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

6. National and Regional Differences in Socioeconomic Characteristics at the Micro Level

Previous

Contents

Next

Table 3 reports means for each country of micro-level socioeconomic characteristics which, in earlier research, have been found to influence the well-being of children.¹⁵ Since the focus of this project is not the association between family-level characteristics and child well-being,¹⁶ yet it is of course necessary to control for the individual-level characteristics which other researchers have found to be important, the research strategy employed in this work is basically to adopt the specification employed by Dooley, et.al. (1998) and then to add the macroeconomic, policy and social context variables which are of central interest in this paper. Since Dooley et.al. also employ the NLSCY, it would obviously be possible to use exactly the same specification for the Canadian data. This is also possible for the U.S. data which are extremely similar to the Canadian, but unfortunately not for the Norwegian data which do not report mother's level of education. It is, of course, essential to have exactly the same information for each country in order to pool the three microdata sets and estimate one regression equation, thus we are forced to choose the 'lowest common denominator.' Fortunately, this requires very little change from the Dooley et.al. (1998) specification. The only significant difference in terms of microeconomic variables included as regressors is that we cannot include education, but add smoking behaviour of the mother, which is highly correlated with education.

TABLE 3: Means - Micro Level Socioeconomic Characteristics (National Level)

	Mom smokes daily	Equivalent income ($Cdn-ppp)	Poor family	Lone mom	Mom < 25 at child's birth	Female child	Child aged 8-11	Only child
Canada	25.6 (0.302)	18,410 (89.57)	17.1 (0.259)	14.6 (0.243)	22.9 (0.291)	48.8 (0.344)	33.5 (0.325)	20.1 (0.276)
Norway	32.5* (1.254)	17,100 (369.84)	11.6* (0.789)	16.1 (0.907)	23.9 (1.070)	51.5 (1.232)	29.2* (1.121)	21.1 (1.006)
United States	25.7 (0.614)	22,149* (257.12)	21.2* (0.571)	23.8* (0.693)	23.1 (0.674)	48.9 (0.698)	34.7 (0.664)	15.8* (0.518)
Note: Standard errors are in parentheses. * indicates significantly different from Canada at 90% level of confidence.

If micro-level characteristics differ substantially across the countries, this could be an important explanation for the observed differences in child outcomes. And, in fact, there are some very important differences: 1) significantly more mothers smoke daily in Norway (32.5 percent) than in Canada or the U.S. (25.6 and 25.7 percent, respectively); 2) equivalent¹⁷ gross incomes, expressed in 1994 Canadian dollars, are significantly higher in the U.S. ($22,149) than in Canada ($18,410) or Norway ($17,100);¹⁸ 3) significantly more young children live in poor families in the U.S. (21.2) than in Canada (17.1) while significantly fewer live in poverty in Norway (11.6);¹⁹ 4) significantly more children live in lone-mother families in the U.S. (23.8 percent) than in Canada (14.6 percent) or Norway (16.1 percent); 5) children are slightly younger in Norway (only 29.2 percent of all children aged 0 to 11 years are in the 8 to 11 year category versus 33.5 percent in Canada and 34.7 percent in the U.S.); 6) more children are only children in Canada and Norway (20.1 percent and 21.1 percent) than in the U.S. (15.8 percent).

It is very important to note that some of these differences in micro-level characteristics across the countries may already largely be the result of differences in policy. Certainly, this is true for household poverty status which is highly dependant upon macroeconomic policy and the level of social transfers available. For example, a simple ordinary least squares regression of regional child poverty rate upon regional unemployment rate and average social transfers received by children in the region shows both unemployment and social transfers to be statistically significant predictors of poverty level.²⁰ For example, a 1 percentage point increase in unemployment is predicted to increase child poverty by 1.4 percentage points; a $500 increase in the average level of social transfers received per child is predicted to reduce child poverty by 2 percentage points, other things equal.

