1The fertility level in France is among the highest in Europe (behind Ireland), thanks largely to the relatively high frequency of families with at least three children, now referred to as “familles nombreuses” (large families). For example, by producing 2.12 children on average, the women born in 1960 have replaced themselves. The contribution of births of order three or above to this fertility is far from insignificant, standing at around 0.5 children per woman, i.e. 24% of the total. And yet in all developed countries, the two-child family model is by far the most common today, though there are recent signs of a slight downward trend. Under this model, the progression from one child to two is very frequent, while the progression from two to three is much rarer. Even in France, where large families are still quite common, only a minority of families have a third child. For example, 90% of women born in 1960 founded a family by producing at least one child; around 80% of these women went on to have a second child, of whom only 44% had a third child. In turn, only 30% of those with three children continued to enlarge their family. So the birth of the second child represents a clear cut-off point, with only a minority of families deciding to have more children. In total, fewer than one woman in three (32%) has at least three children, whereas 72% have at least two.
2This article aims to examine the changes since the 1970s in the proportions of fathers and mothers of two children who have had at least one additional child (or parity progression ratios to the third child) and to analyse the characteristics of these parents who have chosen not to conform to the two-child family norm. By pinpointing the factors which influence the decision to have or, on the contrary, to not have a third child, this research encompasses several issues that are central to the debate among demographers and other specialists interested in the determinants of fertility. These issues concern, in particular, the link between the frequency of third births and the parents’ social status and/or level of education (Kravdal, 1992a; Hoem, 1996; Toulemon, 1995; Ekert-Jaffé et al., 2002), their employment status (Wright et al. 1988, Hoem and Hoem 1989, Kravdal 1992b, Corman 2000), repartnering after union dissolution (Vikat et al. 1999), or the sex of the first two children (Pollard, Morgan, 2002). This research also provides an opportunity to examine the efficacy of French family policy, which specifically encourages births of order three and above by offering more generous welfare benefits and special advantages to large families.
3After presenting some of the main components of French family policy specific to families of three children or more, we will add our contribution to the debate on the efficacy of this targeted approach by comparing the French situation with that of several other European countries and by observing changes in the frequency of progression from two to three children since the 1970s. We will then analyse the characteristics of men and women with a third child and compare them with those who did not enlarge their family after a second birth in order to highlight the factors associated with the birth of a third child.
4The 1999 Study of Family History survey (EHF) was the main data source used for our study. Its large size was a first key advantage, since it enabled us to study sub-populations (Table A in the Appendix). Moreover, the survey contains the main socioeconomic variables generally regarded as determinants of fertility (educational attainment, social status, nationality, etc.)
I – Is French family policy effective in encouraging third births?
1 – Measures targeting families with at least three children
5French family policy has always sought to promote births of order three and above by offering more generous support to families with at least three children. For example, family allowance entitlements have always begun with the birth of the second child, and the amount received increases for the third and following children: in 1946, the monthly allowance paid out for each child of order three or above was more than 50% higher than that paid out for the child of order two [1]. Allowance rates have been adjusted several times since then, generally to the benefit of children of order three and above: in 1970, the difference between allowance entitlements for the second child and those for the third (and fourth) was increased to 68%.
6But targeting of family policy reached its peak at the end of Valéry Giscard d’Estaing’s term as French president:
- in 1978 [2], a new means-tested family income supplement was introduced for families with a child aged below three and for those with three children or more. This new allowance, more generous than those it replaced [3], was received by a larger number of middle-income families, thanks to a higher income ceiling;
- in 1979, three major new measures were announced: for children born from July 1980, the maternity allowance was substantially increased for the third and following children, and set at the symbolic threshold of 10,000 francs [4]; maternity leave for the mother was raised from 16 to 26 weeks; and from the 1980 tax return, dependent children of order three were counted as one unit (instead of 0.5 units previously) for calculation of the “quotient familial” [5] used to determine taxable income;
- lastly, the difference between the family allowances paid out for the second child and those paid out for the third was increased twice more, once in 1979 and once in 1980, up to 83%. However, entitlements for the fourth and following children fell slightly below those of the third child [6].
7Hence in the space of a few years, the “advantages” associated with the third child were substantially reduced. In 1985 however, the creation of the Allocation parentale d’éducation (parental leave allowance, APE), paid to the mother or father if they stop or reduce their working activity to raise a third child, provided a further specific incentive for a third child. However, following its extension to parents of two children in 1994, it no longer targets this specific objective. All in all, though 1978-82 can be seen as the period when third births were most strongly encouraged, the period 1985-94 followed a similar pattern, with the introduction of the APE in 1985 and its improvement in 1986 and 1987 [9], before its extension to parents of two children in 1994.
8Today, three family policy measures still specifically target families with at least three dependent children:
- the family allowance differential, which is now only 28% and which no longer favours the third child with respect to those of higher birth orders;
- the family income supplement, which is still paid to many low- and middle-income families with at least three dependent children [10];
- a method of calculating of taxable income which counts dependent children of order three and above as a whole unit (see fn. 5); this “advantage” mainly benefits higher-income families (those liable to income tax under the French system), though it also enables middle-income families to fall below the threshold for tax liability [11].
