The Effect of Changes in Maternal Employment and Family Composition on Children's Behaviour - May 2002

Results of the analysis are divided into two sections. First an analysis is made of the maternal work experiences and family composition changes over the period 1994 to 1998 using the three cycles of the NLSCY available (Tables 2 to 9). The analysis uses OLS regression to assess the combined effect of these employment and family composition variables on children's behaviour cross-sectionally in 1994 and 1998 (Tables 10 and 11). Secondly, the analysis assesses the longitudinal effect of employment and composition changes on children's behaviour in 1994 and 1998, and changes in behaviour over the period (Tables 12 and 13). The behaviour scales used here are coded such that children with lower scores on the behavioural scales are considered to have better behaviour. Those children with relatively higher scores on the behavioural scales are considered to have worse behaviour. In order to aid in comparing results across the different behaviour scales, since each scale has a different upper limit, the three behaviour variables (Emotional disorder-anxiety, Conduct Disorder-Physical Aggression, Indirect Aggression) are rescaled to have a minimum of zero and a maximum of one.

4.1 Bivariate Results

Over the period from 1994 to 1998 in the NLSCY an increasing proportion of children's mothers were employed at the time of the interview (Table 2). From 64.3 per cent in 1994 the proportion of mothers employed at the time of the NLSCY interview increased to 69.8 per cent in 1996 and 78.1 per cent in 1998. A similar trend is observed for the proportion of children's mothers who were employed at any time in the 12 months prior to the NLSCY interview: from 71.8 per cent in 1994, to 76.0 per cent in 1996, and ultimately 82.9 per cent by 1998. The increases from one cycle to the next may be due to the fact that the collection period for the NLSCY coincided with the recovery period following the recession earlier in the decade. The changing number of employed mothers may simply have been mothers who had lost their jobs earlier in the decade but who were now regaining entry into the labour market. Another explanation is that the trend reflects an increasingly older cohort of mothers with children of school age (6 to 7 years in 1994) who may be re-entering the labour market after caring for these children while they were younger.

Table 2 Cross-sectional Comparisons of Maternal Work Experience (1994, 1996 & 1998)

Selected Maternal Work Variables	1994	1996	1998
Employed at the time of annual interview (%)	64.3	69.8	78.1
Employed at any time in the last 12 months (%)	71.8	76.0	82.9
Employed in the last 12 months - average weeks worked	46.2	44.3	45.5
Average Number of hours worked per week	19.8	23.8	27.3
N=1333
Source: National Longitudinal Survey of Children and Youth, 1994 to 1998

The average number of weeks worked, for those mothers who had been employed in the previous 12 months, did not change much over the period from 1994 to 1998 (Table 2). In fact most of those mothers who had been employed in the 12 months prior to the interview had been employed for most of the year. For those mothers who were employed in 1994 there was a slight decrease from 46.2 weeks of employment on average to 44.3 weeks in 1996, before rising again in 1998 to 45.5 weeks. However, the average number of hours worked each week for this cohort of mothers increased from 19.8 hours per week on average in 1994 to 27.3 hours per week by 1998. Again, this may be due in part, to the better economic environment in 1998 or to a greater number of these mothers entering the workforce due to their children being older and attending school, or some combination of the two.

Mothers who were employed in each of the three cycles of the NLSCY tended to have more highly ranked occupations (using the Pineo index) (Table 3). Mothers who were employed only in one or two of the cycles had relatively lower than average Pineo scores. The Pineo scores for those employed over three years increased from 1994 to 1998, while the scores for those employed for only one or two of the years decreased. Those mothers who were employed in each of the three cycles of the NLSCY tended to have relatively higher occupations (in terms of responsibilities and skills) on average, than mothers who were not.

Table 3 Maternal Average Pineo Difference Score by Number of Years Employed (1994, 1996 & 1998)

Number of Years Mother Employed (Period Covering 1994, 1996 & 1998)	Average Pineo Difference Scores1
	1994	1996	1998
None	0.00	0.00	0.00
One	-0.39	-0.30	-1.21
Two	-0.45	-1.80	-1.38
Three	0.13	0.58	0.60
1 — Scores are average deviations from mean Pineo score for each year.
N=1333
Source: National Longitudinal Survey of Children and Youth, 1994 to 1998

Approximately 17.5 per cent of mothers in the labour force, experienced one or more periods of unemployment of more than 4 weeks in the 12 months prior to the interview in 1994 (Table 4). This figure increases somewhat to 19 per cent in 1996, before falling quite dramatically to 11.0 per cent in 1998. The decline in the proportion of mothers experiencing periods of unemployment may, again, be an indication of improving labour market opportunities later in the period analysed. A small proportion of mothers experienced multiple periods of unemployment over the period (less than 3 per cent in each NLSCY cycle).