In addition, other micro-level variables may also be affected by policy. For example, smoking behaviour could be influenced by tax rates on cigarettes or 'anti-smoking' advertising campaigns; lone-parent status may be influenced by divorce laws, the enforcement of child support payments, etc. Thus, it is important to be clear that this paper is far from providing an exhaustive test of the influence of policy on outcomes for children. Rather, it is attempting to demonstrate a role for policy beyond the impact on the more frequently studied micro variables.

Past work (e.g., Dooley, et.al., 1998) has concluded that children's outcomes are worse if their mothers smoke, if family income is lower, particularly if the family lives with income less than poverty level, if the child lives with a lone mother, or has siblings (particularly many siblings). Thus, except insofar as average family income levels are higher in the U.S., a quick look at these micro variables would lead U.S. to expect outcomes for children to be worse in the U.S. than in Norway or Canada. It is not a priori obvious whether we would, on the basis of Table 3, expect outcomes for young children to be worse in Norway or Canada. Mean characteristics for children are rather similar in Norway and Canada, except that significantly more mothers smoke in Norway (a bad thing), but significantly fewer children live in poverty in Norway (a good thing).

Table 4 reports the results of probit analyses of the probability of young children experiencing each of the 6 negative outcomes discussed in Section 3. For these regressions, microdata files for the 3 countries are pooled together, and dummy variables are introduced to indicate country of residence (with Canada as the base). Estimated equations control for mother's smoking habits, family income status, including a dummy variable to indicate low-income status, mother's age at the time the child was born, child's age, gender and number of siblings. Results for these microlevel variables are consistent with the literature.²¹ For example, children's outcomes are almost always worse if the mother smokes, if she was less than 25 years when the child was born or if she is a lone mother. Outcomes are generally worse if the household is poor.²²

TABLE 4: Probit Analysis of Alternative Child Outcomes Including National Level Dummy 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	-0.960* (0.050)	-1.374* (0.035)	-1.775* (0.050)	-0.236* (0.038)	0.667* (0.039)	-0.73* (0.05)
Dummy=1 if mother smokes daily	0.094* (0.031)	0.099* (0.024)	0.086* (0.033)	0.009 (0.023)	0.228* (0.023)	0.16* (0.03)
Dummy=1 if mother smokes occasionally	-0.031 (0.062)	0.027 (0.045)	-0.072 (0.067)	0.007 (0.042)	-0.004 (0.043)	0.02 (0.06)
Dummy=1 if poor family	-0.090** (0.042)	-0.110* (0.032)	0.011 (0.042)	0.090* (0.029)	0.059** (0.030)	0.08** (0.04)
Dummy=1 if mother was less than 25 at child's birth	0.069** (0.032)	0.068* (0.025)	0.061*** (0.034)	0.071* (0.022)	0.042*** (0.023)	0.06** (0.03)
Dummy=1 if child is aged 8 - 11	0.034 (0.027)	0.133* (0.022)	0.247* (0.096)	0.085* (0.019)	-0.220* (0.019)	0.18* (0.03)
Dummy=1 if child has one sibling	-0.097** (0.040)	0.080* (0.029)	-0.107* (0.039)	-0.133* (0.030)	-0.180* (0.031)	-0.15* (0.04)
Dummy=1 if child has two siblings	-0.164* (0.045)	0.064** (0.033)	0.008 (0.044)	-0.226* (0.033)	-0.269* (0.034)	-0.19* (0.04)
Dummy=1 if child has three or more siblings	-0.361* (0.059)	0.116* (0.041)	-0.064 (0.056)	-0.398* (0.040)	-0.353* (0.040)	-0.26* (0.05)
Dummy=1 if lone mother	0.174* (0.039)	0.129* (0.030)	0.171* (0.039)	0.129* (0.028)	0.171* (0.028)	0.27* (0.04)
Dummy=1 if female child	-0.243* (0.027)	-0.171* (0.020)	-0.167* (0.029)	0.021 (0.019)	-0.359* (0.019)	-0.61* (0.03)
Equivalent gross family income (CN 94 $)	5.29E-7 (1.08E-6)	1.21E-6 (7.83E-7)	-1.87E-6 (1.28E-6)	-3.45E-6* (8.13E-7)	-4.91E-6* (7.97E-7)	-2.36E-6** (1.18E-6)
Dummy=1 if Norway	-0.328* (0.060)	-0.137** (0.048)	-0.301* (0.078)	-0.897* (0.054)	n/a	n/a
Dummy=1 if the United States	n/a	0.007 (0.025)	-0.082** (0.035)	-0.112* (0.0231)	-0.460* (0.023)	0.07** (0.03)
Note: Standard errors are in parentheses.
* indicates significance at the 99% level
** indicates significance at the 95% level
*** indicates significance at the 90% level

