Family Background, Family Income, Maternal Work and Child Development - October 1998

6.1 Descriptive Statistics for Cognitive Development

Table A1 (in the statistical appendix), displays the sample means and standard deviations for the variables used in the regression analysis for PPVT scores. Because of a better fit in the regressions, the log of PPVT scores was used as the dependent variable.

6.1.1 Full sample

Most children included in the sample are from large cities. Half are boys and half are girls. They have 1.32 brothers and sisters and have an average birth order of 1.49. The mean age for mothers is 28.6, family income is 51,000 dollars and 57 percent of mothers work more than 26 weeks per year. More than 60 percent of the children are read to by their parents at least once a day. Mothers have on average 12 and a half years of education and were for the most part born in Canada. Very few children were in poor health and heavily distracted during the test. Finally, about 15 percent of the children were in families receiving at least one payment of welfare assistance in the past year and most children came from neighbourhoods with low levels of poverty. Since the means for the full sample are very similar to the sample of two parent children, the next paragraph compares the means of two-parent children (TPC) with single mother children (SMC).

6.1.2 Comparing two-parent children (TPC) with single-mother children (SMC)

The single mothers of the children in the sample gave birth much earlier than mothers in two-parent families. SMC have 0.89 brothers and sisters as compared to 1.40 for TPC. TPC have mothers with one more year of education who work considerably more weeks per year. The most important differences between these children are the income their family takes in, as two-parent families generate three times more income per year than single mothers, and the proportion of SMC in families receiving welfare is twice the proportion for TPC. PPVT scores of TPC are 5 percent higher than SMC.

6.1.3 Comparing children with mothers who have a strong attachment to the labour market (SAM) with children who have mothers with a low attachment to the labour market (LAM)

In the case of two-parent children (TPC), SAM are exactly one year older than LAM, which is consistent with a Beckerian model of dynamic fertility decisions where it is optimal to delay fertility in order to capitalize on investments in human capital. In the same vein, they have less children and are more educated. However, SAM read as frequently to their children as LAM, which is surprising because they use up more time in the labour market. Average family income is of course much less in LAM families, concomitantly, they receive proportionally more welfare payments. Finally, slightly more LAM are immigrants and PPVT scores for children with SAM are slightly higher than for children with LAM (about 4 percent higher). In the case of single mother children (SMC), SAM are almost, on average, two years older than LAM, they also have less children, are better educated, and have ten thousand dollars more in income while receiving welfare payments in a much smaller proportion than LAM. Almost all children with SAM who are immigrants have settled in Canada more than five years ago, they also read slightly more frequently than LAM. So differences between LAM and SAM for SMC are similar to differences for TPC. However, PPVT scores are lowest for SMC with LAM (6 percent lower than in the case of the full sample).

6.2 Regression Results for PPVT

6.2.1 Full sample

Table 1 singles out from the regression results the coefficients associated with the work and income variables. For the full sample, twelve regressions were performed. A set of four regressions were done with each of the three labour supply indicators: WORK, WEEKS, and WEEKFT and WEEKPT. There are two regressions for each measure of income (total family income, other income), one regression that includes a dummy for welfare receipt and one that does not. Columns 2 and 4 are the results for the regressions with income from other sources and Columns 1 and 3 present the results for regressions with the full income of the family.

Table 1: Regression Coefficients on Work and Income Variables

	All Families			Two-parent Families			One-parent families
	Model 1	Model 2	Model 3	Model 4	Model 3	Model 4	Model 3
A. WORK	0.0097	0.0151	0.0054	0.0097	0.0054	0.0086	0.0245
	(1.65)	(2.63)	(0.96)	(1.64)	(0.85)	(1.37)	(1.16)
Family income	0.0044		0.0032		0.0021		-0.0078
	(3.56)		(2.50)		-0.0035		(0.43)
Income square	-0.0059		-0.0044		(1.65)		0.0000
	(2.93)		(2.15)				(0.81)
Other income		0.0054		0.0039		0.0032
		(4.30)		(2.98)		(2.10)
Other income square		-.0000		-0.0000		-0.0000
		(3.43)		(2.51)		(2.09)
Welfare			-0.0364	-0.337	-0.033	-0.0325
			(4.26)	(3.86)	(2.83)	(2.79)
Spouse weeks					0.0005	0.0004
					(2.12)	(1.87)
B. WEEKS	0.0002	0.0003	0.0001	0.0002	0.0002	0.0002	0.0005
	(1.91)	(2.94)	(1.08)	(1.86)	(1.19)	(1.72)	(1.04)
Family income	0.0043		0.0031		0.0020		-0.0074
	(3.46)		(2.45)		(1.42)		(0.41)
Income square	-0.0058		-0.0044		-0.0034		0.0000
	(2.87)		(2.12)		(1.58)		(0.80)
Other income		0.0054		0.0096		0.0033
		(4.31)		(3.02)		(2.14)
Other income square		-0.0000		-0.0000		-0.0000
		(3.44)		(2.53)		(2.12)
Welfare			-0.0360	-0.0328	-0.0324	-0.0317	-0.0262
			(4.20)	(3.75)	(2.77)	(2.71)	(1.21)
Spouse weeks					0.0005	0.0004
					(2.13)	(1.86)
C. WEEKS full-time	0.0002	0.0004	0.0001	0.0002		0.0002	0.0004
	(1.59)	(2.80)	(0.86)	(1.77)		(1.66)	(0.78)
C. WEEKS part-time	0.0002	0.0003	0.0002	0.0002		0.0002	0.0006
	(1.69)	(2.04)	(1.03)	(1.35)		(1.21)	(1.03)
Family income	0.0043		0.0031				-0.0064
	(3.48)		(2.47)				(0.35)
Income square	-0.0058		-0.0044				0.0000
	(2.88)		(2.14)				(0.76)
Other income		0.0055		0.0040		0.0033
		(4.33)		(3.01)		(2.15)
Other income square		-0.0000		-0.0000		-0.0000
		(3.46)		(2.54)		(2.13)
Welfare			-0.0360	-0.0328		-0.0317	-0.027
			(4.20)	(3.74)		(2.71)	(1.24)
Spouse weeks						0.0004
						(1.86)
Sample size	2,840	2,840	2,840	2,840		2.436	381