Table 4 Periods of Maternal Unemployment of more than 4 weeks (1994, 1996 & 1998)

Periods of Maternal Unemployment of more than 4 weeks within a year (%)	1994	1996	1998
None	82.5	80.9	89.0
One	14.7	15.6	8.6
Two	2.7	2.0	1.5
Three	0.1u	1.4u	0.9u
Total	100.0	100.0	100.0
u - these data do not meet Statistics Canada's quality standards. Conclusions based on these data will be unreliable, and most likely invalid.
N=1333
Source: National Longitudinal Survey of Children and Youth, 1994 to 1998

This situation alters somewhat when observing the number of periods of unemployment over time for the whole of the interval rather than in each discrete year. Over one-third of children's mothers experienced at least one episode of unemployment of 4 weeks or more from 1994 to 1998 (Table 5). Just over 13 per cent of the mothers (13.09 per cent) had two or more periods of unemployment, and 6.2 per cent experienced three or more periods of unemployment from 1994 to 1998. We see, therefore, that a considerable proportion of mothers encountered changes in their employment situation over the period. Many mothers only experienced a single period of unemployment in any one year, but these did not seem to be isolated events. A large proportion of mothers have experienced continual periods of unemployment with any attendant effects that these might have on the family as a whole and most particularly on their children.

Table 5 Periods of Maternal Unemployment of more than 4 weeks over time (1994 to 1998)

Periods of Maternal Unemployment of more than 4 weeks (1994 to 1998) (%)
None	63.92
One	22.99
Two	6.90
Three	4.59
Four	1.19*
Five or more	0.40u
Total	100.00*
u — these data do not meet Statistics Canada's quality standards. Conclusions based on these data will be unreliable, and most likely invalid.
N=1333
Source: National Longitudinal Survey of Children and Youth, 1994 to 1998

For those mothers who were employed at any time during the years 1994, 1996 or 1998, the average number of weeks worked (34.7 weeks) was considerably lower than the cross-sectional results (Table 6). Observing the total number of weeks worked on average over the period (104.1 weeks) suggests a large number of mothers have a strong attachment to the labour market, at least with regard to the number of weeks they work during the period. In fact, the majority of the mothers (63.0 per cent) were employed in each of the years 1994, 1996 and 1998, while approximately 12 per cent (11.8 per cent) had not been employed in any of these years (Table 7). As well, there is a trend of an increasing average number of weeks worked as the number of years employed increases. Those mothers who worked in each of the three cycles were employed for the majority of the year — 47.3 weeks on average.

Table 6 Longitudinal Comparisons of Maternal Work Experience (1994, 1996 & 1996)

Mothers with any Employment 1994 to 1998	Mean	S.D.
Total Weeks Employed	104.1	57.8
Average weeks worked in years employed	34.7	19.3
N=1333
Source: National Longitudinal Survey of Children and Youth, 1994 to 1998

 

Table 7 Maternal Number of Years Employed and Average Weeks Employed (1994, 1996 & 1998)

Maternal Number of Years Employed (1994, 1996 & 1998)	Percent	Average Weeks worked in years employed
None	11.8	n/a
One	9.8	10.0
Two	15.4	25.8
Three	63.0	47.3
Total	100.0	34.7
N=1333
Source: National Longitudinal Survey of Children and Youth, 1994 to 1998

Family composition patterns over the same time period show quite interesting results (Table 8). Over four-fifths (82.22 per cent) of the children were in biological mother and father families in 1994. By 1998 this figure was slightly lower, but still very high, at 78.08 per cent. As the proportion of children in biological mother/father families decreased over the period from 1994 to 1998, the proportion in biological mother and step-father families rose from 2.71 to 5.56 per cent, while the proportion in lone mother families stayed relatively stable — 13.08 to 13.15 per cent.

Table 8 Cross-sectional Family Composition (Children aged 6 to 7 years in 1994)

Cross-sectional Family Composition	1994 (6-7 years)	1996 (8-9 years)	1998 (10-11 years)
Biological mother and father present	82.22	80.67	78.08
Biological mother and step-father	2.71	3.84	5.56
Biological father but no biological mother	0.66u	1.33u	2.22
Two parents (exclusive of categories above)	1.25u	1.08u	0.92u
Lone Parent Mother	13.08	13.03	13.15
Other	0.08u	0.04u	0.07u
Total	100.00	100.00	100.00
u — these data do not meet Statistics Canada's quality standards. Conclusions based on these data will be unreliable, and most likely invalid.
N=1333
Source: National Longitudinal Survey of Children and Youth, 1994 to 1998

As observed above, there were relatively small but nonetheless important shifts in family composition when comparing cross-sectional changes over the period from 1994 to 1998. However, what is most interesting is the extent to which the same children's family composition remained stable or changed over the time interval. Around nine in ten (88.9 per cent) children lived in the same family situation in 1998 as they did in 1994 (Table 9). Approximately eight in ten (77.91 per cent) children remained in family situations where both biological parents were present in 1994 and 1998. Almost one in twelve children (7.69 per cent) remained in lone mother families, 2.09 per cent remained in families with a biological mother and step-father, and 1.21 per cent remained in some other stable family composition from 1994 to 1998.