However, the central question addressed in Table 4 is: once we have controlled for relevant micro-level determinants of child well-being, are there still statistically significant differences across countries which might be due to other influences of policy? With respect to Norway, the answer to this question is unambiguous. For all outcomes for which we have comparable microdata, children in Norway are better off than children in Canada (i.e., they are less likely to experience negative outcomes) conditional on standard micro characteristics. Moreover, the magnitudes of these effects are very large — twice the size, for example, of 'lone-mother' effects (except in the case of injury). Note, again, that we are already controlling for household income and poverty status, variables which will reflect macroeconomic conditions such as the unemployment rate and which include contributions by the state to family income in the form of social transfers and thus incorporate a very important dimension of policy.

For Canada and the U.S., the patterns observed from a simple comparison of proportions basically hold for the multivariate comparisons. Children in the U.S. are less likely to be fearful/anxious or to be restless/overly active than children in Canada. They are also less likely to experience activity limitation, once we have controlled for other relevant factors (this is a change from the basic descriptive analysis presented earlier). There is again no statistically significant difference between Canada and the U.S. in the likelihood that a child will have an accident/be injured. Children in the U.S. are significantly more likely to be disobedient at school. While not so large as the Canada/Norway effects, the U.S. dummies, where significant, compare in magnitude with the effects of mother smoking, for example.

Table 5 illustrates the extent of regional variation in micro-level characteristics. For example, roughly 30 percent of mothers smoke daily in the Atlantic provinces and Quebec versus only 20.6 percent in BC. In Newfoundland, 30 percent of young children live in poor families versus only 14.9 percent in Alberta. Regional variation is, if anything, even larger in the U.S. For example, 34.9 percent of mothers smoke daily in the Mountain region versus only 15.9 percent in the Pacific. In the East-South-Central, 36.2 percent of children are poor in East-South-Central versus only 10.7 percent in New England.

And, Table 6 indicates that even after controlling for micro-level characteristics, region is a statistically significant predictor of child outcomes. That is, from one to two-thirds of the regional dummy variables are statistically different from the province of Ontario (the base case) for all outcomes.²³ Specific regional patterns are not the focus of this paper (see Phipps, 1999 for a discussion of provincial differences in child outcomes). The key message of Table 6 is again that standard micro-level determinants do not tell the whole story about child outcomes.

TABLE 5: Means - Micro Level Determinants (Regional Level)

