Keeping Contact with Children: Assessing the Father/Child Post-separation Relationship from the Male Perspective

IV. FACTORS AFFECTING THE FREQUENCY OF FATHER/CHILD CONTACT

The aim of this section is to identify the net effect of factors influencing the frequency of contact fathers have with their children after separation. Our objective, in other words, is to measure the net effect of certain variables, such as the child's age at separation or the socio-economic situation of fathers, on the amount of time fathers spend with children once all the variables under consideration have been controlled for.

The fact that fathers can have several children whom they see at different frequencies gives the data a particular structure (termed "hierarchical"). The dependent variable (the frequency of father/child contact) whose variation we are attempting to explain is measured separately for each child. Certain of the independent variables included in the analysis are measured for each child (such as the child's age at separation), while others concern the characteristics of fathers (such as his level of education). The traditional regression methods cannot be used to analyze this type of data without introducing a statistical bias. Consequently, we employed a multi-level regression analysis.[8]

In the first analysis, we explored the extent to which father/child contact varies according to the socio-demographic characteristics of fathers and their children. In the second, we attempted to measure to what point the fathers' behaviour is linked to their attitudes and perceptions towards their paternal role, once their socio-demographic characteristics and those of their children are controlled for.

A. SOCIO-DEMOGRAPHIC CHARACTERISTICS

Our findings, presented in earlier sections of this report, influenced the choice of variables to be included in the multi-level regression analysis. The following "child" characteristics were selected: sex, age at survey, age at separation, time elapsed since separation, parents' conjugal situation at the child's birth and the distance between the parents' respective homes (see Table 16). The "father" characteristics include: age at survey, education, work pattern, and subsequent family history: forming a new union, having another child or living with a new partner's children. The "work pattern" variable combines information gathered at the time of the survey, and classifies fathers into one of four categories: 1) those who did not work during the year preceding the survey; 2) those who worked part time;[9] 3) those with full-time employment and day-time hours (the reference category in the equation); and 4) those with full-time employment who regularly worked evenings, nights or weekends.

Table 16 presents the regression coefficients associated with the socio-demographic characteristics of fathers and children, as well as a series of other statistics. The constant in model 1 corresponds to the average value of the dependent variable, and is equal to the square root of the average number of days fathers spent with their children. Once squared, the number indicates that each child spent an average of 70 days a year with his or her father. Most of the observed variation in the number of days fathers and children spent together is located at the father level:[10] 78 percent of the total variation in the number of days comes from differences between the fathers; consequently, only 22 percent of the variation is situated at the child level. In other words, the analysis reveals that fathers have much the same behaviour with each of their children and that differences observed between children of the same father are relatively small.

Model 4 integrates both the father and the child characteristics. The regression coefficients presented for this model indicate the net effect that each of the independent variables has on the time fathers spent with children when the father and child characteristics are both controlled for. A negative coefficient indicates that the variable in question decreased the amount of contact while a positive coefficient increased it.

Examining the coefficients associated with the child variables shows that the frequency of father/child contact is positively linked to the child's age at the time of the survey (a coefficient of 0.214). In other words, fathers have more contact with older children and this effect remains even when the length of time since separation was taken into account. Moreover, contact tends to decrease as the time since separation (or since birth, in the case of children born outside a union) lengthens, and this decline occurs at a slightly quicker rate than the increase of contact linked to children's age (standardised coefficients of -0.192 compared with +0.179; data not shown). The distance separating the parents' homes also has a significant influence on the frequency of contact between fathers and children. Thus, when mothers and fathers live 50 kilometres or more apart, children see their father much less often than when the distance separating the two households is less than 10 kilometres (reference group in the equation).

The child's sex does not seem to have any significant link with the time children spend with fathers, once the other socio-demographic characteristics have been controlled. Likewise, the conjugal situation of parents at the birth of the child has no significant impact on the frequency of contact. The link between these variables observed earlier may have been caused by differences in the age of children or in the time elapsed since separation rather than by differences in the type of union itself. For instance, children born within a cohabiting union were perhaps younger when their parents separated than were those born to parents who had married directly, which would partly explain why they saw their father less frequently.