Table 9 Family Composition Over Time (1994 to 1998)

	Per cent of total
Stable
Both biological parents present	77.91
Lone Parent Mother and no Father	7.69
Biological mother and step-father	2.09
Other adults present	1.21u
Change
Biological father year 1, step-father year 5	0.15u
Enter Lone Parent Mother	3.95
Exit Lone Parent Mother	3.54
Varies between single mother and other adults	0.99u
One other change	2.30
Two or more changes	0.15u
Total	100.00
u — these data do not meet Statistics Canada's quality standards. Conclusions based on these data will be unreliable, and most likely invalid.
N=1333
Source: National Longitudinal Survey of Children and Youth, 1994 to 1998

Of the approximately 11.0 per cent of children who experienced some form of change in their family situation, the largest proportion lived in families that went from a two parent to a lone mother situation (3.95 per cent). An almost equal proportion of children (3.54 per cent) lived in families that changed from a single mother situation to a two parent situation. Quite a large proportion of children living in other family situations experienced changes in their family structure — about 2.45 per cent experienced one, or even two or more changes in family composition over the period from 1994 to 1998. It should be noted that these changes only capture the compositional changes which are detailed by the data collection of the NLSCY. Changes which took place in 1995 and 1997 are not necessarily reflected in these data.

4.2 Multivariate Models

4.2.1 Contemporaneous Effects

As observed, there were some changes in the employment situations of children's mothers over the period from 1994 to 1998, and also some changes in the family composition of a small proportion of children. What is needed is to assess to what degree the changes in mother's employment and family composition have occurred together, as well as an assessment of the effect of these changes on children's behaviour. That is, we want to investigate the consequence for children's behaviour scores if we simultaneously control for maternal employment and family composition variables. Using regression analysis we investigate the effect of maternal employment and family composition on our three behavioural scales — emotional disorder-anxiety, conduct disorder-physical aggression, and indirect aggression — for 1994 and 1998 (Table 10 contains the descriptive statistics, and Table 11 the regression coefficients). The analysis uses two sets of equations. First variables are entered which related to our two conceptual areas of interest, namely a number of variables relating to mother's employment and family composition, with controls for the gender and health status of the child. For the second equation further controls are added relating to important parental characteristics identified in our literature review such as mother's education, age and immigration status, and information relating to spousal employment.

Table 10 Cross-sectional Multivariate Regression Descriptive Statistics, Children aged 6-7 in 1994.

	1994		1998
	Mean	S.D.	Mean	S.D.
Maternal Employment Characteristics
Pineo score	0.00	3.85	0.00	4.09
Mother employed	0.64	0.48	0.78	0.41
Number of hours worked per week	19.85	18.73	27.27	17.42
Family Composition Patterns
Two Biological Parent (reference)	0.84	0.37	0.79	0.41
Biological mother and step-father family	0.03	0.16	0.06	0.23
Lone Mother family	0.13	0.34	0.13	0.34
Other families	0.02	0.14	0.03	0.18
Gender of the Child (0=Male)	0.48	0.50	0.48	0.50
Children with Health Problems (0=Very Good to Excellent health)	0.12	0.32	0.10	0.31
Mother's Education
Less than High School	0.14	0.35	0.11	0.31
High School (reference)	0.21	0.41	0.21	0.40
More than High School	0.65	0.48	0.68	0.46
Immigrant Status of Mother (0=non-immigrant)	0.15	0.36	0.14	0.35
Mother's Age	35.13	5.12	39.00	5.24
Spousal Employment Patterns
Pineo Score	0.58	0.49	0.68	0.47
Spouse Employed	0.00	3.69	0.00	3.72
Number of hours worked per week	16.68	18.50	23.52	18.67
Dependent Variables
Emotional-Disorder Anxiety	0.15	0.14	0.16	0.15
Conduct Disorder-Physical Aggression	0.11	0.15	0.10	0.13
Indirect Aggression	0.11	0.17	0.11	0.17
N=1333
Source: National Longitudinal Survey of Children and Youth, 1994 to 1998

Mother's employment is measured using variables measuring whether the mother is currently employed, their Pineo socio-economic classification (as mentioned above, this is a Pineo difference score variable), and the number of hours per week they are employed. Family composition is calculated using four dummy variables. Comparisons are made to the reference category, that is, families with both biological parents present. The other variables include a measure of lone mother families, families with the biological mother and step-father present, and lastly all other family situations. Mother's education is determined by three dummy variables measuring education in three groups: less than high school, high school graduation (the reference category), and education beyond high school.

Results from the first regression equations for 1994 and 1998 (see Equation 1 in Table 11) show that children with mothers who have higher scores on the Pineo socio-economic classification tend to have better (i.e. lower) behavioural scores; statistically significantly so for each behavioural scale in 1994. Children in families where the mother is currently employed tended to have worse (i.e. higher) scores for behaviours associated with conduct disorder-physical aggression and indirect aggression in 1994 and 1998. Higher number of hours per week worked by mothers is associated with higher emotional disorder-anxiety behaviour scale scores in 1994 and higher indirect aggression scores in 1998, but with lower behaviour scale scores for conduct disorder-physical aggression in 1994. These results differ somewhat from those found by Menaghan, Mott and Cooksey (1997). Maternal employment characteristics in their study were associated with lower behavioural scores for children's oppositional action (1997).