	Mother smokes daily	Equivalent income ($Cdn-ppp)	Poor family	Lone mother	Mother < 25 at child's birth	Female child	Child aged 8-11	Only child
Newfoundland	30.5* (1.345)	14,901* (350.89)	30.0* (1.335)	13.8 (1.006)	37.0* (1.411)	49.1 (1.456)	37.2 (1.408)	23.2* (1.238)
Prince Edward Island	29.9* (1.697)	13,764* (328.54)	25.1* (1.594)	14.1 (1.278)	27.2* (1.650)	48.2 (1.838)	34.8 (1.751)	15.1* (1.340)
Nova Scotia	29.3* (1.205)	15,302* (249.13)	24.4* (1.133)	19.8* (1.051)	29.8* (1.210)	49.2 (1.318)	34.7 (1.255)	17.6 (1.009)
New Brunswick	31.1* (1.284)	14,885* (249.66)	21.4* (1.133)	13.3 (0.939)	30.1* (1.272)	49.4 (1.380)	35.0 (1.317)	19.6 (1.102)
Quebec	31.1* (0.754)	17,656* (182.77)	17.9 (0.620)	13.7 (0.558)	21.4 (0.668)	48.9 (0.810)	33.1 (0.762)	25.1 (7.035)
Ontario	22.6 (0.564)	19726 (186.02)	14.8 (0.476)	15.3 (0.482)	20.5 (0.544)	48.7 (0.669)	33.2 (0.630)	19.4 (0.531)
Manitoba	24.1 (1.063)	16,116* (275.69)	21.7* (1.018)	12.2* (0.806)	23.4 (1.053)	48.0 (1.233)	32.8 (1.116)	18.4 (0.961)
Saskatchewan	30.3* (1.113)	15,742* (275.66)	24.4* (1.035)	15.7 (0.876)	30.7* (1.112)	49.5 (1.204)	34.6 (1.146)	15.3* (0.872)
Alberta	25.1 (0.973)	19,466 (370.79)	14.9 (0.797)	11.9* (0.723)	26.7* (0.993)	48.7 (1.118)	34.0 (1.059)	16.7 (0.836)
British Columbia	20.6 (0.973)	18,606* (281.65)	15.5 (0.866)	16.1 (0.879)	21.9 (0.995)	48.7 (1.195)	33.8 (1.130)	17.7 (0.914)
New England	24.9 (3.064)	27,524* (1236.0)	11.7 (2.252)	16.5 (2.907)	10.7* (2.397)	50.0 (3.509)	27.0 (3.114)	16.1 (2.603)
Mid-Atlantic	27.1 (1.874)	26,837* (949.1)	18.2 (1.625)	24.1* (2.081)	18.9 (1.849)	48.8 (2.106)	28.1 (1.895)	16.0 (1.571)
East-North Central	25.6 (1.398)	21,412 (538.9)	18.1 (1.221)	21.8* (1.502)	20.9 (1.456)	49.3 (1.584)	34.2 (1.503)	13.3* (1.093)
West-North Central	31.6* (2.459)	18,257 (733.3)	23.6* (2.223)	23.1* (2.528)	22.7 (2.489)	44.1 (2.599)	38.6 (2.548)	13.4* (1.806)
South Atlantic	26.7 (1.336)	21,346* (528.9)	22.7* (1.260)	23.1* (1.486)	24.1 (1.492)	49.1 (1.502)	34.2 (1.425)	18.7 (1.195)
East-South Central	32.0* (2.870)	17,949 (1043.3)	31.0* (2.830)	28.7* (3.428)	36.2* (3.611)	48.4 (3.058)	43.8* (3.036)	25.5 (2.737)
West-South Central	21.6 (1.708)	21,931* (821.3)	26.0* (1.798)	26.1* (2.123)	29.6* (2.150)	50.6 (2.049)	37.4 (1.984)	14.2* (1.464)
Mountain	34.9* (2.973)	20,753 (1195.8)	27.6* (2.788)	28.1* (3.330)	29.5* (3.232)	44.0 (3.096)	44.1* (3.096)	16.0 (2.320)
Pacific	15.9* (1.327)	22,449* (647.7)	21.5* (1.480)	27.6* (1.892)	24.3 (1.797)	52.2 (1.801)	35.4 (1.723)	15.4* (1.319)
Note: Standard errors are in parentheses. * indicates significantly different from Ontario.