Table 16: Effect of a number of socio-demographic characteristics (including the children's age at the time of the survey and the time elapsed since separation) on the time that separated fathers spend with their children (Coefficients γ of the multi-level regression model)a
Variable b Model 1 Model 2 Model 3 Model 4
Constant 8.328 8.879 -11.230 -8.984
Child variables
Sex (boys)   0.173   -0.061
Age at time of survey   0.219 *   0.214 *
Time elapsed since separation   -0.263 *   -0.342 *
Type of union at birth (marriage)
- Out of union   -0.078   0.160
- Marriage preceded by cohabitation   -1.232   -1.011
- Cohabitation   -0.279   0.073
Distance (< 10 km)
- 10 - 49 km   0.100   -0.390
- 50 - 499 km   -2.010 *   -1.827 *
- 400 km and over   -3.278 *   -3.063 *
Father variables
Age     1.167 * 1.094 *
(Age) 2     -0.015 * -0.014 *
Highest level of education (Secondary)
- < Secondary     -1.158 -1.211
- Post-secondary     -0.280 -0.123
- University     -1.398 -0.959
Employment (Full-time/day)
- Unemployed     -2.450 -1.598
- Part-time     -2.654 * -2.634 *
- Full-time/evening, night or weekend     -1.140 -1.318
New union (no)     -0.204 0.536
Child born in new union (no)     -1.516 -0.440
Stepchild (no)     0.640 0.428
Variance - level 2 26.16 23.36 22.93 20.33
Variance - level 1 7.24 6.82 7.24 6.86
Deviance 2747.25 2711.99 2730.75 2696.44
χ 2 (degrees of freedom) n.a. 35.26 (9) 16.51 (11) 50.80 (21)
R 2 2 n.a. 0.10 0.10 0.20
R 2 1 n.a. 0.10 0.10 0.18

Source: Statistics Canada, General Social Survey (GSS), Cycle 10: the Family, 1995.

Among the father characteristics, only their age at the time of the survey and the work pattern in the year preceding the survey appear to be significantly linked to the degree of father/child contact. Neither education, nor a more recent union, and having new children or living with step-children within this union, show any significant influence on the amount of time fathers and children spend together.

Fathers' age has a non-linear effect on frequency of contact. The regression coefficients attached to the age and to the age-squared are both significant. The first coefficient is positive, showing that the number of days spent with children increases as fathers' age at survey increases, up to a certain age; after this age, contact levels start to decline, as indicated by the negative coefficient for the age-squared. Constructing the curve attached to these regression coefficients shows that contacts increase until the age of 39 years, and then start to decrease beyond this age.

Moreover, fathers working part-time spend less time with their children than fathers employed full-time during the day (coefficient of -2.634 for the former compared with 0 for the latter). One might have expected the opposite result, on the assumption that fathers who work part-time would have more free time to devote to their children. This finding may, however, reflect the income effect. Fathers with only part-time work are likely to have a lower average employment income than those working full-time. As mentioned earlier, fathers with higher incomes tend to see their children more regularly, and this could explain the association between part-time work and less father/child contact. The small sample size,[11] however, makes it impossible to ascertain to what extent this association is caused by income variations, since full-time workers generally have higher incomes, or by differences in the fathers' availability, since part-time work is often characterised by atypical working hours.

The children's age at the time of the survey, their age at separation, and the time elapsed since the separation are three variables closely linked to one another; they cannot therefore be included simultaneously in the equation. To evaluate the relative importance of each of these three variables on the frequency of father/child contact, tables 17 and 18 repeat the analysis presented in Table 16, but with a change in the variables in the equation.

Table 17: Effect of a number of socio-demographic characteristics (including the children’s age at the time of the survey and at the time of separation) on the time that separated fathers spend with their children (Coefficients γ of the multi-level regression model)a
Variable b Model 1 Model 2 Model 3 Model 4
Constant 8.328 6.986 -11.230 -10.470
Child variables
Sex (boys)   0.184   -0.016
Age at time of survey   -0.021   -0.099
Age at separation   1.627 *   1.504 *
(Age at separation) 2   -0.231 *   -0.200 *
(Age at separation) 3   0.010   0.009 *
Type of union at birth (marriage)
- Out of union   1.059   1.148
- Marriage preceded by cohabitation   -1.546   -1.273
- Cohabitation   -0.074   0.240
Distance (< 10 km)
- 10 - 49 km   0.261   -0.211
- 50 - 499 km   -1.910 *   -1.758 *
- 400 km and over   -3.002 *   2 -2.855 *
Father variables
Age     1.167 * 1.081 *
(Age) 2     -0.015 * -0.014 *
Highest level of education (Secondary)
- < Secondary     -1.158 -1.320
- Post-secondary     -0.280 -0.253
- University     -1.398 -0.908
Employment (Full-time/day)
- Unemployed     -2.450 -1.443
- Part-time     -2.654 * -2.340 *
- Full-time/evening, night or weekend     -1.140 -1.169
New union (no)     -0.204 0.380
Child born in new union (no)     -1.516 -0.445
Stepchild (no)     0.640 0.343
Variance - level 2 26.16 22.81 22.93 20.15
Variance - level 1 7.24 6.60 7.24 6.63
Deviance 2747.25 2709.43 2730.75 2692.91
2 (degrees of freedom) n.a. 37.82 (11) 16.51 (11) 54.33 (22)
R 2 2 n.a. 0.12 0.10 0.21
R 2 1 n.a. 0.12 0.10 0.20

Source: Statistics Canada, General Social Survey (GSS), Cycle 10: the Family, 1995.