The results for family composition for 1994 show that children who are not in two biological parent situations tend to have higher scores on the behavioural scales, although this changes somewhat for 1998. In particular, children living in lone mother situations in 1994 had significantly higher emotional disorder-anxiety scores than children in two biological parent family situations (Equation 1). Children in biological mother and step-father families, and those in other parental situations, also obtain higher scores on average for emotional disorder-anxiety, as well as conduct disorder-physical aggression. The most interesting results are for indirect aggression where children in non-two biological parent situations in 1994 were statistically significantly more likely to have higher scores. Results for 1998 are similar to those found for 1994, with some important differences. The results for indirect aggression are now only significant for biological mother and step-father families. Children in lone mother families tend to now experience better (i.e. lower) behavioural scores than two biological parent families for emotional disorder-anxiety and conduct disorder-physical aggression, significantly so for conduct disorder-physical aggression. Our results for family composition are in the same general direction and somewhat similar to those found by Menaghan, Mott and Cooksey (1997).

For the remaining variables in Equation 1, children with poorer health were more likely than children with better health to have significantly higher scores on each of emotional disorder-anxiety and conduct disorder-physical aggression in both 1994 and 1998. However, children with poorer health had slightly lower scores for indirect aggression in 1994 but significantly higher scores in 1998, compared to children with very good to excellent health. Girls were more likely to have statistically significantly lower conduct disorder-physical aggression behavioural scale scores than boys in both 1994 and 1998, but significantly higher indirect aggression scores in 1998.

Overall, we see that maternal employment, family composition and other variables accounted for some, but relatively little, of the variation in children's behavioural scale scores in 1994 and 1998, with the exception of indirect aggression. The initial Equation 1 model in Table 11 accounted for approximately 7.5 per cent of the variation (adjusted) in 1994 and 5 per cent (adjusted) in 1998 in the variation of children's indirect aggression scores. Aspects of family composition and maternal employment are observed in Equation 1 to be associated with changes in children's indirect aggression, and to a lesser degree emotional disorder-anxiety, and conduct disorder-physical aggression.

The addition of further controls in Equation 2 (Table 11) for each behavioural scale somewhat changes the results observed above. Mother's occupational characteristics measured with the Pineo socio-economic classification remain negatively associated with children's behavioural scale scores in 1994, significantly so for conduct disorder-physical aggression and indirect aggression. However, in 1998 there is no association between mother's Pineo difference score and children's behavioural scores. The children of mothers who are currently employed no longer have statistically significantly higher (i.e. worse) conduct disorder-physical aggression or indirect aggression scores in 1994. Similarly, the number of hours that mothers work is no longer significantly related with children's behavioural outcomes in 1994 or 1998.

Family composition figures for Equation 2 are consistent with those for Equation 1, although in some cases they are no longer significantly related to children's behaviour. Children living in either lone mother, biological mother and step-father, or in other family situations, have higher (i.e. worse) scores on average for each of the three dependent behavioural variables than children in two biological parent families in 1994. In particular, children in lone mother or in other family situations, have statistically significantly worse indirect aggression scores than children in two biological parent families in 1994. Family composition is still a significant factor affecting conduct disorder-physical aggression in 1998 — children in lone mother families are significantly more likely to have better scores, and those in biological mother and step-father families are significantly more likely to have worse scores.

Girls still tended to have significantly better (i.e., lower) conduct disorder-physical aggression behaviour scores than boys after the application of further controls in both 1994 and 1998. However, girls continued to have significantly worse (i.e. higher) scores than boys on indirect aggression in 1998. Children's health is not significantly related to emotional disorder-anxiety or to conduct disorder-physical aggression in 1994 after the application of further controls. This is not the case in 1998, where children's health is still statistically significantly related with worse scores (i.e. higher scores) on each of the behaviour scales.

The control variables for Equation 2 form some interesting relationships with our dependent variables. Mother's highest level of education had a significant impact on children's indirect aggression in 1994, and emotional disorder-anxiety and conduct disorder-physical aggression in 1998. Children of mothers with less than high school education scored significantly worse on indirect aggression (1994) and on emotional disorder-anxiety (1998), while children of mothers with more than a high school education scored significantly worse on conduct disorder-physical aggression (1998). Children of immigrant mothers had significantly lower conduct disorder-physical aggression scores in 1994, and also significantly lower indirect aggression scores in 1998. The age of a child's mother is seen to be negatively associated with children's behavioural scores, but only significantly so for emotional disorder-anxiety and conduct disorder-physical aggression in 1994, and conduct disorder-physical aggression and indirect aggression in 1998. The spousal employment variables are significantly negatively related to emotional disorder-anxiety in 1998. But here it is the fact that a spouse is employed in a higher socio-economic classification score occupation, rather than the number of hours worked or the fact that they are employed, which affects children's behaviour to a greater extent. However, the higher the Pineo difference score the more likely children are to experience worse conduct disorder-physical aggression scores (in 1994) and the greater the number of hours worked by the spouse the more likely the child will have higher indirect aggression scores (in 1998).