2 – Large families are a key factor in fertility differences between countries
9Table 1 shows fertility by birth order of women of the 1960 cohort in the eleven countries of Western Europe for which data is available. The countries were classified by level of completed fertility, which appears to be more closely linked to the number of births of order three and above than to the number of births of orders one and two.
Completed fertility of the 1960 cohort and distribution by birth order in eleven Western European countries (number of children per woman)
Completed fertility of the 1960 cohort and distribution by birth order in eleven Western European countries (number of children per woman)
10This table shows that France, where completed fertility is highest [13] (2.12 children per woman), is the country where the contribution of births of order three and above is the largest (0.5 children per woman) and Italy the country where these two values are the lowest (respectively 1.66 and 0.21). Moreover, the classification of countries by level of high birth order fertility follows pratically the same ranking as that of completed fertility, though this is certainly not the case for the sum of first and second order fertilities. For example, the ranking of Spain, Denmark and Greece by completed fertility does not correspond to their fertility ranking for birth orders one and two. In these three countries, childlessness is quite rare and the majority of mothers have a second child (respectively 71% in Spain, 79% in Denmark and 82% in Greece). But few mothers wish to enlarge their family any further [14], and fertility at higher birth orders is relatively low, resulting in quite low completed fertility. Conversely, England-Wales and Finland achieve completed fertility of almost two children per women, despite relatively low first and second order fertility (respectively 1.51 and 1.50 children per woman), thanks to fertility at birth orders three and above that ranks among the highest in the list (0.47). A high level of childlessness (19% of women in England-Wales and 17% in Finland) is partly offset by a relatively high frequency of families with three or more children. Through a sort of “specialization” process, some women opt out of motherhood while others produce large numbers of children (Ekert-Jaffé et al., 2002).
11Hence, fertility at birth orders three and above appears to be a determining factor for the level of completed fertility. For the eleven countries observed, these two variables are closely correlated (rs = 0.83) and almost two-thirds of the difference between the levels of completed fertility in France (2.12 children per woman) and in Italy (1.66) can be attributed to births of order three or above (0.29/0.46).
12So if one aim of family policy is to maintain fertility levels, it therefore makes sense to encourage births of order three or above. Note, however, that fertility at birth order three and above in Norway, Sweden, England-Wales and Finland reaches levels similar to those of France without the help of a specially targeted policy [15]. But in France, would these births still be as frequent without such specific encouragement?
3 – Have these measures been effective?
13To evaluate the effectiveness of specific measures to encourage third births over the periods 1978-82 and 1985-94, we will compare the change in parity progression ratios (PPR) of women with two children (A2) observed since the 1970s with that of women with one child (A1) (Figure 1). Each PPR is represented by two curves: the thin lines represent the period PPRs, a2 and a1, which are the sum of duration-specific fertility rates for third (respectively second) order births by time elapsed since the second (first) order birth [16]; the thick lines represent the PPRs A2 and A1 in the parity cohorts, i.e. the proportion of women who have a third (respectively second) child among those who had a second (first) child in a given year,—or parity two (parity one) cohorts. The parity cohort PPRs have been shifted by 4 years (or 3.5 years) so that they can be shown on the same graph. This shift represents the approximate mean interval between second and third (or first and second) births in the 1975-85 cohorts [17].
Parity progression ratios at parity one and parity two since 1970. Period (a1 and a2) and cohort (A1 and A2) parity progression ratios
Parity progression ratios at parity one and parity two since 1970. Period (a1 and a2) and cohort (A1 and A2) parity progression ratios
14The variations in the cohort PPR A2 reflect those of the frequency (or intensity) of progression to a third child among women with two children, and comparison of these variations with those of the period PPR a2 reveals the changes in fertility timing: whenever the period PPR is below that of the cohorts, this indicates a postponement, i.e. a prolongation of the interval between the second and third child and conversely, when a2 exceeds A2, this signifies a shortening of the interval [18]. Since the 1970s, the periods of postponement have been much longer than periods of anticipation: it is only in the periods 1980-82 and, to a lesser extent, 1989-90, that a transient reversal of the trend towards a longer interval between the second and third birth is observed.
15This postponement is especially marked between 1973 and 1979 and is accompanied by a rapid decline in the frequency of progression from a second to a third child: within a few short years, the PPR A2 fell from 57% (1966 cohort) to 46% (1973 cohort). The signs of a renewed upturn in the period PPR coincides with the introduction of the family policy to encourage third births. And this policy appears to have worked, since a2 increased dramatically between 1978 and 1981 (from 40 to 52%), while the interval between births shortened for a while and the PPR A2 rose to 47.6% in the 1975 to 1977 cohorts. Over the period 1983-85, when family policy was less specifically targeted, the birth interval started to grow again and A2 began to fall. However, the progressive introduction of parental leave appears to have provided a second stimulus for third births: in the 1983-84 cohorts A2 saw a further upturn and the interval between births shortened temporarily once more. But within a few years, intervals between births started increasing again and A2 dropped rapidly: from the 1987 cohort, the PPR fell below 45% then decreased slowly to a level slightly under 44% in the 1990-92 cohorts [19].