 

TABLE 6: Probit Analysis of the Probability of Alternative Child Outcomes Including Regional Level Dummy 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	-0.972* (0.054)	-1.387* (0.039)	-1.727* (0.055)	-0.157* (0.041)	0.639* (0.042)	0.64* (0.04)
Dummy=1 if mother smokes daily	0.091* (0.031)	0.104* (0.024)	0.091* (0.033)	0.014 (0.023)	0.228* (0.023)	0.23* (0.02)
Dummy=1 if mother smokes occasionally	-0.030 (0.062)	0.035 (0.045)	-0.080 (0.068)	-0.019 (0.043)	-0.008 (0.043)	-0.01 (0.04)
Dummy=1 if poor family	-0.099** (0.043)	-0.113* (0.032)	0.016 (0.043)	0.094* (0.029)	0.049*** (0.030)	0.05*** (0.03)
Dummy=1 if mother was less than 25 at child's birth	0.064** (0.032)	0.065** (0.025)	0.052 (0.034)	0.070* (0.023)	0.033 (0.023)	0.03 (0.02)
Dummy=1 if child is aged 8 - 11	0.033 (0.027)	0.139* (0.022)	0.251* (0.030)	0.087* (0.019)	-0.222* (0.019)	-0.22* (0.02)
Dummy=1 if child has one sibling	-0.094** (0.040)	0.073** (0.029)	-0.123* (0.040)	-0.143* (0.030)	-0.178* (0.031)	-0.18* (0.03)
Dummy=1 if child has two siblings	-0.160* (0.045)	0.063*** (0.033)	-0.005 (0.044)	-0.238* (0.033)	-0.272* (0.034)	-0.27* (0.03)
Dummy=1 if child has three or more siblings	-0.351* (0.059)	0.104** (0.041)	-0.093 (0.057)	-0.413* (0.040)	-0.349* (0.040)	-0.35* (0.04)
Dummy=1 if lone mother	0.181* (0.039)	0.123* (0.030)	0.172* (0.039)	0.127* (0.028)	0.176* (0.029)	0.18* (0.03)
Dummy=1 if female child	-0.244* (0.027)	-0.173* (0.020)	-0.169* (0.029)	0.020 (0.019)	-0.361* (0.019)	-0.36* (0.02)
Equivalent income	7.08E-7 (1.08E-6)	9.69E-7 (7.91E-7)	-2.26E-6*** (1.30E6)	-3.66E-6* (8.23E-7)	-4.86E-6* (8.02E-7)	-4.86E-6* (8.02E-7)
Dummy=1 if Norway	-0.319* (0.062)	-0.116** (0.051)	-0.331* (0.080)	-0.962* (0.056)	n/a	n/a
Dummy=1 if Newfoundland	0.085 (0.094)	0.024 (0.088)	-0.106 (0.124)	-0.084 (0.079)	0.111 (0.080)	0.11 (0.08)
Dummy=1 if Prince Edward Island	0.459* (0.167)	-0.196 (0.192)	0.006 (0.224)	-0.107 (0.157)	0.127 (0.156)	0.13 (0.16)
Dummy=1 if New Brunswick	0.175** (0.083)	-0.017 (0.079)	0.022 (0.102)	-0.104 (0.071)	-0.021 (0.070)	-0.02 (0.07)
Dummy=1 if Nova Scotia	0.264* (0.072)	0.155** (0.066)	0.002 (0.092)	-0.129** (0.064)	0.112*** (0.064)	0.11*** (0.06)
Dummy=1 if Quebec	0.0004 (0.036)	-0.008 (0.032)	-0.229* (0.047)	-0.139* (0.029)	0.019 (0.029)	0.02 (0.03)
Dummy=1 if Manitoba	-0.031 (0.075)	-0.034 (0.066)	-0.065 (0.089)	-0.121** (0.060)	0.105*** (0.059)	0.10*** (0.06)
Dummy=1 if Alberta	0.013 (0.048)	0.058 (0.041)	0.155* (0.052)	-0.102* (0.039)	0.091** (0.039)	0.09** (0.04)
Dummy=1 if Saskatchewan	-0.114 (0.079)	0.070 (0.064)	-0.033 (0.089)	-0.088 (-0.060)	0.183* (0.060)	0.18* (0.06)
Dummy=1 if British Columbia	-0.057 (0.046)	0.091** (0.038)	0.028 (0.052)	-0.052 (0.036)	0.003 (0.036)	0.00 (0.04)
Dummy=1 if New England	n/a	-0.155 (0.103)	-0.073 (0.139)	-0.083 (0.091)	-0.355* (0.090)	-0.36* (0.09)
Dummy=1 if Mid-Atlantic	n/a	0.159* (0.059)	-0.051 (0.088)	-0.315* (0.063)	-0.500* (0.060)	-0.50* (0.06)
Dummy=1 if East-North Central	n/a	-0.047 (0.049)	-0.107 (0.068)	-0.095** (0.043)	-0.458* (0.043)	-0.46* (0.04)
Dummy=1 if West-North Central	n/a	0.090 (0.072)	-0.283** (0.119)	-0.240* (0.070)	-0.566* (0.069)	-0.57* (0.07)
Dummy=1 if South Atlantic	n/a	0.038 (0.052)	0.105 (0.066)	-0.318* (0.049)	-0.463* (0.047)	-0.46* (0.05)
Dummy=1 if East-South Central	n/a	-0.427* (0.132)	0.019 (0.129)	0.060 (0.089)	-0.154*** (0.090)	-0.15*** (0.09)
Dummy=1 if West-South Central	n/a	0.069 (0.075)	-0.606* (0.164)	-0.221* (0.072)	-0.061 (0.069)	-0.06 (0.07)
Dummy=1 if Mountain	n/a	0.038 (0.103)	-0.256 (0.163)	-0.179*** (0.094)	-0.412* (0.093)	-0.41* (0.09)
Dummy=1 if Pacific	n/a	0.127** (0.062)	-0.296* (0.106)	-0.098*** (0.058)	-0.541* (0.058)	-0.54* (0.06)
Note: Standard errors are in parentheses.
* indicates significance at the 99% level
** indicates significance at the 95% level>
*** indicates significance at the 90% level