The application of the control variables has tended to increase, quite substantially in one case, the proportion of variation which the independent variables explain in the dependent behavioural variables, but lowers it in another (Table 11). The regression models now explain approximately 6 per cent of the variation in emotional disorder-anxiety in 1994, and 4 per cent in 1998. The proportion of variation in conduct disorder-physical aggression is still very low in 1994 (less than 2 per cent) but is slightly higher in 1998 (approximately 3 per cent). The amount of variation explained for indirect aggression in 1998 remained stable at approximately 5 per cent. The largest change in explained variation is the drop in the proportion of indirect aggression, by about 5 percentage points (from 7.5 per cent in 1994 to 2.4 per cent in 1998).

4.2.2 Longitudinal Effects

For the longitudinal analysis two major sets of variables were constructed to measure change in maternal employment and family composition from 1994 to 1998. As mentioned above, the dependent variables in the longitudinal analyses include the children's 1998 behavioural variable scores as well as a measure of behavioural change for each scale — measured as the difference in their scale score between 1994 and 1998. The children's scale score in 1994 for the relevant behavioural variable are included as a control when investigating the change in behaviour from 1994 to 1998³. The descriptive statistics of this analysis are presented in Table 12 and the coefficients in Table 13.

Table 12 Longitudinal Multivariate Regression Descriptive Statistics, Children aged 6-7 in 1994

	Mean	S.D.
Maternal Employment Patterns
Change in Pineo Score 1994-1998	0.22	3.05
For those employed in 1998
No transitions in employment status	0.53	0.50
Some transitions in employment status	0.24	0.43
High transitions in employment status	0.06	0.24
For those not employed in 1998
No transitions in employment status (reference)	0.11	0.31
Some to High transitions in employment status	0.06	0.24
Family Composition Patterns
Stable
Two Biological Parent (reference)	0.78	0.41
Lone Parent Mother	0.08	0.27
Biological Mother, step-father	0.02	0.14
Other family	0.01	0.11
Change
Two Biological Parent to biological mother and step-father	0.001	0.04
Enter Lone Parent Mother	0.04	0.20
Exit Lone Parent Mother	0.04	0.19
Varies between Lone Parent Mother and Two Parent	0.01	0.10
One other change	0.02	0.15
Two other changes	0.00	0.04
Gender of the child (0=Male)	0.48	0.50
Children with Health Problems (0=Very Good to Excellent health)	0.10	0.30
Mother's Education
Less than High School	0.11	0.31
High School (reference)	0.21	0.40
More than High School	0.68	0.46
Immigrant Status of Mother (0=non-immigrant)	0.14	0.35
Mother's Age (1998)	39.00	5.24
Dependent Variables
Change in Emotional-Disorder Anxiety 1994-1998	0.02	0.16
Change in Conduct Disorder-Physical Aggression to 1994-1998	-0.02	0.14
Change in Indirect Aggression 1994-1998	0.00	0.18
N=1333
Source: National Longitudinal Survey of Children and Youth, 1994 to 1998

 

Table 13 Multivariate Regression of Behavioural Scale Scores at ages 10-11 (1988) and Change from 1994, Predicted from Long-Term Patterns of Maternal Employment and family Composition