16Last, an increase in the cohort PPR A2 is observed during both of the periods in which family policy specifically encouraged third births, indicating that this policy did indeed have a positive impact. Moreover, when this explicit third-child incentive was abolished, the PPR started falling again. But before concluding that the French third-child policy is effective, we must ensure that these trends are specific to third births.
17Comparison of changes in the PPR at parity two with the PPR at parity one suggests that this conclusion should be qualified, since several similarities are observed (Figure 1). Despite several time lags of one or two years, the period PPRs a1 and a2 often move in the same direction, proving that the periods favourable or unfavourable to family enlargement are generally the same, whatever the parity already reached. But the targeting of family policy on the third child does appear to have had a specific effect on a2, since it varies much more than a1 over the period 1979-83. For example, during this period, only the interval between the second and third birth shortens temporarily, while the trend towards longer intervals between the first and second birth simply tails off for a while. Moreover, the second period of upturn in a1 (between 1984 and 1987) is shorter than that of a2 (1984-89), probably due to the introduction of the parental leave allowance. The extension of this allowance to children of birth order two gives a clear boost to a1, the period PPR at parity one, which has risen substantially since 1994 [20], unlike a2. Last, comparing the evolution of cohort PPRs, we observe that the variations in A1 are slow and of limited amplitude, fluctuating between 78 and 80%, while those of A2 follow a continuous downward path, broken only by the small number of cohorts which appear to have responded positively to the third-child policy.
18Assuming that without such measures A2 would have declined regularly between the 1973 cohort (46%) and the 1990 cohort (44%), this effect represents an average of 1.5 percentage points for all intermediate cohorts. This is a minimum however, since we cannot rule out the possibility that A2 would have dropped even further if all support for families of three children had been withdrawn.
19In short, though it is very likely that these two periods of targeted family policy did indeed influence the frequency of progression from two to three children, the effect is limited and difficult to quantify, since we cannot observe what would have happened in the absence of such measures. And the effect of maintaining a certain number of specific provisions for families of three children or more is even more difficult to calculate. Tempo effects are easier to identify however.
20Moreover, a possible “knock-on” effect of third-child measures and associated information campaigns on births of other orders, as suggested by the comparative changes in the period PPRs a1 and a2, cannot be ruled out.
II – Who are the men and women with a third child?
21To observe the most recent possible cohorts at the time of the 1999 survey, we will limit our analysis to births occurring in the 11 years following the second birth. Though 12 to 15 years may elapse between the birth of the second and third child, such cases are rare and their inclusion would not affect the characteristics observed [21]. So the proportion of men and women who have had a third child over the 11 years following the birth of their second child will be our measure of intensity of A2.
22We will begin by observing changes in the intensity and tempo of the PPR A2 for men and for women. We will show that the age at second birth and the length of the interval between the first and second birth have a determining impact on this ratio. Lastly, we will assess the role of other family, economic or social variables: the sex of the first two children, nationality, parents’ sibship size, educational attainment, social group, individual employment trajectories and union histories. Whenever the behaviour of men and women differs, separate analyses will be performed for each sex. In the last part, we will use logistic regression models to verify whether all these identified factors stand up to analysis with “all other things being equal”.
1 – Little difference between men and women in progression from two to three children
23The proportion of mothers and fathers of two children who have a third child in the 11 years following the second birth follows a very similar pattern [22] to the cohort PPR A2 analysed above (Figure 1). The figure in the Appendix shows the values of this PPR and of the mean interval between second and third birth in the cohorts of men and women at parity two. Table 2 summarizes this information by grouping these parity cohorts by five-year groups centred on “round” years (1970, 1975, 1980 and 1985). The group sizes thus obtained are large enough (Table A in the Appendix) for analysis, though they obscure the intensity and timing variations of the annual cohorts (Table 2), and notably the small effects on specific cohorts of measures to encourage third births. Hence after a drop between the cohorts labelled 1970 and 1975 (the 1967-72 and 1973-77 cohorts respectively), the PPR tends to be relatively stable in the three following groups of cohorts, at around 44.5% for women and slightly higher (45%) for men. The mean interval between second and third birth also stabilizes, at around four years, after rising sharply between the first two groups of cohorts.
PPR to third child in the 11 years following a second order birth and mean interval between second and third birth
PPR to third child in the 11 years following a second order birth and mean interval between second and third birth
24Figure 2 shows the distribution of third order births by length of interval since the birth of the second child. It is always during the second year following this second birth that third births are most common, though their frequency after short intervals (one and two years after the second child) decreased considerably after the 1970 cohort (Figure 2a). The increase in births after longer intervals (four years and more) did not fully offset this decline, resulting in a first decrease in the PPR (Figure 2b). Intensity then stabilized in the 1975 to 1985 cohorts, with a relatively stable distribution of births by interval since previous birth (Table 2).