First, for the results obtained with the work dummy as the labour supply indicator, the effects are strongest (as for the other labour supply indicators) for the specification with no welfare variable and the income from other sources variable. For this latter measure of income, the work variable will capture some of the income effects that come with mother's work since the income from their work is not included in the other income variable, however the effect of the work dummy is to increase approximately the PPVT score by only 1.5 percent. This would be an upper bound on the total effects of mother's employment. The three other specifications have lower values for the work dummy parameter and none are significant at the 95 percent level of confidence. Both the inclusion of the welfare variable and the use of total family income reduce the effect of the work dummy. The welfare effect is relatively strong and significant at the 99 percent level of confidence reducing scores by approximately 3.5 percent. However, we cannot be sure that this identifies a "pure" effect of welfare participation, it may rather capture a non-linearity in the income effects, which may be strongest for the very poor. This assumption is explored more thoroughly in Lefebvre and Merrigan (1999). Income effects, despite being significant, are extremely weak. An increase of income in the order of 20,000 dollars will barely increase scores by one percent.

The results for the two other indicators of labour supply variables (WEEKS; WEEKFT and WEEKPT) basically replicate the results with the work dummy. The results are very similar to those of Hill and O'Neill. Since they control for more factors , such as grandparents education and the mother's skills, and they demonstrate that the introduction of these factors decreases considerably the effect of working hours on PPVT scores, it is feasible that the presence of these factors in our specification could make the effect of the work variable negative and significant. Table 2 presents results of other specifications that provide evidence for this hypothesis. Starting from a purely demographic model and the work dummy variable, it is enriched with other variables to see whether the work dummy effect would be sensitive to specification choices. Again, results match those of Hill and O'Neill. In the demographic model, work has a relatively strong positive effect on scores. However, as more human capital and income variables are added, the work dummy variable becomes weaker and finally non-significant. Therefore, it is feasible that the introduction of these variables could render the work effects negative.

Table 2: Effects of Changes in Model Specification on the Mother's Work Variable* Coefficients (t-ratios in parentheses)

Model Specification	Full Sample	Two-parent families	One-parent families
1. Includes only work plus child's gender, mother's age at child birth, provinces and urban area	0.0260 (5.00)	0.0184 (2.98)	0.0682 (3.99)
2. Also Includes PPVT circumstances (child's health problem and distraction during test)	0.0237 (4.48)	0.0181 (2.84)	0.0664 (3.87)
3. Also includes child's siblings	0.0226 (4.96)	0.0217 (3.51)	0.0644 (3.78)
4. Also includes mother's education and frequency of reading to the child (and spouse/partner's weeks of work and education two-parent families)	0.0161 (2.81)	0.0099 (1.59)	0.0450 (2.57)
5. Also includes family income	0.0102 (1.72)	0.0074 (1.17)	0.0382 (1.98)
6. Also includes low income neighbourhood	0.0097 (1.65)	0.0073 (1.16)	0.0349 (1.78)
7. Also includes welfare	0.0054 (0.90)	0.0054 (0.85)	0.0245 (1.16)
5a Also includes other income	0.0161 (2.,81)	0.1100 (1.77)	0.0440 (2.49)
6a Also includes low income neighbourhood	0.0151 (2.63)	0.1080 (1.48)	0.0377 (2.11)
7a Also includes welfare	0.0097 (1.64)	0.0085 (1.37)	0.0251 (1.23)
* Paid Work for 26 weeks or more (see text).

Full results are found in Table 3 (columns 2, 3 and 4) for the specification with the WORK variable, the total family income variable, and the welfare dummy.²⁴ The mothers age at the birth of the child, region of residence, city size and immigration status all play a significant role in the PPVT score. The only demographic variable playing no role is the child's sex. Except for the case of immigration, the effects are relatively small. However, all other things being equal, a child from PEI with a recent immigrant mother who is 20 years old will, on average, have a PPVT score almost 15 percent lower than a child living in Manitoba with a 30 year old mother who is not an immigrant. Having younger siblings is very detrimental to scores while having older siblings has a relatively small and positive effect. So there can be relatively large differences in scores between children from families of different types.

Human capital variables have strong effects as well since the coefficient on the number of years of education is 0.011 (or 1.1 percent per year of education), almost three times larger than the coefficient on the age of the mother at birth of the child. Children with parents who read little or almost never to their child are expected, all other things kept equal, to score 5.8 percent lower than children with parents reading several times per day, and 3.9 percent lower than children read to once a day. This type of intervention seems to be very noteworthy as a tool to increase PPVT scores. Also, living in a neighbourhood with a high incidence of low income families will produce statistically significant lower PPVT scores. However, the effect is quite small as the incidence is measured in percentage points. It is difficult to ascertain what this effect is capturing, perhaps unobserved heterogeneity correlated with the incidence of low income families or the decreased chances of interacting with children from higher income families.