Emotional-Disorder Anxiety
	Equation 1	Equation 2	Equation 3 Change (94-98)
Maternal Employment Patterns
Change in Pineo Score 1994-1998	-0.003b	-0.003b	-0.004a
For those employed in 1998
No transitions in employment status	0.007	0.016	0.010
Some transitions in employment status	0.011	0.020	0.009
High transitions in employment status	0.059a	0.062a	0.041b
For those not employed in 1998
No transitions in employment status (reference)	--	--	--
Some to High transitions in employment status	0.026	0.041b	0.012
Family Composition Patterns
Stable
Two Biological Parent (reference)	--	--	--
Lone Parent Mother	0.006	-0.008	-0.019c
Biological Mother, step-father	-0.037	-0.043c	-0.052b
Other family	0.036	0.035	0.010
Change
Two Biological Parent to biological mother and step-father	-0.003	0.002	-0.005
Enter Lone Parent Mother	0.060a	0.050a	0.048a
Exit Lone Parent Mother	-0.002	-0.011	0.002
Varies between Lone Parent Mother and Two Parent	0.013	0.015	0.016
One or more other changes	0.033	0.025	0.018
Gender of the child (0=Male)		0.005	0.003
Children with Health Problems (0=Very Good to Excellent health)		0.0071a	0.062a
Mother's Education
Less than High School		0.030b	0.023c
High School (reference)		--	--
More than High School		0.002	-0.007
Immigrant Status of Mother (0=non-immigrant)		-0.007	-0.013
Mother's Age (1998)		-0.001	0.000
1994 Behaviour Score			-0.584a
R2	0.043	0.078	0.253
R2 (adjusted)	0.031	0.061	0.239
Conduct Disorder-Physical Aggression
Maternal Employment Patterns
Change in Pineo Score 1994-1998	0.002b	0.002c	-0.001
For those employed in 1998
No transitions in employment status	0.008	0.016	0.014c
Some transitions in employment status	0.035a	0.034a	0.020c
High transitions in employment status	0.000	0.003	0.002
For those not employed in 1998
No transitions in employment status (reference)	--	--	--
Some to High transitions in employment status	0.046a	0.055a	0.048a
Family Composition Patterns
Stable
Two Biological Parent (reference)	--	--	--
Lone Parent Mother	0.004	-0.007	-0.016
Biological Mother, step-father	-0.033	-0.037c	-0.030c
Other family	0.063	0.050c	-0.008
Change
Two Biological Parent to biological mother and step-father	0.000	-0.011	-0.007
Enter Lone Parent Mother	0.042b	0.043b	0.030b
Exit Lone Parent Mother	-0.017	-0.029c	-0.011
Varies between Lone Parent Mother and Two Parent	-0.020	-0.034	-0.015
One or more other changes	-0.010	-0.003	-0.020
Gender of the child (0=Male)		-0.040a	-0.011c
Children with Health Problems (0=Very Good to Excellent health)		0.055a	0.032a
Mother's Education
Less than High School		0.023b	0.018c
High School (reference)		--	--
More than High School		0.026a	0.015b
Immigrant Status of Mother (0=non-immigrant)		-0.010	-0.001
Mother's Age (1998)		-0.002a	-0.001c
1994 Behaviour Score			-0.555a
R2	0.034	0.071	0.335
R2 (adjusted)	0.022	0.053	0.322
Indirect Aggression
Maternal Employment Patterns
Change in Pineo Score 1994-1998	0.002	0.001	-0.001
For those employed in 1998
No transitions in employment status	0.027c	0.036a	0.026b
Some transitions in employment status	0.046a	0.054a	0.037a
High transitions in employment status	0.045b	0.040c	0.039b
For those not employed in 1998
No transitions in employment status (reference)	--	--	--
Some to High transitions in employment status	-0.001	0.026	-0.001
Family Composition Patterns
Stable
Two Biological Parent (reference)	--	--	--
Lone Parent Mother	0.007	-0.004	-0.023c
Biological Mother, step-father	-0.045c	-0.062b	-0.084a
Other family	0.131	0.125	0.072b
Change
Two Biological Parent to biological mother and step-father	0.077	0.070	0.094
Enter Lone Parent Mother	0.062a	0.045b	0.034c
Exit Lone Parent Mother	0.060b	0.036c	0.057a
Varies between Lone Parent Mother and Two Parent	-0.008	-0.005	-0.020
One or more other changes	-0.036	-0.054c	-0.053b
Gender of the child (0=Male)		0.040a	0.036a
Children with Health Problems (0=Very Good to Excellent health)		0.048a	0.042a
Mother's Education
Less than High School		0.010	0.000
High School (reference)		--	--
More than High School		-0.004	-0.016
Immigrant Status of Mother (0=non-immigrant)		-0.053a	-0.058a
Mother's Age (1998)		-0.004a	-0.003a
1994 Behaviour Score			-0.560a
R2	0.027	0.028	0.314
R2 (adjusted)	0.016	0.014	0.300
a = p<0.001, b = p<0.01, c = p<0.05, two-tailed tests
N=1333
Source: National Longitudinal Survey of Children and Youth, 1994 to 1998

Information presented in Table 13 consists of three equations. In the first equation for each of the dependent behavioural variables measured in 1998, the relationship between mother's employment change and family composition change and the dependent behaviour scales is observed, without any controls for significant mother and child characteristics. In the second equation controls are included for important mother and child characteristics, including child's health status, gender of child, mother's age, and highest level of education and immigrant status, but still using the three dependent variables as measured in 1998. The third and final equation includes all of the previous controls as well as the time 1 (i.e. 1994) behaviour scores as a control to observe the relationship between mother's changing employment situations and changing family composition, to changes in children's behavioural outcomes. The dependent variable is simply the difference or change between children's 1994 behavioural scale score and their score on the corresponding behavioural scale in 1998.

To capture change in mother's occupation and skill level their Pineo difference score in 1994 is subtracted from that in 1998. Changes in mother's employment are divided into two broad areas, those employed (defined as one or more weeks of work) and those not employed in 1998. Conceptually it would have been preferable to have divided each of these two areas into three groups: those who experienced no change in employment status from 1994 to 1998, those who experienced a low number of transitions (one or two stretches of looking for employment), and finally those who experienced many transitions (three or more stretches of looking for employment). However, due to sample size problems it was necessary to combine the categories of mothers who experience some and those who experience high levels of transition in employment status but who were not employed in 1998. This yields five categories for the regression model. The reference category for the analysis is those mothers with no employment and no transitions over the period from 1994 to 1998.

Family composition is divided into two broad groups — stable and changed (see Table 9). Stable family composition includes: both biological parents present — the reference category; lone parent mothers; two parent families where the biological mother is present; and other families where at least one or more adults are present. Changed family composition is conceptually divided into six categories (see Table 9) but again sample size problems have lead to the creation of five categories: families that included two biological parents in 1994 but became biological mother and a step-father in 1998; families which became single biological mother from 1994 to 1998; families that started as single mother in 1994 but which by 1998 had exited this situation; families which varied between single mother and other adults present (such as two parent families) over the period from 1994 to 1998; other family composition change situations not listed above which changed one or more times over the period from 1994 to 1998.