Third-birth fertility in the different cohorts by interval since the second order birth
Third-birth fertility in the different cohorts by interval since the second order birth
Population: men and women25The 1990 cohort can be observed for seven years. We see that the low fertility after short interval lengths has not been offset by an especially high frequency of births after four to six years (Figure 2a). Six years after their second child, only 34.4% of these fathers and mothers had a third child, compared with around 37.5% in the previous cohorts (Figure 2b). So it is practically certain that the frequency of progression from two to three children will decrease in this cohort, given that fertility at long interval lengths is unlikely to increase by the required amount (Figure 2a).
2 – The importance of age at first two births and of the intervals between these two births
26Among all the variables linked to the probability of having a third child, the age at first two births and the spacing of these births are of primary importance. The youngest mothers and fathers, or those with a short interval between the first and second births are much more likely to have a third child than older parents or those whose first two children were spaced far apart. A short interval between the first and second birth and a young age at second birth often have a common cause, i.e. a desire for a large family.
27Despite the increase in age at the first two births (Table 3), we observe that the PPR remained stable in the parity two cohorts from 1975 to 1985 (Table 2) [23]. This can be explained by looking at changes in the behaviour of men and women by age group.
The consequences of a later age at second birth on the PPR to the third child
28Given the age difference between partners, the mother’s age at the second birth is two to three years lower on average than that of the father, and the fathers’ ages are more widely dispersed than those of the mothers [24]. The increase in maternal and paternal age at childbearing is producing a rapid increase in the proportion of fathers aged at least 40 at the second birth, and a reduction in the number of fathers aged under 25 (Table B in Appendix). For mothers, the number of second births after age 40 is increasing, though the proportion remains very small. For both sexes, the mean age at the birth of the second child rises rapidly between 1975 and 1995 (+3.2 years for men and +3.1 years for women). This is a consequence of the increase in both age at first birth and the interval between first and second birth. The first component explains almost 80% of the increase in mean age at second birth between 1975 and 1995, and more than 90% of the increase between 1985-95 (Table 3).
Change in mean age at second birth and its components in cohorts of men and women who had their second child between 1975 and 1995 (in years)
Change in mean age at second birth and its components in cohorts of men and women who had their second child between 1975 and 1995 (in years)
29Whatever the cohort, the later the second birth, the lower the probability of having a third child (Figure 3). The age effect is especially strong for women: their PPR at parity two is between 60% and 70% when they have their second child before age 25, compared with only 10% to 20% when it is born between ages 35 and 45. For men, the age effect is also substantial up to age 35, in the 1985 cohort especially. Beyond this age, increased age does not greatly decrease the PPR at parity two.
Parity progression ratio A2 of parity two cohorts by age at second birth, 1975 and 1985 cohorts
Parity progression ratio A2 of parity two cohorts by age at second birth, 1975 and 1985 cohorts
30Between the 1975 and 1985 male cohorts, the PPR at parity two increased for all ages at second birth, except in the 30-34 age group, where it remained stable (Figure 3). For women, an increase is observed under age 35 only, while PPRs for mothers aged over 35 fell very slightly. In the younger age groups, this trend is explained by a selection effect: men and women who have children at a very young age have a particular profile (social group, educational attainment, nationality, etc.) and are increasingly selected among those who will go on to have a large family. In the intermediate age groups, there has been a “sideways shift” in behaviour, which corresponds to the postponement of fertility: for example, though the age at second birth has increased by two years, the women who had a second child at age 28 in 1985 behave in the same way as those who were 26 in 1975. For the oldest age groups, this shift appears to affect men but not women: the probability of having a third child remains low for women when they have a second child after age 35 [25], probably due to an increase in hypofecundity and sterility with age, but perhaps also because the desire for more children after a second birth is closely linked to age [26].
31Hence, it is because parity progression ratios have increased for practically all ages at second birth that the increase in age at second birth has not, until now, lowered the PPR in the 1975 to 1985 cohorts: if age-specific behaviour had not changed, the PPR A2 would have fallen to 40.2% in the 1985 cohort (Table 4, first line), instead of its actual level of 44.5%.
Estimate of PPR A2 in cohorts of mothers who had a second child in 1975, 1985 and 1995 on the basis of different behaviour hypotheses (%)
Estimate of PPR A2 in cohorts of mothers who had a second child in 1975, 1985 and 1995 on the basis of different behaviour hypotheses (%)
32What might the PPR of the 1995 cohort be, given the age structure—already known—of mothers at second birth in this cohort? If the behaviours observed in the 1985 cohort do not change, i.e. if within each group of ages at the birth of the second child the proportion of women having a third child is the same as that observed in the 1985 cohort (Figure 3a), the PPR will fall to 38.5%, due to the higher age of mothers at second birth (Table 4, second line). It is more realistic to assume that PPRs by age at second birth will continue to change. The scenario chosen here involves prolonging the increase in PPRs observed between 1975 and 1985 in a linear manner at young ages (under 35), and, for the ages at which the PPR fell slightly in 1985, retaining the PPR observed in 1975 (high hypothesis on the last line of Table 4). The prospective results thus obtained show that cohort ageing should inevitably lower the proportion of women having at least three children (from 44.5% in the 1985 cohort to a maximum of 43% in the 1995 cohort, under the high hypothesis).