Table 3: OLS Regression Results for 4- and 5-year-old Children: Dependant Variable Logarithm of Standardized PPVT-R Secore (t-ratios in parentheses)

	Full sample	Two-parent	One-parent	Mothers working full time			Mothers working part time or unemployed
				All	With spouse	Lone	All	With spouse	Lone
Child characteristics
PPVTHealth	-0.025 (1.74)	-0.038 (2.36)	0.031 (0,87)		-0.022 (1.15)
PPVTDistraction	-0.003 (3,85)	-0.002 (2.31)	-0.007 (3.15)	-0.004 (3.30)	-0.002 (1.87)	-0.010 (2,63)	-0.003 (2.28)	-0.002 (1.57)	-0.003 (1.12)
Male	0.002 (0.32)	0.002 (0.27)	-0.004 (0.23)	-0.004 (0.64)	0.003 (0.43)	-0.019 (0.73)	0.009 (1.09)	0.009 (0.98)	0.006 (0.27)
Number of siblings	-0.021 (6.30)	-0.020 (5.90)	-0.021 (2,00)	-0.013 (2.69)	-0.011 (2.16)	-0.033 (1.91)	-0.026 (5.57)	-0.028 (5.36)	-0.014 (0,98)
Birth order	0.021 (4.23)	0.029 (3.96)	-0.157 (1.91)	0.057 (2.11)	0.067 (2.45)	-0.184 (1,09)	0.056 (0.20)	-0.004 (0.12)	-0.221 (2.22)
Mother's characteristics
Age at child's birth	0.004 (5.65)	0.004 (5.74)	0.002 (0.14)	0.002 (2.45)	0.002 (2.61)	0.002 (0.77)	0.005 (5.27)	0.006 (5.28)	-0.001 (0.45)
Years of education	0.010 (6.70)	0.008 (4.37)	0.014 (2.69)	0.011 (5.66)	0.010 (4.26)	0.004 (0.41)	0.009 (3.75)	0.005 (1.82)	0.024 (3.65)
Immigrant1 (>9years)	-0.042 (3.67)	-0.049 (3.96)	-0.028 (0.82)	-0.031 (2.12)	-0.040 (2.59)	0.023 (0.45)	-0.055 (2.96)	-0.060 (2.94)	-0.065 (1.37)
Immigrant2 (5-9years)	-0.011 (5.65)	-0.105 (5.11)	-0.230 (3.69)	-0.084 (2.83)	-0.103 (3.40)	0.014 (0.09)	-0.128 (4.78)	-0.108 (3.68)	-0.037 (4.52)
Immigrant3 (<5 years)	-0.083 (2.75)	-0.090 (3.05)		-0.080 (1.02)	0.070 (1.00)		-0.085 (2.46)	-0.096 (2.70)
Paid work	0.016 (0.27)	0.001 (0.17)	0.022 (1.05)
Spouse's characteristics
Years of education		0.005 (3.48)			0.005 (2.68)			0.005 (2.20)
Weeks of work		0.001 (1.70)			0.000 (1.75)			-0.000 (0.01)
Family characteristics
Family income/10,000	0.003 (2.57)	0.002 (1.44)	-0.001 (0.38)	0.003 (1.87)	0.001 (0.89)	-0.003 (1.16)	0.005 (1.50)	0.002 (0.70)	0.007 (1.06)
Family income squared	-0.000 (2.18)	-0.000 (1.66)	0.000 (0.58)	-0.000 (1.79)	-0.000 (1.38)	0.000 (1.37)	-0.000 (0.91)	-0.000 (0.40)	-0.000 (1.17)
Received welfare	-0.036 (4.28)	-0.034 (2.95)	-0.023 (1.10)	-0.033 (2.17)	-0.020 (1.49)	-0.056 (1.88)	-0.033 (2.80)	-0.038 (2.41)	0.020 (0.64)
Readchild2: weekly	0.021 (0.13)	0.001 (0.06)		-0.014 (0.63)	-0.014 (0.59)		0.013 (0.53)	0.015 (0.54)
Readchild3: daily	0.036 (2.22)	0.029 (1.62)	0.058 (3.49)	0.020 (0.89)	0.016 (0.67)	0.034 (1.32)	0.048 (2.02)	0.042 (1.49)	0.073 (3.13)
Readchild4: >daily	0.052 (2.83)	0.046 (2.27)		0.030 (1.17)	0.029 (1.10)		0.072 (2.64)	0.065 (2.07)
Area characteristics
Newfoundland	0.006 (0.47)	0.012 (0.87)	0.005 (0.15)	0.019 (0.98)	0.020 (0.87)	0.043 (0.87)	0.001 (0.07)	0.009 (0.41)	-0.028 (0.63)
Prince Edward Island	-0.022 (1.43)	-0.027 (1.66)	0.043 (1.00)	-0.018 (0.96)	-0.020 (1.00)	-0.015 (0.20)	-0.027 (1.04)	-0.043 (1.47)	0.089 (1.59)
Nova Scotia	0.011 (0.92)	-0.002 (0.13)	0.064 (2.05)	-0.003 (0.20)	-0.016 (0.95)	0.074 (1.41)	-0.023 (1.29)	0.012 (0.63)	0.045 (1.09)
New Brunswick	-0.018 (1.57)	-0.019 (1.48)	-0.024 (0.72)	-0.039 (2.54)	-0.045 (2.70)	-0.015 (0.30)	0.004 (0.22)	0.008 (0.39)	-0.052 (1.04)
Québec	0.025 (3,00)	0.023 (2,51)	0.046 (1,72)	0.033 (3,41)	0.033 (2,89)	0.068 (1,64)	0.013 (0,93)	0.012 (0,82)	0.031 (0,84)
Manitoba	0.030 (2.54)	0.025 (1.90)	0.066 (1.96)	0.032 (2.24)	0.020 (1.27)	0.097 (1.98)	0.020 (0.97)	0.026 (1.15)	0.000 (0.01)
Saskatchewan	0.018 (1.62)	0.013 (1.20)	0.039 (1.22)	0.018 (0.13)	0.013 (0.98)	0.037 (0.76)	0.015 (0.75)	0.011 (0.46)	0.031 (0.73)
Alberta	0.023 (2.18)	0.021 (1.85)	0.063 (1.59)	0.012 (0.88)	0.005 (0.40)	0.053 (0.90)	0.036 (2.08)	0.040 (2.12)	0.045 (0.84)
British Columbia	-0.007 (0.68)	-0.014 (1.17)	0.040 (1.37)	-0.011 (0.77)	-0.012 (0.79)	0.003 (0.06)	-0.007 (0.41)	-0.016 (0.83)	0.055 (1.50)
Urban area1 (+500)	-0.012 (1.35)	-0.014 (1.47)	-0.008 (0.26)	-0.014 (1.22)	-0.016 (1.01)	-0.010 (0.21)	-0.009 (0.60)	-0.008 (0.46)	-0.018 (0.41)
Urban area2 (100-500)	0.018 (2.75)	0.014 (1.59)	0.047 (1.79)	0.026 (2.52)	0.020 (1.89)	0.071 (1.58)	0.010 (0.75)	0.006 (0.42)	0.010 (0.27)
Urban area3 (30-100)	0.015 (1.50)	0.011 (1.00)	0.018 (0.63)	0.007 (0.58)	0.002 (0.16)	0.033 (0.74)	0.025 (1.55)	0.027 (1.46)	-0.032 (0.77)
Urban area4 (15-30)	-0.001 (0.01)	-0.003 (0.29)	-0.008 (0.25)	-0.017 (1.24)	0.012 (0.77)	0.022 (0.48)	-0.023 (1.33)	-0.021 (1.00)	-0.063 (1.38)
Urban area5 (<15)	0.008 (0.83)	0.002 (0.43)	0.029 (0.96)	0.016 (1.29)	0.011 (0.85)	0.056 (1.12)	-0.001 (0.03)	-0.005 (0.30)	-0.010 (0.24)
Low income neighbourh.	-0.055 (1.75)	-0.015 (0.43)	-0.186 (2.51)	-0.066 (1.46)	-0.054 (1.05)	-0.094 (0.63)	-0.054 (1.20)	0.022 (0.05)	-0.175 (1.93)
Constant	4.34 (143.8)	4.30 (128.1)	4.517 (44.9)	4.355 (93.5)	4.284 (85.1)	4.670 (26.8)	4.317 (84.7)	4.293 (74.9)	4.391 (28,0)
Sample size	2,840	2,422	381	1,558	1,387	156	1,282	1,035	225
Adjusted R-squared	0.141	0.130	0.185	0.107	0.113	0.085	0.150	0.133	0.200
lnPPVT score	4.5899	4.5951	4.5574	4.6043	4.6047	4.5978	4.5711	4.5808	4.5308

6.2.2 Splitting the sample in two-parent children (TPC) and single-mother children (SMC)

Columns four, five and six of Table 1 present results for two-parent and single-mother children. For the sample of TPC, the results presented are only with the welfare receipt dummy variable but still for two measures of income and for the three work indicators. For the SMC only one specification, with the welfare participation dummy and family income, is presented. In this case, it makes less sense to include income form other sources as a regressor since in more than 90 percent of these families the mother is the sole provider. However, results are presented for the three different labour supply specifications.

In the case of two-parent children (TPC), the sample permits the introduction of controls for the spouse or partner's level of education (not necessarily the father's as step-families are included) and the weeks worked in the preceding year. Again the mother's work effects are strongest with the income from other sources as the income variable. However, for none of the cases are the labour supply variables significant at the 95 percent level. The introduction of the two spouses related variables reduces considerably the effect of both types of income. The spouse's weeks worked variable has a positive and weak effect. The welfare effect is very similar to the full sample case. The other sociodemographic effects (see Table 3, column 3) are similar to the full sample case. The spouse's years of education have a positive and significant effect. The incidence of low income neighbourhood families is no longer significant possibly reflecting the increasing control of the child's activities when two parents are present in the family or could simply reflect the reduction in the variance of the variable in this sample.

For single-mother children (SMC), the regression provides different results. The mother's work effects are non significant. The welfare effect is not significant while it is for TPC, given that welfare is very strongly correlated with income in this sample, including both may probably wash out both effects. In the case of demographic variables (see Table 3, column 4), the urban and provincial dummies are jointly significant, the immigration dummies have also a significant effect, but the age effect is not significant. Years of education have a positive effect. Being in a low income neighbourhood, in contrast with TPC, has a negative and significant effect on scores.

6.2.3 Comparing children with SAM and children with LAM

The results presented in Table 3 are for regressions performed with a sample of children with mothers working more than 26 weeks (SAM) in the preceding year, and a sample of children with mothers working less than 26 weeks in a year or not working at all (LAM). First, for two-parent children (TPC), the age of the mother, reading, immigration status, have significantly stronger effects for LAM (see Table 3, columns 6 and 9). The opposite is true for mother's and spouse's education and hours of work. The large difference in the frequency of reading effects could reflect differences in time used for reading and the quality of reading time. LAM can more easily find periods of the day that are more optimal for the child's concentration. The same can be said of the age at birth variable, it is easier for LAM in the labour market to use the human capital built up by experience. This reasoning however should apply to the education variable. However, the results in this case are counterintuitive and could reflect a sample selection if unobserved parental skills are positively correlated with work, increasing the probability a child has a SAM with higher education. On the other hand, the reading effects could reflect a negative correlation between preferences for investment in child-rearing and work.

Second, for single-mother children (SMC), the most obvious difference between both groups are the effects of the human capital variables (see Table 3, columns 7 and 10). The education and frequency of reading effect are positive and significant only for children with LAM. In fact, for these children, the difference between mothers reading at least once a day and those reading less than once a day is very large at about 7 percent, one of the strongest effects found in the regressions. These results are consistent with two assumptions. First, there is less time for SAM to read to their children even if they do it frequently, the time could be of poor quality, given that these mothers work and have to compromise with domestic production, child-rearing and the demands of work. The other possible reason for these results is that unobserved preferences for investing in children are negatively correlated with preferences for work. This is crucial in terms of policy, because if self-selection is the reason to these findings, shifting policy towards generating incentives for SAM to stay home will not produce a positive increase in their children's scores.²⁵ The recent immigrant effect is very negative for children with LAM, however there are very few recent immigrants in this sample. The welfare effect is very large and negative for SAM. Finally, the income effect is not significant for both groups.