• ¹⁵'Micro-level' variables are known for each child in the outcomes data sets and enter the multivariate analysis as such. However, for the purposes of discussion here, we compare the average for all children within each country.
• ¹⁶See Phipps, 1998a which focusses upon exactly this issue, asking whether particular characteristics of the child's family, such as lone-parent status, have the same impact on child well-being in each country.
• ¹⁷'Equivalent' income adjusts for family size using the OECD equivalence scale.
• ¹⁸Following Hanratty and Blank (1992), we convert all currencies to 1994 Canadian dollars, using the 1990 OECD estimate of purchasing power parity (PPP) for individual consumption by households (OECD, 1990, Table 1.5, pp. 30/31, line 1). We extrapolate PPP to the appropriate year using country-specific deflators for private final consumption (OECD,1996, pp. 102,104, 123).
• ¹⁹A child is deemed 'poor' if his or her family gross equivalent income is less than 50 percent of median equivalent income for that country. This procedure allows for comparability across countries, and within Canada, leads to estimates which are qualitatively very similar to those which would be obtained using the LICOs.
• ²⁰Poverty = 36.35 + 1.41 Unemployment -0.004 Social Transfers (9.62) (2.68) (-3.37) where t-ratios are presented in parenthesis and the adjusted r-squared is 0.33.
• ²¹See Phipps, 1998a for a detailed comparison of microdata results for the 3 countries.
• ²²Although the regression results for various outcomes are presented in a single table to conserve space, it is not appropriate to compare magnitude of effects across equations which have different dependent variables.
• ²³Given the Canadian focus of this project, a Canadian province seemed a sensible choice for the base case, and Ontario is the most populous province. However, it is important to note that there is nothing definitive about saying that another province or U.S. region has outcomes which are statistically different from Ontario. That is, it is entirely possible that two regions which are not statistically different from Ontario are nonetheless statistically different from one another.