Interval between first and second birth and parity progression ratio to the third child
33The shorter the interval between the first two births, the higher the parity progression ratio to the third child (Figure 4). Interval length zero is an exception. This very specific category, representing around 2% of second order births, mainly comprises twins. After a twin birth, the risk of having a third child is slightly lower (Pison, Couvert, 2004). So the PPR is highest (between 60% and 70%) when the second child is born during the year following the birth of the first. From then on, the longer the interval between the first two children, the lower the probability of having a third child. We know that in large families the intervals between births are shorter than in small families (Desplanques 1985, Lavertu, 1997).
Parity progression ratio A2 in cohorts of parity two women by interval between first and second birth
Parity progression ratio A2 in cohorts of parity two women by interval between first and second birth
34From one cohort to the next, for all intervals between first and second birth, the PPR at parity two increased (Figure 4). But the overall PPR remained stable since, over the same period, the interval between the first two births grew longer (Table 3 and Table C in the Appendix). Shorter intervals have become less frequent and increasingly select couples who will end up having a large family.
3 – Other characteristics linked to the frequency of third births
Importance of family characteristics
35The parents’ sibship size is closely correlated with the level of fertility (Deville, 1979; Desplanques, 1985), so it is hardly surprising that progression to a third child is more frequent when the father or mother come from a large family (Table 5). But there is no clear difference between the behaviour of men and women who are only children and those with one brother or sister, the former even having a higher propensity to enlarge their family. It is only for those who belong to sibships of three or four children that the probability increases significantly, and only for very large sibships (at least five children) that the frequency of progression to a third child exceeds the average.
Parity progression ratio A2 by certain characteristics of the family of origin (% of women and men with a third child in the 1975 and 1985 parity two cohorts)
Parity progression ratio A2 by certain characteristics of the family of origin (% of women and men with a third child in the 1975 and 1985 parity two cohorts)
36Nationality seems to play a much more definite role: it is much more common for African men and women (from North or sub-Saharan Africa) to have a third child than for the others. This is logical, since these parents of foreign nationality, mostly recent immigrants, are unlikely to have switched to the European two-child family model, which is relatively rare in their home country [27].
37Since people of African nationality generally have large sibships, cross-tabulation between nationalities and sibship sizes highlights the link between parents’ sibship size and PPR. In each group of nationalities (European/non-European) the fathers and mothers were grouped according to their sibship size: small (one or two children for persons of European nationality, one to four for the others), medium (respectively three or four children and five to seven children), or large (five or more/eight or more). The link between these two variables shows up clearly for men and women of European nationality and, once again, only those who belong to the largest families (five children or more) have a third child more often than the average. Among non-European foreigners, the probability is always much higher and appears to be less directly linked to sibship size.
Having at least one boy and one girl
38When the two first-born are of the same sex, the probability of having an additional child is slightly higher (between 46% and 50% depending on the cohort) than when they are of a different sex (41% to 42%) (Table 6). There is no sex preference however: no difference in behaviour is observed between parents of two boys and parents of two girls. These observations tally with findings based on the previous INSEE Family surveys (Calot, Deville, 1971), though it would appear that the large drop in the PPR A2 has widened the difference in behaviour between parents who already have a boy and a girl and those with two children of the same sex [28].
Repartnering favours the birth of a third child
39When the first and second child are born of different unions, the PPR at parity two is slightly higher than when they are born in the same union (Table 6). In the most recent cohort however, this relationship is less pronounced for women than for men. When the first two children are both born in a stable union, this PPR is even slightly lower if the parents are married at the time of the second birth. So marriage does not appear to have any particularly positive effect on the third birth. However, in the 1975 and 1985 cohorts, married couples still represent a large majority and their PPR strongly influences that of the whole cohort.
Parity progression ratio A2 by certain characteristics of the couple or of the family being formed (% of women and men with a third child in the 1975 and 1985 parity two cohorts)
Parity progression ratio A2 by certain characteristics of the couple or of the family being formed (% of women and men with a third child in the 1975 and 1985 parity two cohorts)
40Compared with a union that is undissolved after the birth of the second child, union dissolution followed by repartnering favours the birth of a third child when these two events occur within eleven years of the second birth. Conversely, and in all logic, it is the parents whose union dissolution is not followed by repartnering who least often have a third child [29]. Hence union dissolution both discourages and encourages third births. And although the number of union dissolutions rose by one-third between the 1975 and 1985 cohorts, the negative effects on third births were offset by repartnering.
Relationship between educational attainment and fertility: a backward J-curve
41In France, the relation between women’s educational attainment and their completed fertility was first described as a U-curve (Calot, Deville, 1971), then as a backward J-curve (Desplanques, 1985). Since the 1982 Family survey, following the inclusion of all women (and not only ever-married women, as in previous surveys) this pattern has disappeared, since female celibacy and childlessness increase with the level of qualification (Desplanques 1985, 1994; Daguet, 2000), so the completed fertility of the most highy educated women decreases accordingly.