Table 18: Effect of a number of socio-demographic characteristics (including the children's age at the time of separation and the time elapsed since separation) on the time that separated fathers spend with their children (Coefficients γ of the multi-level regression model) a
Variable b Model 1 Model 2 Model 3 Model 4
Constant 8.328 6.986 -11.230 -10.470
Child variables
Sex (boys)   0.184   -0.016
Age at time of survey   -0.021   -0.099
Age at separation   1.627 *   1.504 *
(Age at separation) 2   -0.231 *   -0.200 *
(Age at separation) 3   0.010   0.009 *
Type of union at birth (marriage)
- Out of union   1.059   1.148
- Marriage preceded by cohabitation   -1.546   -1.273
- Cohabitation   -0.074   0.240
Distance (< 10 km)
- 10 - 49 km   0.261   -0.211
- 50 - 499 km   -1.910 *   -1.758 *
- 400 km and over   -3.002 *   -2.855 *
Father variables
Age     1.167 * 1.081 *
(Age) 2     -0.015 * -0.014 *
Highest level of education (Secondary)
- < Secondary     -1.158 -1.320
- Post-secondary     -0.280 -0.253
- University     -1.398 -0.908
Employment (Full-time/day)
- Unemployed     -2.450 -1.443
- Part-time     -2.654 * -2.340 *
- Full-time/evening, night or weekend     -1.140 -1.169
New union (no)     -0.204 0.380
Child born in new union (no)     -1.516 -0.445
Stepchild (no)     0.640 0.343
Variance - level 2 26.16 22.81 22.93 20.15
Variance - level 1 7.24 6.60 7.24 6.63
Deviance 2747.25 2709.43 2730.75 2692.91
2 (degrees of freedom) n.a. 37.82 (11) 16.51 (11) 54.33 (22)
R 2 2 n.a. 0.12 0.10 0.21
R 2 1 n.a. 0.12 0.10 0.20

Source: Statistics Canada, General Social Survey (GSS), Cycle 10: the Family, 1995.

Comparing the results of the two tables shows the importance of the child's age at separation for the contact they have subsequently with their father. When the child's age at separation is introduced into the model, the effect associated with the time elapsed since the separation (Table 18) and with the child's age at the time of the survey (Table 17) is no longer significant, underlining the crucial role of the point in time at which parental separation occurs in children's lives.

It should be noted, however, that the effect of the child's age at separation is not linear. To correctly model the link between age at separation and frequency of contact, we added two variables representing the age at separation squared and cubed. The first coefficient is positive, indicating at first view a positive relation between children's age at separation and the number of days spent with their father; at a particular age, this relationship is inversed and becomes negative (the coefficient of the age squared is negative) before becoming positive once again (the coefficient of the age cubed is positive). When the time since separation is controlled for (Table 18), the curve established using the regression parameters shows that the frequency of father/child contact rises as the child's age at separation increases until approximately 5.5 years; it then decreases slightly until the age of 10 years, from which point the amount of contact starts to increase sharply again. This result is hardly unexpected. During the early years of life, one might assume that maintaining father/child contact after separation is first and foremost linked to the length of time during which a father created close ties with his child. Subsequently, between 5 and 10 years, contact remains at a relatively stable level, with the rise observed from the age of 10 years probably reflecting a combination of the concrete links children established with fathers before the separation and their greater autonomy in the decision to see their father. Finally, the effect of the other variables in the analysis remains more or less stable from one table to the other, as does the proportion of the variation explained by the model.