6.3 Results for Social Adjustment Indicators

6.3.1 Samples means

Table 4 presents the means of children's scores on instruments measuring problematic behaviours and pro-social behaviour for the full sample, a sample of TPC, a sample of SMC. The same samples are split up into samples of children with SAM and children with LAM. For the sample of 4 to 11 year old, the worst mean scores, by far except for pro-social behaviour, for all cases are obtained for children with single LAM. The best scores are obtained by TPC with LAM for HI, ED, IA, and for TPC in families with SAM, for CD and PB. However, scores for children in both types of families are very similar when children are in two-parent families, and children in single-mother families do much worse, on average, than children with two parents.

For the four and five year old, we notice that the younger children are more hyperactive, score higher for conduct disorders and exhibit less pro-social behaviour, however they score lower on emotional disorders and indirect aggression. More importantly, in three out of the five indicators, children with single SAM are on average worst off. Only for the case of ED, do children with single LAM have a higher score. For children with two parents, we observe that for two indicators, children with LAM have higher scores, HI and PS, the opposite is true for CD and IA while the mean is practically the same for ED. Therefore, it seems that younger children could possibly be affected by the absence of mothers in the home when they are young and when they are in single-mother families.

Table 4: Weighted Samples Means for 4- to 11-year-old and 4- and 5-year Old Children, Behavioural Scores (standard deviation in parentheses)

	Behavioural scores
Samples	Hyperactivity/inattention HI	Emotional disorder ED	Conduct disorder CD	Indirect aggression IA	Pro-social behaviour PS
4- to 11-year-old
Full sample	4.56 (3.59)	2.56 (2.59)	1.37 (1.85)	1.19 (1.69)	12.35 (3.89)
Mothers work full time	4.56 (3.57)	2.54 (2.53)	1.29 (1.77)	1.17 (1.66)	12.40 (3.86)
Mothers no work/part time	4.57 (3.63)	2.61 (2.68)	1.45 (1.98)	1.24 (1.75)	12.27 (3.92)
Two parents	4.40 (3.52)	2.43 (2.48)	1.27 (1.74)	1.12 (1.62)	12.39 (3.88)
Mothers working full time	4.47 (3.51)	2.46 (2.47)	1.25 (1.71)	1.14 (1.64)	12.43 (3.84)
Mothers no work/part time	4.28 (3.53)	2.39 (2.51)	1.31 (1.79)	1.10 (1.59)	12.31 (3.94)
Lone mothers	5.47 (3.85)	3.31 (2.99)	1.84 (2.32)	1.59 (2.00)	12.12 (3.92)
Mothers work full time	5.20 (3.87)	3.11 (2.86)	1.62 (2.09)	1.38 (1.76)	12.15 (4.00)
Mothers no work/part time	5.74 (3.81)	3.53 (3.11)	2.06 (2.52)	1.81 (2.19)	12.10 (3.83)
4- and 5-year-old
Full sample	4.90 (3.44)	2.10 (2.20)	1.57 (1.92)	0.79 (1.38)	11.17 (4.09)
Mothers work full time	4.98 (3.45)	2.05 (2.19)	1.55 (1.88)	0.79 (1.41)	10.98 (4.07)
Mothers no work/part time	4.81 (3.42)	2.15 (2.20)	1.60 (1.97)	0.79 (1.34)	11.40 (4.11)
Two parents	4.71 (3.40)	2.01 (2.16)	1.52 (1.85)	0.74 (1.33)	11.22 (4.14)
Mothers working full time	4.83 (3.44)	2.00 (2.16)	1.49 (1.81)	0.75 (1.39)	11.04 (4.05)
Mothers no work/part time	4.55 (3.35)	2.02 (2.15)	1.56 (1.91)	0.72 (1.23)	11.48 (4.25)
Lone mothers	6.00 (3.43)	2.62 (2.35)	1.86 (2.26)	1.09 (1.62)	10.86 (3.78)
Mothers work full time	6.23 (3.34)	2.50 (2.44)	2.04 (2.41)	1.13 (1.55)	10.51 (4.16)
Mothers no work/part time	5.86 (3.49)	2.69 (2.29)	1.75 (2.16)	1.07 (1.67)	11.10 (3.49)
Source: Micro-data from the NLSCY, cycle 1.

6.3.2 Regression results

Table A2, in the statistical appendix, displays the samples mean and standard deviation for the variables used in the regression analysis for behavioural scores. Since the results for SAM and LAM are similar, only the results for three samples (full, TPC and SMC) are presented in Table 5. For the sample of 4 and 5 year old the results are not as significant as for the full sample of 4- to 11-year-old children but are nevertheless presented for the full sample. The specifications are with the total family income variable. The specifications with the other income measure produce exactly the same effects as in the PPVT regressions, decreasing slightly the negative impacts of work on the outcomes.

Table 5: OLS Regression Results for 4- to 11-year-old and 4- and 5-year-old Children: Dependent Variables Behavioural Scores (t-ratios in parentheses)