42When we observe the PPR at parity two, a similar selection process is under way, since only men and women with at least two children are concerned. This explains why the configuration obtained is very similar to this famous backward J-curve: in the 1975 cohort, the PPR A2 is highest among unqualified women (this category includes many foreign women, though they are not the majority), much lower among women with intermediate qualifications and rises substantially again among women with a higher education degree. Among the most qualified women, A2 is similar to that of women with only a primary school certificate, while women with intermediate qualifications (BEPC, BEP, general, technical and vocational baccalauréat) have the lowest PPRs. In the 1985 cohort the order is slightly different: holders of a technical or vocational baccalauréat have the lowest PPR, while parity progression at parity two is no longer especially high among women with only a primary school certificate. A similar configuration is observed for men from the 1975 cohort [30].
Female labour force participation is still a decisive factor
43Women who have never worked before the birth of their second child much more frequently have a third child than those who have worked. This category of women who have never worked, which decreases from one cohort to the next, still represents 10% of the 1985 cohort. It includes far more foreign women of non-European nationality and from a working class environment, who are most likely to have a third child [31]. For the other women, the frequency of third births appears to be quite closely linked to the continuity of their labour force participation: those who have always worked and did not stop for the birth of their first two children, and who represent the majority (61% of the 1975 cohort and 66% of the 1985 cohort [32]), less frequently have a third child than those who have already stopped working in the past. Among the latter, there is no real difference between those who stop at the time of the second birth and those who have already stopped working by that time. This correlation between labour force participation and fertility, already demonstrated on many occasions, is generally interpreted as a two-way link: though some women are reluctant to have another child because they do not wish or cannot afford to stop working, others definitely do not want more children, so it is easier for them to remain economically active. But a third factor influencing the decision to not have a further child and to remain in the labour force cannot be ruled out.
Behaviour that varies from one socio-occupational category to another
44Only women who have never worked (either before or after the birth of their second child) could not be included in a socio-occupational category. Logically, it is these women who have by far the highest PPRs (Table 7).
Parity progression ratio A2 by educational attainment and socioeconomic characteristics of the parents (% of women and men with a third child in the 1975 and 1985 parity two cohorts)
Parity progression ratio A2 by educational attainment and socioeconomic characteristics of the parents (% of women and men with a third child in the 1975 and 1985 parity two cohorts)
45All other respondents were classified by their occupation reported at the time of survey, or by their most recent occupation. Self-employed mothers followed by those with higher- and middle-level occupations had the lowest PPRs at parity two, while those in the blue-collar worker and farmer categories most often had a third child. These differences grew much smaller between the 1975 and 1985 cohorts however. From the fathers’ side, things are not very different, though we note that farmers are no characterized by high fertility. Only blue-collar workers have a PPR at parity two significantly above the average.
46For the frequency of progression to a third child by partner’s [33] occupation, the results are similar: on the female side, it is the partners of blue-collar workers who have the highest fertility, and increasingly rarely those of farmers, while on the male side, it is the partners of economically inactive women and, far behind, those of female blue-collar workers who most often have a third child. The partners of self-employed women have a low PPR to the third child.
47But it is above all when the socio-occupational category (SOC) of both partners is taken into account that the differences become pronounced [34]. For example, for men in the 1985 cohort, whatever their occupation, it is not surprisingly those with economically inactive partners who have the highest PPR at parity two (up to 80% among blue-collar workers whose wife is economically inactive). The only exception is farmers (42% when their wife is economically inactive, the same as for farmers as a whole, see Table 7). For a given SOC, on the contrary, it is those whose partner is self-employed or has a higher-level occupation who are least likely to have a third child, with the exception of couples who both have higher-level occupations. Men with a blue-collar partner tend to have a high PPR at parity two (except in the case of the non-farming self-employed) while the opposite is true when the partner is a white-collar worker. For male blue- and white-collar workers with partners in a middle-level occupation, the PPRs at parity two are very low, while for non-farming self-employed men they are relatively high. Generally speaking, in each socio-occupational category, homogamy leads to behaviours close to the average for the social group, while exogamy may lead to PPRs that are often much lower, and sometimes higher than those of the SOC as a whole.
48Seen from the woman’s side, for a given socio-occupational category, in the 1985 cohort, the influence of the partner’s occupation is generally less determining, except among economically inactive and non-farming self-employed women. Practically always, except for women with higher-level occupations, having a blue-collar partner increases the PPR at parity two, while having a partner with a higher-level occupation sometimes increases it (PPR of 51% for female blue-collar workers, 46% for women in middle-level occupations and 43% for those in higher-level occupations) and sometimes lowers it (38% for female white-collar workers and 26% for non-farming self-employed women). Women whose partner is a white-collar worker or has a middle-level occupation always have low PPRs at parity two, with the exception of women in higher-level occupations. It is only when women are economically inactive or self-employed that the partner’s socio-occupational category has a determining impact on the level of A2, exactly following the social hierarchy: blue-collar worker, white-collar worker, middle-level occupation and higher-level occupation (from highest to lowest).