B. Attitudes to, and perceptions of, the paternal role

In this second step, we repeated the analysis, but integrated the elements of men's perception of the paternal role shown earlier to be linked to the frequency of father/child contact. The first of these is a variable measured for the children, opposing fathers who declared themselves satisfied with the existing custody arrangements (where and with whom the child lived) with those who were not satisfied. Also included is a series of variables measured for each father. The first combines information concerning how close fathers felt to their own father during childhood and how they compare with him in terms of their paternal role, organized into four categories:1) the reference category, comprising fathers who were close to their own father during childhood but who feel that they themselves are better fathers;2) fathers who were close to their father but who do not consider themselves better;3) fathers who were not close to their own father and who feel they are better fathers; and 4) fathers who were not very close to their father,[12] but do not feel they are better fathers (see Table 19). The other variables included in the model contrast:1) fathers who strongly disagree with the statement that "everyday tasks linked to children are not principally men's responsibility" with the other fathers (including those without an opinion); 2) fathers who felt that "the fact of having children made them happier" with those who did not; and last 3) fathers who were satisfied with the "time spent in general with the children" with those who were dissatisfied.

The results of this analysis, with the child's age at separation and the time elapsed since separation included in the model, are presented in Table 19. First, including the variables linked to fathers' perception of their paternal role increases the proportion of variation explained between fathers and between children, from around 20 percent to 30 percent (compare the R2 in model 4 of tables 18 and 19). Introducing these variables into the analysis, however, hardly changes the impact of fathers' and children's socio-demographic characteristics on the amount of time they spend together. Thus, the child's age at separation, the father's age at the time of the survey and his work pattern remain significantly linked to the level of father/child contact. However, the distance separating the parents' domiciles is no longer significant once the attitude and perception variables are added to the model. This is because fathers who live far from their children and who see them less often are also less likely to declare themselves happy to have had children and satisfied with the time they spend with them in general.

Table 19: Effect of a number of socio-demographic characteristics (including the children's age at the time of separation and the time elapsed since separation) on the time that separated fathers spend with their children (Coefficients γ of the multi-level regression model)a
Variable b Model 1 Model 2 Model 3 Model 4
Constant 8.528 5.360 -8.783 -11.230
Child variables
Sex (boys)   0.309   -0.056
Time elapsed since separation   -0.059   -0.121
Age at separation   1.444 *   1.156 *
(Age at separation) 2   -0.215 *   -0.173 *
(Age at separation) 3   0.009 *   0.008 *
Type of union at birth (marriage)
- Out of union   1.501   1.267
- Marriage preceded by cohabitation   -1.059   -1.351
- Cohabitation   0.593   1.080
Distance (< 10 km)
- 10 - 49 km   0.297   -0.032
- 50 - 499 km   -1.458   -0.561
- 400 km and over   -2.997 *   -1.562
Satisfied with custody (no)   2.499 *   1.446 *
Father variables
Age     0.802 * 0.876 *
(Age) 2     -0.011 * -0.012 *
Highest level of education (Secondary)
- < Secondary     -1.112 -1.111
- Post-secondary     -0.664 -0.575
- University     -1.517 -1.193
Employment (Full-time/day)
- Unemployed     -0.726 -0.190
- Part-time     -3.103 * -2.538 *
- Full-time/evening, night or weekend     -1.278 -0.965
New union (no)     0.120 0.541
Child born in new union (no)     -1.103 -0.725
Stepchild (no)     -0.497 -0.717
Relationship with own father
(very close / better)
- Very close / not better     -0.730 -1.126
- Not very close / better     0.406 0.102
- Not very close / not better     -0.956 -1.171
Not responsible for childcare tasks (strongly disagree)     1.581 * 0.847
Happy to have had child (no)     2.495 * 2.293 *
Satisfied with time with child (no)     3.843 * 3.001 *
Variance - level 2 25.49 21.76 17.60 16.61
Variance - level 1 7.68 6.79 7.66 6.93
Deviance 2533.76 2488.43 2472.08 2447.13
χ 2 (degrees of freedom) n.a. 45.33 (12) 61.68 (17) 86.60 (29)
R 2 2 n.a. 0.14 0.25 0.30
R 2 1 n.a. 0.14 0.24 0.29

Source: Statistics Canada, General Social Survey (GSS), Cycle 10: the Family, 1995.

The results presented in Table 19 are as we expected. It is impossible, however, to draw any conclusion as to the direction of the observed relation between fathers' attitudes and perceptions and the amount of contact they have with their children: is the satisfaction expressed by fathers concerning custody arrangements or the fact of having had children, for example, the cause or the consequence of how much time they spend with their children? In other words, we cannot determine how far the degree of satisfaction that fathers express about their paternal role has a direct influence on amount of father/child contact or if, on the contrary, this satisfaction is the direct result of the number of days they spend with them. To do this, it would be necessary to carry out further analyses that are outside the framework of the present study.


Date modified: