0.35* (0.04)	-0.025* (0.05)
Dummy=1 if single mother	0.18* (0.04)	0.12* (0.03)	0.16* (0.04)	0.13* (0.03)	0.18* (0.03)	0.27* (0.04)
Dummy=1 if female child	-0.24* (0.03)	-0.17* (0.02)	-0.17* (0.03)	0.02 (0.02)	-0.36* (0.02)	-0.61* (0.03)
Equivalent income	7.43E-7 (1.08E-6)	1.02E-6 (7.89E-7)	-2.32E-6*** (1.30E-6)	-3.66E-6* (8.20E-7)	-4.79E-6* (8.0E-7)	-2.57E-6** (1.19E-6)
Unemployment	-2.92E-3 (2.6E-3)	0.02** (0.01)	0.04* (0.01)	0.03* (0.01)	0.02*** (0.01)	1.43E-3 (0.02)
Average social transfers	5.5E-5 (6.1E-5)	-2.0E-5 (2.9E-5)	-1.6E-4* (4.2E-5)	-8.0E-5* (2.7E-5)	-7.0E-5** (3.4E-5)	-1.3E-4* (4.8E-5)
Student teacher ratio	0.02 (0.07)	0.02** (0.01)	-7.50E-3 (0.01)	-0.01*** (0.01)	-0.02** (0.01)	4.13E-3 (0.01)
Physicians per 100,000	1.70E-3 (1.36E-3)	5.49E-4 (4.44E-4)	4.33E-4 (6.27E-4)	-2.65E-3* (4.15E-3)	-1.76E-3* (4.69E-4)	-3.2E-4 (6.44E-4)
% of immigrants as heads of households	-9.44E-3 (1.27E-3)	1.58E-3 (1.88E-3)	8.75E-3* (2.71E-3)	0.01* (0.00)	2.48E-3 (2.44E-3)	0.01* (3.35E-3)
% of public expenditures on health Care	-0.03*** (0.01)	1.16E-3 (1.87E-3)	0.01* (2.68E-3)	-6.86E-3* (1.78E-3)	0.01* (2.37E-3)	-0.01* (3.24E-3)
GDP per capita	-4.56E-6 (1.3E-5)	-8.77E-7 (7.23E-6)	-2.0E-5*** (9.81E-6)	-1.0E-5*** (6.83E-6)	3.1E-6 (7.0E-6)	-1.0E-5 (9.54E-6)
% of children receiving transfers in the top income quintile	-3.29E-3 (6.89E-3)	-1.06E-3 (1.30E-3)	-3.79E-3** (1.87E-3)	1.85E-3 (1.21E-3)	6.73E-4 (1.30E-3)	0.01* (1.81E-3)
Note: Standard errors are in parentheses.
* significant at the 99% level
** significant at the 95% level
*** significant at the 90% level

• ²⁴We have collected and tested a number of other macro and context variables: poverty in the region, percent of lone mothers, percent of mothers in the labour force, gender wage ratio, unemployment history, level of trust in the region. Some of these variables were too highly correlated with almost everything else (e.g., poverty in the region; percent of lone mothers). Others were neither statistically significant nor central to our story, hence they are not discussed further.
• ²⁵Ideally, we should use a measure of 'social spending on children' but such a measure is not available (and is not even easy to conceptualize since some fraction of spending on general infrastructure, such as highways or street lights, is for the benefit of children as well as adults). Later specifications consider average levels of social transfers received by families with children.
• ²⁶While we have social spending per capita by Canadian provinces, this is not available for U.S. regions. Thus, the cross-country data is supplemented with regional variation only for Canada.
• ²⁷All results reported in this paper employ sampling weights. We have also run all specifications without using weights and find basically the same story. Note is made of important exceptions.
• ²⁸None of the child outcome data sets contains particularly good measures of income, hence we cannot use, for example, pre-transfer income.
• ²⁹The coefficient on accidents/injuries is positive and significant in the unweighted regression.
• ³⁰This research is exploratory in nature. Hence, we choose not to push this too far, and so look only for very broad policy indicators.
• ³¹This is calculated for all children aged 0 to 11, regardless of whether or not a transfer was received. Thus, it is a measure which reflects both the percentage of the population receiving transfers and the level of transfers received. It is basically an average per child expenditure figure.
• ³²Table 15, discussed later in this section, reports our efforts to 'purge' the effects of unemployment from transfer receipt using multivariate work with the LIS data.
• ³³We have also estimated all equations using only the sample of 6 to 11 year-olds, for whom student/teacher ratios are presumably most relevant. Significant differences are noted as appropriate through the text, though the basic qualitative story is much the same.
• ³⁴Once again, we have only national level data on this measure for the U.S.
• ³⁵In both the unweighted regressions and the regressions with 6 to 11 year-olds only, restless/overly active behaviour is not statistically affected by unemployment and the negative sign on disobedience at school remains.
• ³⁶Average social transfers are not significant in the average social transfers probit in the unweighted regressions and for the 6 to 11 year-old sample.
• ³⁷That is, if transfers are high because of economic hard times, and if economic hard times are bad for child outcomes, then we will be associating high levels of transfers with low levels of child outcomes and concluding that the transfers don't matter as much as may actually be true.
• ³⁸In Tables 4 and 5 we report estimates, for Canada and the U.S., of a two-stage estimation procedure in which we first estimate the probability of receiving a transfer and then estimate the level of transfer for those who receive. However, we do not use these estimates in the procedure which follows for two reasons: 1) it is not possible to estimate a probability of benefit receipt for Norwegian children, since all Norwegian children receive benefits; 2) we want a measure which is comparable to that used in the previous section which reflects spending per child. The two-stage results are simply reported to demonstrate that they basically tell the same story, and thus to reassure economists who are more accustomed to the two-stage procedure.
• ³⁹For Norway, we are forced to proxy this with someone in the household received unemployment insurance.
• ⁴⁰We have also done this for children living with single mothers. Results are not much different. We choose the two-parent results because more children live with two parents.
• ⁴¹The relatively small sample size of the World Values Survey necessitated aggregation of some Canadian provinces (Newfoundland, PEI, Nova Scotia and New Brunswick were aggregated; Manitoba and Saskatchewan were aggregated).
• ⁴²In this section, we are treating the 'lazy' variable as a potential proxy for the underlying generosity of the social safety net. However, it could also be interpreted as an important indicator of social context.
• ⁴³Average Social Transfers = 2808 + 52.73 % Top Quintile. (7.58) (7.78) Adjusted R-squared = 0.75. T-ratios are in parenthesis.