4 – Multivariate analysis confirms the importance of each factor
49Multivariate analysis shows that each variable taken separately has an effect on the probability of having a third child. But these variables are often correlated. This is the case, for example, for sibship size and nationality, but also for nationality and SOC. Do these effects stand up to an analysis with “all other things being equal”? To verify this, we used several logistic regression models [35] whose dependent variable is the fact of having had or not had a third child in the 11 years following the year of the second birth. The independent variables are all the preceding descriptive variables.
50We chose to synthesize the “age at second birth” and “educational attainment” variables in order to control for the correlation between them. Age at second birth is no longer taken as an absolute value, but relative to all persons with the same level of education. For example, a woman who had a second child at age 24 will be included in the “intermediate age” categories if she stopped school after the baccalauréat, and in the “young age” category if she has a higher education degree. This methodological choice is consistent with the methods used in other studies of the impact of educational attainment on progression to a third child conducted in France and in other European countries (Toulemon 1995; Hoem, 1996; Hoem et al., 2001).
51The results of the multivariate analysis shown in Table 8 correspond exactly to those obtained in univariate analysis, indicating that the differences do not simply reflect structural effects.
Factors influencing the probability of having a third child in the 11 years following the second birth (results of logistic regression models)
Factors influencing the probability of having a third child in the 11 years following the second birth (results of logistic regression models)
52Whatever the cohort and whatever the sex, nationality is always the factor which most strongly influences the PPR at parity two. The odds ratios [36] are always higher among women and men of non-European nationality. The size of family of origin also has a very significant effect, mainly among non-Europeans. The odds ratio (OR) associated with the “non-European nationality and from a large family” modality is the highest in the model, though it decreases between the 1975 and 1985 cohorts, falling from 16.6 to 4.1 in the female cohorts, and from 8.2 to 5.6 in the male cohorts.
53The demographic behaviour variables are also highly discriminative: a very short interval between the first two children and a relatively young age at the birth of the second child, especially among men with a higher education degree, are associated with a high PPR to the third child (in the 1985 cohort, OR = 2.4 for men with a higher education degree who had a second child at a relatively young age). Conversely, a long interval between the first two births and a late age at the birth of the second child reduce the relative probability of having a third child, especially among the least educated women (OR = 0.5). Between the 1975 and 1985 cohorts, the effect of short birth intervals is heightened by a selection effect, while the effect of longer birth intervals (4 years or more) is slightly reduced. It is probably a selection effect which also explains the increase in odds ratios among persons with low levels of education who had a second child at a young age, women especially.
54The variables which combine births of children and union histories of individuals always play a highly significant role: when the first two children are born of different unions, the risk of having a third child is increased (OR between 2.4 and 3). On the other hand, the fact that both children are or are not born of a married couple does not significantly modify the probability of having a third child.
55The effect of occupational history up to the second birth is also significant, though appears to be weak, especially among women who never worked before the second birth. However, the effect of economic inactivity is largely absorbed by the “socio-occupational category” variable, one of whose modalities is “no economic activity”. In the absence of this variable, the fact of never having worked before the second birth has a much larger positive effet (model not presented here).
56Lastly, the models show that the specific effect of the socio-occupational category weakens between 1975 and 1985, among women especially, with the exception of those who are economically inactive. Likewise, the influence of the partner’s socio-occupational category also weakens, mainly among women. Conversely, the effect of the female partner’s economic inactivity increases for men (OR = 2.0 in 1985). The different social groups are associated with specific demographic and social behaviours and this is taken into account in the model by means of control variables (nationality, age at second birth, interval between births, etc.).
57Overall, the direction (positive or negative) of the effects of the variables is the same for men and for women. Only the values of the parameters vary slightly over the ten-year period separating the two cohorts (decrease in the effect of nationality and sibship size variables, increase in the effect of a short interval between first and second birth). The demographic variables (age at previous births and spacing of first two births) and nationality are the variables with the most determining influence on progression to a third child. The effects of employment and union histories are confirmed. More detailed analyses would be necessary however to observe the specific effects of these variables (analysis of sub-cohorts, life history models). Such analyses are beyond the scope of this article, whose aim was above all to identify the main factors associated with a third birth.
Overview and outlook
58In France, family policy focuses more specifically on third births than in other European countries. We have shown that if one of the objectives of family policy is to sustain fertility, then this policy is well-targeted, given the close correlation between the number of births of order three and above and the fertility level in Western European countries. Moreover, the proportion of men and women having more than two children fallen below half since the 1970s, so there is substantial scope for increasing the parity progression ratio to the third child in France. The measures specifically targeting third births introduced between 1978 and 1980 primarily influenced the timing of births. We have nevertheless shown that these measures, along with the creation of the parental leave allowance (APE) in 1985, may have helped to maintain and even increase the probability of having a third child. The recent trend has been towards a decrease in the PPR A2, which has remained below 45% since the 1987 cohort, while targeting of family policy on the third child has been substantially reduced and the extension of the APE to the second child in 1994 now serves as an incentive for second births.