	Full sample 4- to 11-year-old					Two-parent 4- to 11-year-old
	Hyperactivity/inattention HI	Emotional disorder ED	Conduct disorder CD	Indirect aggression IA	Pro-social behaviour PS	Hyperactivity/inattention HI	Emotional disorder ED	Conduct disorder CD	Indirect aggression IA	Pro-social behaviour PS
Child characteristics
Age	-0.137 (9.24)	0.111 (10.3)	-0.076 (9.48)	0.091 (12.5)	0.242 (15.0)	-0.141 (8.95)	0.106 (9.38)	-0.076 (9.23)	0.090 (11.8)	0.231 (13.3)
Male	1.381 (22.5)	0.098 (2.20)	0.556 (16.8)	-0.145 (4.85)	-1.452 (21.8)	1.364 (21.0)	0.075 (1.62)	0.501 (14.7)	-0.151 (4.82)	-1.471 (20.5)
Number of siblings	-0.213 (5.71)	-0.314 (11.5)	0.126 (6.23)	0.044 (2.39)	-0.338 (8.30)	-0.233 (5.92)	-0.330 (11.7)	0.104 (5.04)	0.030 (1.55)	-0.312 (7.16)
Birth order	-0.167 (3.38)	0.450 (12.6)	0.093 (3.49)	-0.050 (2.06)	0.313 (5.84)	-0.144 (2.79)	0.450 (12.2)	0.097 (3.57)	-0.045 (1.83)	0.336 (5.92)
Mother's characteristics
Age at child's birth	-0.033 (4.29)	-0.001 (0.11)	-0.018 (4.35)	-0.017 (4.42)	-0.019 (2.26)	-0.031 (3.77)	-0.003 (0.43)	-0.010 (4.25)	-0.017 (4.24)	-0.014 (1.52)
Years of education	-0.108 (6.69)	-0.002 (0.14)	-0.010 (1.18)	-0.027 (3.40)	0.103 (5.84)	-0.105 (6.20)	-0.003 (0.24)	-0.008 (0.92)	-0.023 (2.72)	0.092 (4.83)
Immigran1t(>9 years)	-0.276 (2.23)	-0.163 (1.82)	-0.260 (3.89)	-0.043 (0.71)	0.277 (2.06)	-0.141 (1.09)	-0.005 (0.57)	-0.202 (2.98)	-0.000 (0.01)	0.282 (1.97)
Immigrant2(5-9 years)	-0.539 (2.24)	-0.324 (1.86)	-0.467 (3.61)	-0.002 (0.02)	-0.419 (1.60)	-0.449 (1.79)	-0.186 (1.03)	-0.425 (3.12)	0.131 (1.07)	-0.379 (1.36)
Immigrant3(< 5 years)	-1.092 (3.59)	-0.624 (2.82)	-0.595 (3.61)	-0.180 (1.19)	-0.291 (0.87)	-1.012 (3.21)	-0.514 (2.28)	-0.525 (3.16)	-0.043 (0.21)	-0.397 (1.13)
Paid work	0.121 (1.75)	0.105 (2.10)	0.076 (2.05)	0.040 (1.18)	-0.025 (0.33)	0.125 (1.74)	0.085 (1.66)	0.075 (1.98)	0.042 (1.21)	-0.019 (0.24)
Spouse's characteristics
Years of education						-0.001 (0.49)	-0.007 (3.16)	-0.004 (2.32)	-0.003 (2.30)	0.004 (1.30)
Family characteristics
Family income (000)	-0.004 (3.55)	-0.002 (2.81)	-0.002 (2.90)	-0.000 (0.28)	-0.001 (1.02)	-0.004 (3.49)	-0.002 (2.66)	-0.002 (2.95)	-0.000 (0.41)	-0.001 (0.76)
Received welfare	0.583 (5.44)	0.472 (6.07)	0.284 (4.96)	0.289 (5.52)	0.076 (0.65)	0.606 (4.66)	0.451 (4.85)	0.264 (3.85)	0.223 (3.56)	-0.076 (0.11)
Step-family	0.925 (8.58)	0.451 (5.76)	0.166 (2.85)	0.223 (4.25)	0.031 (0.27)	0.932 (8.76)	0.455 (5.96)	0.174 (3.11)	0.242 (4.72)	0.037 (0.32)
Female-head family	0.644 (6.05)	0.594 (7.68)	0.385 (6.69)	0.302 (5.82)	-0.392 (3.37)
Area characteristics (*)
Newfoundland	-0.785 (5.44)	-0.762 (7.28)	-0.562 (7.20)	-0.134 (1.88)	-0.829 (5.27)	-0.694 (4.55)	-0.725 (6.63)	-0.500 (6.22)	-0.038 (0.51)	-0.770 (4.56)
Prince Edward Island	0.115 (0.65)	-0.176 (1.37)	-0.163 (1.71)	-0.120 (1.41)	-0.618 (3.24)	0.081 (0.44)	-0.288 (2.16)	-0.184 (1.88)	-0.029 (1.45)	-0.440 (2.15)
Nova Scotia	0.005 (0.38)	-0.102 (1.04)	-0.066 (0.90)	-0.011 (0.16)	0.201 (1.30)	0.036 (0.25)	-0.012 (0.12)	-0.052 (0.69)	0.040 (0.57)	0.349 (2.18)
New Brunswick	-0.325 (2.34)	-0.252 (2.49)	-0.245 (3.26)	-0.264 (3.90)	-0.167 (1.11)	-0.023 (1.59)	-0.197 (1.88)	-0.172 (2.25)	-0.234 (3.32)	-0.062 (0.39)
Quebec	0.406 (4.18)	0.200 (2.83)	-0.075 (1.44)	-0.070 (1.48)	-0.678 (6.41)	0.434 (4.20)	0.232 (3.14)	-0.014 (0.26)	-0.031 (0.63)	-0.599 (5.25)
Manitoba	-0.219 (1.66)	0.078 (0.81)	-0.005 (0.07)	-0.229 (3.54)	-0.320 (2.23)	-0.251 (1.81)	0.049 (0.49)	0.051 (0.70)	-0.211 (3.14)	-0.273 (1.77)
Saskatchewan	-0.027 (0.21)	-0.083 (0.91)	-0.010 (0.15)	0.027 (0.44)	-0.206 (1.51)	0.004 (0.03)	-0.049 (0.51)	0.035 (0.49)	0.064 (0.99)	-0.193 (1.31)
Alberta	-0.021 (0.18)	0.003 (0.04)	-0.008 (0.12)	-0.077 (1.33)	-0.249 (1.92)	-0.020 (0.16)	0.015 (0.16)	0.070 (1.07)	-0.041 (0.69)	-0.223 (1.62)
British Columbia	-0.020 (0.16)	0.130 (1.45)	0.021 (0.31)	-0.041 (0.68)	0.089 (0.67)	-0.001 (0.00)	0.127 (1.32)	0.046 (0.65)	0.011 (0.17)	0.102 (0.69)
Low income neigh.	-0.001 (0.34)	0.004 (1.52)	-0.002 (0.85)	0.006 (3.39)	0.012 (2.76)	0.000 (0.06)	0.004 (1.23)	-0.004 (1.75)	0.004 (2.04)	0.015 (3.05)
Constant	7.600 (24.6)	1.141 (5.08)	2.032 (12.2)	1.272 (8.41)	10.77 (32.1)	7.537 (22.8)	1.347 (5.69)	2.090 (12.0)	1.294 (8.10)	10.64 (29.2)
Sample size	12,329	12,342	12,312	11,939	12,070	10,579	10,587	10,566	10,249	10,372
Adjusted R-squared	0.095	0.063	0.064	0.046	0.080	0.085	0.053	0.054	0.035	0.082
	Lone-mothers 4- to 11-year-old					Full sample 4- to 11-year-old
	Hyperactivity/inattention HI	Emotional disorder ED	Conduct disorder CD	Indirect aggression IA	Pro-social behaviour PS	Hyperactivity/inattention HI	Emotional disorder ED	Conduct disorder CD	Indirect aggression IA	Pro-social behaviour PS
Child characteristics
Age	-0.122 (2.74)	0.136 (3.97)	-0.067 (2.52)	0.100 (4.36)	0.315 (7.05)	-0.548 (4.75)	0.117 (1.58)	-0.377 (4.93)	0.240 (5.02)	0.732 (5.33)
Male	1.491 (8.23)	0.244 (1.76)	0.901 (8.34)	-0.119 (1.20)	-1.342 (7.41)	0.920 (7.97)	-0.002 (0.03)	0.454 (6.63)	-0.078 (1.62)	-1.324 (9.69)
Number of siblings	-0.040 (0.59)	-0.182 (2.01)	0.271 (3.85)	0.149 (2.47)	-0.502 (4.26)	-0.236 (3.42)	-0.308 (7.05)	0.156 (3.86)	0.045 (1.58)	-0.538 (6.63)
Birth order	-0.346 (2.11)	0.413 (3.29)	0.050 (0.51)	-0.114 (1.35)	0.163 (1.08)	-0.057 (0.55)	0.642 (9.73)	0.070 (1.15)	-0.053 (1.24)	-0.850 (6.95)
Mother's characteristics
Age at child's birth	-0.046 (2.21)	0.003 (0.20)	-0.017 (1.39)	-0.018 (1.69)	-0.038 (1.83)	-0.064 (4.52)	0.006 (0.67)	-0.028 (3.31)	-0.016 (2.66)	-0.013 (0.79)
Years of education	-0.111 (2.26)	0.023 (0.62)	-0.007 (0.24)	-0.028 (1.10)	0.138 (2.81)	-0.069 (2.23)	-0.009 (0.47)	-0.003 (0.18)	-0.013 (0.98)	0.052 (1.41)
Immigrant1(>9 years)	-1.128 (3.11)	-0.877 (2.78)	-0.627 (2.56)	-0.307 (1.46)	0.250 (0.61)	-0.140 (0.58)	-0.131 (0.85)	-0.034 (0.24)	0.032 (0.32)	0.316 (1.09)
Immigrant2(5-9 years)	-1.130 (1.47)	-1.214 (2.06)	-0.765 (1.67)	-0.991 (2.57)	-0.616 (0.80)	-0.333 (0.82)	-0.344 (1.32)	-0.306 (1.27)	0.171 (1.00)	-1.160 (2.36)
Immigrant3(< 5 years)	-1.871 (1.69)	-1.577 (1.86)	-1.031 (1.56)	-1.205 (2.16)	0.878 (0.81)	-0.687 (1.17)	0.402 (1.07)	-0.262 (0.75)	0.107 (0.43)	-1.316 (1.86)
Paid work	0.039 (0.16)	0.217 (1.18)	-0.093 (0.65)	0.105 (0.86)	0.088 (0.37)	0.138 (1.07)	0.112 (1.36)	0.163 (2.13)	0.033 (0.62)	-0.195 (1.27)
Spouse's characteristics
Years of education
Family characteristics
Family income (000)	-0.002 (0.28)	0.002 (0.29)	-0.003 (0.61)	0.001 (0.22)	-0.000 (1.58)	-0.005 (1.90)	-0.004 (2.03)	-0.002 (1.09)	-0.001 (0.71)	-0.000 (0.07)
Received welfare	0.475 (1.92)	0.640 (3.38)	0.223 (1.51)	0.408 (3.23)	0.266 (1.08)	0.501 (2.56)	0.383 (3.06)	0.007 (0.06)	0.076 (0.93)	0.212 (0.91)
Step-family						0.472 (2.22)	0.357 (2.62)	0.082 (0.65)	0.039 (0.44)	-0.008 (0.03)
Female-head family						0.631 (3.04)	0.527 (3.97)	0.436 (3.54)	0.342 (3.98)	-0.541 (2.19)
Area characteristics (*)
Newfoundland	-1.422 (3.25)	-1.049 (3.14)	-0.961 (3.66)	-0.694 (3.05)	-1.125 (2.56)	-0.838 (2.92)	-0.754 (4.12)	-0.422 (2.48)	-0.138 (1.16)	-1.242 (3.64)
Prince Edward Island	0.448 (0.83)	0.710 (1.71)	0.052 (0.16)	-0.056 (0.21)	-1.643 (3.02)	0.558 (1.68)	-0.025 (0.12)	0.204 (1.04)	0.136 (1.00)	-0.807 (2.06)
Nova Scotia	-0.158 (0.43)	-0.528 (1.86)	-0.225 (1.02)	-0.361 (1.92)	-0.405 (1.10)	0.172 (0.68)	0.021 (0.13)	-0.134 (0.89)	0.067 (0.63)	-0.392 (1.29)
New Brunswick	-0.978 (2.28)	-0.601 (1.83)	-0.754 (2.93)	-0.414 (1.89)	-0.799 (1.86)	-0.010 (0.04)	-0.197 (1.19)	-0.148 (0.97)	-0.178 (1.67)	-0.409 (1.34)
Quebec	0.215 (0.73)	0.007 (0.03)	-0.455 (2.58)	-0.250 (1.65)	-1.121 (3.78)	0.302 (1.66)	0.094 (0.81)	0.153 (1.44)	0.000 (0.01)	-1.062 (4.90)
Manitoba	-0.060 (0.15)	0.299 (0.96)	-0.381 (1.57)	-0.391 (1.86)	-0.542 (1.33)	0.254 (1.01)	0.389 (2.41)	0.398 (2.67)	-0.176 (1.69)	-0.524 (1.76)
Saskatchewan	-0.217 (0.58)	-0.136 (0.48)	-0.261 (1.17)	-0.160 (0.83)	-0.310 (0.83)	0.332 (1.43)	0.118 (0.78)	0.272 (1.97)	-0.068 (0.71)	-0.244 (0.88)
Alberta	0.075 (0.19)	0.169 (0.55)	-0.460 (0.94)	-0.266 (1.32)	-0.509 (1.29)	0.153 (0.67)	-0.019 (0.13)	0.191 (1.42)	0.006 (0.07)	-0.530 (1.97)
British Columbia	0.026 (0.08)	0.174 (0.69)	-0.152 (0.75)	-0.297 (1.76)	-0.083 (0.25)	0.655 (2.86)	0.398 (2.71)	-0.502 (3.70)	0.041 (0.44)	-0.315 (1.15)
Low income neigh.	-0.001 (0.64)	0.006 (0.90)	-0.004 (0.73)	0.013 (2.86)	0.002 (0.25)	-0.299 (0.45)	0.067 (0.16)	-0.487 (1.22)	-0.283 (1.00)	-0.715 (0.88)
Constant	9.000 (9.80)	1.077 (1.53)	2.428 (4.43)	1.451 (3.09)	10.77 (11.7)	9.880 (13.1)	0.737 (1.52)	3.283 (7.32)	0.323 (1.03)	8.542 (9.49)
Sample size
Adjusted R-squared	1,750	1,755	1,746	1,690	1,698	3,255	3,262	3,250	3,181	3,174
	0.069	0.057	0.071	0.048	0.089	0.073	0.067	0.046	0.021	0.077
* Urban area variables are also included in the estimation.

First, for the full sample, it can be noted immediately that for HI, ED and CD children with SAM, all other things equal, have worst scores than children with LAM, and this difference is statistically significant. However, this difference is relatively small in comparison with other effects. For example, the family characteristics have a much stronger effect than work in the cases where the work dummy effect is statistically significant. For the case of HI, where the mean score is 4.56, the work dummy parameter is 0.121 while the welfare coefficient is 0.583, the step-family coefficient is 0.925 and the female headed family coefficient is 0.644. These characteristics have a strong and significant effect on all indicators, except for the pro-social indicator where only being in a female-headed family has a significant effect. The income effects, even when significant, are particularly weak. The strongest effect, in general, is the gender effect, in particular for HI, CD and PB, where these effects evaluated at their respective means are, 30, 41 and 12 percent. Indirect aggression is however used more frequently by girls and the estimated effect of being a girl is to increase by 12 percent the value this indicator. The other child characteristics in the regression are statistically significant for all indicators. The elasticities for age are respectively for, HI, ED, CD, IA, and PB, 0.22. 0.32, 0.41, 0.57, and 0.14. The effect of the number of siblings is also, for all cases, statistically significant. The elasticities are, respectively, 0.06, 0.16, 0.11, 0.04, and 0.03. Finally, birth order is also always statistically significant with elasticities of 0.05, 0.25, 0.10, 0.06, and 0.03. Therefore, the age of the child is second to sex in terms of the amplitude of the effects of child characteristics. Aging has beneficial effects on HI, CD and PB, the number of siblings has beneficial effects on HI, ED and birth order is beneficial for HI, IA and PB.

For the mother's characteristics, the age of the mother is significant for 4 out of 5 indicators, while years of education change significantly 3 out of the 5 indicators. For the age of the mother, the elasticities are for HI, CD, IA and PB, 0.21, 0.38, 0.41, and 0.04. For years of education, elasticities are for HI, IA, PB are 0.31, 0.29 and 0.11. An intriguing result is that children of immigrant mothers do quite well, particularly with recent immigrant mothers for HI, ED and CD. The differences between recent immigrant and Canadian born mothers is very large. For example, in the case of CD it is almost 50 percent. Finally, the low income neighbourhood variable is significant for IA and PS. It increases IA and PB.

The results are very similar for the sample of TPC. The spouse's education, an added explanatory variable, is significant for three indicators, and in all cases increases the child's outcomes. For the case of ED, the spouse's education has a significant impact even though the mother's does not. For the two other cases it is significant, the effect is however considerably smaller than the mother's. Controlling for the spouse's labour supply changed the results very little, since it was not significant the final specifications did not used the variable.

Surprisingly, for the sample of SMC, child characteristics, age of mother, years of education, immigration status of the mother coefficients are quite similar to the coefficients found with the sample of TPC. The major differences are that the work dummy and income effects are never significant for SMC. The means of the dependent variables for SMC left the impression that work was a determinant factor for these children.

• ²⁴Since the regression results are basically the same for the alternatives (mothers) work indicator variables (as for the effect associated with the spouse/partner work indicators - hours or weeks), they are not presented but can be obtained from the authors.
• ²⁵To check whether mothers with stronger preferences for work will invest less in their children, a regression was performed which included only children with mothers who did not work the preceding year and a dummy variable which took a value of 1 if the mother said she did not work because she wanted to stay with her children and 0 if the primary reason she was home was because she was laid off or could not find work. The dummy variable had no effect on PPVT scores. Therefore, it is possible that the frequency of reading effect is due to the fact that LAM mothers have more time to read to their children.