Considering the results from the data (Table 13) before the controls and after their application (Equations 1 and 2 only) we see that the children of mothers who increased their Pineo socioeconomic classification between 1994 and 1998 were significantly more likely to have lower emotional disorder-anxiety and conduct disorder-physical aggression behavioural scores in 1998 than children of mothers who had decreased Pineo scores. Even after controlling for mother's and children's characteristics the relationship is still significant. However, an increasing Pineo socio-economic classification from 1994 to 1998 is associated with statistically significant higher conduct disorder-physical aggression behavioural scores, both before and after the application of controls. This result is interesting given that Menaghan, Mott and Cooksey (1997) have found that mother's increasing occupational complexity over time is negatively associated with children's oppositional action behaviour.

The changing employment status of mothers over the period from 1994 to 1998 has a somewhat different effect on children's behavioural scale scores. Children's mothers who were employed in 1998 and who experienced a high number of transitions in employment status from 1994 to 1998, scored significantly higher (worse) emotional disorder-anxiety scores than children of mothers who were not employed in 1998 and who had experienced no employment transition (i.e. they were outside the labour force for each of 1994 and 1998) — even after the application of controls (Table 13). Both children of mothers who were employed in 1998 and experienced some transitions in employment status from 1994 to 1998, as well as children of mothers not employed in 1998 but who experienced some or a high number of transitions between 1994 and 1998 in employment status, obtained statistically significant higher conduct disorder-physical aggression scores (both before and after applying controls) than children of mothers who were not employed in 1998 and who had experienced no employment transition. Turning to indirect aggression behaviour, we see that children of mothers in each of the three categories of those employed in 1998 (No, Some and High Transitions) had statistically significant higher (worse) scores before and after applying control variables, compared to children whose mothers did not work in 1998 and had experienced no employment status transitions from 1994 to 1998.

These results suggest that mother's employment status may be a factor in affecting children's behavioural outcomes. When compared to children of mothers who are not employed and experience no transitions — children of mothers who experience a high degree of transition in employment may have worse emotional disorder-anxiety and indirect aggression scores, and children of mothers that experience some changes in employment status and are currently employed, or who are not currently employed but have experienced some transitions in employment status, tend to have relatively worse conduct disorder-physical aggression scores. This result differs from that of Menaghan, Mott and Cooksey (1997) where mother's employment was seen to reduce children's oppositional action behaviour compared to mothers that were not employed.

A possible reason for why our results differ from the existing literature is that the behaviour scales we use, while similar in many ways to those employed in the USA, such as children's oppositional action as used by Menaghan, Mott and Cooksey (1997), they are not identical. Mothers who are employed and experiencing many employment status changes may not be able to respond to the children's emotional and anxiety behaviours to the same degree due to time and other employment pressures — particularly if they are in a more constant state of transition seeking employment.

Turning to changes in family composition, similarly mixed results are observed. Children in stable lone biological mother and step-father families had significantly lower (better) emotional disorder-anxiety scores than children in stable two biological parent families but only after applying control variables (Equation 2, Table 13). However, children in families that became lone parent mother families had significantly worse emotional disorder-anxiety scores both before and after applying controls compared to children in stable two biological parent families.

After applying control variables, children in stable biological mother and step-father families, and those in families that exited from lone female parent families scored statistically significantly lower (better) than children in stable two biological parent families on the conduct disorder-physical aggression scale. Children in families that entered lone parent mother situations also scored significantly worse conduct disorder-physical aggression scores, both before and after applying controls compared to children in stable two biological parent families, while children in stable other family situations also scored significantly worse.

Children in stable biological mother and step-father families have significantly better (lower) indirect aggression scores than children in stable two biological parent families. However, children in families which change, either exiting or entering lone parent mother families, are observed to have worse indirect aggression scores compared to children in stable two biological parent families. Family composition seems to be important for affecting children's behavioural scale scores, particularly stable biological mother and step-father, as well as lone parent mother families.

Regarding the control variables in Table 13, gender is a factor in that girls are more likely to have better conduct disorder-physical aggression scores, but worse indirect aggression scores than boys. Children with lower health are observed to have significantly worse behavioural scale scores for each of the three dependent variables compared to children with very good or excellent health. Children of mothers with less than a high school education score significantly worse on emotional disorder-anxiety and conduct disorder-physical aggression compared to children of mothers with just a high school education. However, children of mothers with more than a high school education score worse on the conduct disorder-physical aggression scale when compared to children of mothers with just a high school education. If the child's mother was an immigrant to Canada they are significantly more likely to have lower (better) indirect aggression. As well, children of mothers who are older are significantly more likely to have lower conduct disorder-physical aggression and indirect aggression scores (in 1998) compared to children of younger mothers.

Observing the proportion of variation which our regressions explained, we see that applying controls increases the proportion for two of our dependent variables (Table 13). Emotional disorder-anxiety explained variation increases from 3.1 per cent to 6.1 per cent, conduct disorder-physical aggression increases from 2.2 per cent to 5.3 per cent, while the proportion of explained variation for indirect aggression remains relatively steady dropping slightly from 1.6 per cent to 1.4 per cent. Nonetheless, despite these increases the explained variation for each regression is still quite low.

The analysis to this point has relied upon comparisons between mother's employment change and family composition change with children's behaviour in discrete years. Yet, what if the children in 1998 who experienced relatively higher behaviour scores, did not do so in 1994? The results discussed so far have not captured whether and if so how, children's behavioural change may vary with changes in mother's employment status and family composition change. To measure the extent of change in behaviour it makes sense to have as the dependent variable a measure of this change covering the period from 1994 to 1998. As discussed above, children's behaviour scores from 1994 are included as a control of unmeasured variables on the new dependent behavioural change variable (see endnote 3). Results for the variables which measure change over time in children's behaviour are observed for each behaviour scale in Equation 3 (Table 13).

Results show that for most of the relationships which had been significant after the application of control variables in Equation 2, there still are many which are significantly related to the dependent behavioural variables in the form of behaviour changes over time in Equation 3. For example, as shown, the Pineo difference scores are significantly negatively related to emotional disorder-anxiety in 1998. The result for the Pineo difference scores and changes in emotional disorder-anxiety from 1994 to 1998 in Equation 3 is also significantly and negatively related. This means that children of mothers with higher Pineo difference scores are more likely to experience a decrease in their emotional disorder-anxiety score than children of mothers with lower Pineo difference scores.

Mother's employment transition relationships identified in Equation 2 remained significantly related to the dependent behavioural variables in Equation 3, when compared to mothers not employed in 1998 and experiencing no transitions from 1994 to 1998. Children of mothers who were employed in 1998 and experienced high transitions in employment status remained positively and significantly related to changes in emotional disorder-anxiety, while children of employed mothers in 1998 with some transitions remained positively and significantly related to conduct disorder-physical aggression changes from 1994 to 1998. All mothers employed in 1998, whether experiencing no, some or high transitions in employment status from 1994 to 1998, were also still positively and significantly related to changes in indirect aggression. As well, the relationship between mothers not employed in 1998, but who experienced some or high transition in employment status from 1994 to 1998, and conduct disorder-physical aggression also remained positive and significant.

However, in two cases the original relationship in Equation 2 was no longer significant in Equation 3: change in Pineo difference scores is no longer significantly related to change in conduct disorder-physical aggression, and mothers not employed in 1998 and having some or high transitions in employment status are no longer significantly related to emotional disorder-anxiety. This suggests that in these instances the scores for these particular children were already high in 1994. That is, these children likely already had high scale scores on the behavioural variables before the change in maternal employment. In the case of the relationships which remained significant, the stability in the significance suggests that the changes in mothers employment may very well have contributed to changes in children's behaviour scores.

Turning to family composition we see that stable family situations of biological mother and stepfather are still significantly related to lower (better) scores for emotional disorder-anxiety, conduct disorder-physical aggression, and indirect aggression. Additionally, stable other family situations is no longer significantly related to higher (worse) scores on conduct disorder-physical aggression, but is now significantly related to higher scores on indirect aggression. Stable lone parent mother situations for emotional disorder-anxiety and indirect aggression which were not significant in Equations 1 or 2, are now significantly and negatively related to changes in these two variables. That is, children in stable lone parent mother families experience a much greater lowering of their behaviour scale scores from 1994 to 1998 than children of stable two parent families.

Children in family situations where there is an entry into a lone parent mother family are still positively significantly related to changes in each of the three behavioural variables. As well, children in families which exit from a lone parent mother situation are positively and significantly related to indirect aggression, but are no longer significantly related to conduct disorder-physical aggression. Children in families that experience one or more changes in the composition of the family are also still negatively significantly related to changes in indirect aggression.

The relationships for the control variables which were significant in Equation 2 remain significant in Equation 3 and in the same direction, for child's gender, health, mother's education, mother's age and mother's immigration status (Table 13). The shift to a regression that captures change over the period of time from 1994 to 1998 shows that except for a few cases, most of the relationships that had been identified in Equation 2 as significant remained so in Equation 3. Change in the magnitude of some coefficients, such as maternal education, have been thought to indicate that their influence on children's behavioural development rests in the children's early years (less than 6 years old), with diminishing importance afterward (Menaghan, Mott and Cooksey, 1997). Our results, at least for maternal education lend some support to this since all of the regression coefficients from Equation 2 to Equation 3 become smaller, although only the relationships for conduct disorder-physical aggression and emotional disorder-anxiety are significant. Since Equation 3 explicitly looks at the changes in behaviour over time, it may mean that most of the effect of maternal attributes is being applied earlier than the period under investigation, and is relatively smaller over the period we observe.

• ³Menaghan, Mott and Cooksey (1997) use the same technique to investigate changes over time in family composition and maternal employment. They argue that by controlling for a behavioural variable in year 1 researchers are able to ascertain the extent to which children's higher behavioural scores in a later year were in fact already present at the earlier point in time. This will be observed as a reduction or elimination of the effects in the coefficients. Menaghan, Mott and Cooksey (1997) base their method on the work of Kessler and Greenberg (1981). Kessler and Greenberg (1981: 7-20) contend that in the absence of measurement error of some variable (X), the inclusion of a measure in time 1 (X₁) to predict a measure in time 2 (X₂) permits researchers to interpret the regression coefficients as measuring the effect on the amount of change in the dependent variable (in this case X₂).