59The key variables influencing the PPR to the third child are behaviour (age at second birth and interval between first and second birth) and nationality, even after controlling for structural effects. The steady increase in age at first two births, without a significant shortening of birth intervals, should produce an ongoing decrease in A2 in the cohorts of men and women who had a second child in the 1990s.
60All other variables—sex of first two children, parents’ sibship size, union and occupational histories, educational attainment and socio-occupational category of the person and of their partner—remain significant. The effect of the socio-occupational category is tending to weaken however.
61Our analysis focused on cohorts of women and men who had a second child between 1973 and 1977 (grouped under the label “1975 cohort”), and between 1983 and 1987 (“1985 cohort”). The 1999 EHF survey is too recent to see if each variable still has the same effect on the PPR at parity two of the 1995 cohort (1993-97), though the characteristics of the men and women of this cohort at the birth of their second child are already known. Union histories have become more complex (almost 10% of second children are born in a different union from the first, compared with 5% in 1975), women are more economically active (the proportion of women who had never worked before the second birth fell to 7% in 1995 compared with 12% in 1975) and age at second birth has risen sharply. Though repartnering tends to favour parity progression, union dissolution without repartnering, increased female employment and a higher age at childbearing have the opposite effect. Nevertheless, as is the case for female employment, delayed first and second births and the lower PPR to the third child may have a common cause, i.e. the desire of both men and women to have smaller families. In this case, the increasing age at childbearing is not “responsible” for the declining frequency of third births.
62A family policy that addresses the needs and desires of women and men wishing to have a third child must take account of changes in society, and in particular the growth in female labour force participation. Maintaining a PPR A2 close to 45% will depend above all on the desire of men and women to have a third child. A policy that supports families with three or more children certainly helps to foster this desire.
Change in intensity and timing of third order fertility in parity two cohorts (1968-87 cohorts)
Change in intensity and timing of third order fertility in parity two cohorts (1968-87 cohorts)
Size of cohorts of fathers and mothers of two children observed in the 1999 EHF survey by year of birth of second child
Size of cohorts of fathers and mothers of two children observed in the 1999 EHF survey by year of birth of second child
Distribution of men and women in the cohorts by age at second birth (%)
Distribution of men and women in the cohorts by age at second birth (%)
Distribution of second births in the cohorts by interval since first birth (%)
Distribution of second births in the cohorts by interval since first birth (%)
Notes
-
[*]
Université Marc Bloch, Strasbourg 2 and Institut national d’études démographiques.
-
[**]
Institut national d’études démographiques.
Translated by Catriona Dutreuilh -
[1]
In 1946 the rate was 20% of the BMAF (fixed monthly base used to calculate the various benefit entitlements) for the second child, and 30% of the BMAF for the third, i.e. 50% more. See Laroque (1985) and CNAF (2005) for a presentation of French family policy from 1945 to the present day.
-
[2]
Law of 12 July 1977
-
[3]
The family income supplement replaced the single wage allowance (or non-working mother’s allowance) and the childcare allowance introduced in 1972.
-
[4]
This was known as the “Giscard million” since 10,000 francs = 1 million old francs. The purchasing power of 10,000 francs in 1980 was the equivalent of 3,590 euros in 2004.
-
[5]
Under the French system known as “Le quotient familial” (family quota) a taxable income is divided into “units” reflecting the family status of the tax-payer. People living alone (single, widowed etc.) represent one unit and are taxed on the whole of their income. A married couple are considered as two units and their joint income will be divided by two. The first two children count as one-half a unit each and, since 1986, each additional child has been counted as a full unit.
-
[6]
The allowance was set at 23% of the monthly base for the second child, at 42% for the third, and at 39.5% for each of the following children.
-
[7]
The allowance was raised to 32% of the monthly base for the second child and set at 40% for the third and following children. These rates have been adjusted just once since then: in 1986 the rate was raised to 41% of the BMAF for the third and following children, taking the difference to 28%.
-
[8]
This abolition was announced in 1982 in the familly allowance reform bill adopted by the Council of Ministers on 21 April 1982 (Laroque, 1985, p. 284).
-
[9]
Initially granted for a maximum period of 24 months, the APE was small and the conditions of entitlement were strict (proof of two years of working activity in the last 30 months); Its amount was gradually increased, the conditions of previous working activity were relaxed and the duration of payment extended to the child’s third birthday. In fact, only 27,000 families received the APE in 1986, through their number increased rapidly, reaching 162,000 in 1987 and 186,000 in 1988 (CNAF, 2005).
-
[10]
Since 1990, there have been around 900,000 beneficiaries of family income supplement (granted exclusively to families with three or more children since 1986).
-
[11]
A married couple with three dependent children and a monthly income of 2,500 euros is not liable to income tax, thanks to the third child. The same couple with two children would pay 480 euros. Moreover, families which do not pay income tax are entitled to local authority benefits and to rebates on certain services.
-
[12]
The rate of discount depends on the number of children: 30% for 3 children, 40% for 4 children, 50% for 5 children, and 75% for 6 children or more.
-
[13]
We do not have data for Ireland, where completed fertility is higher than in France (2.41 children per woman), though it is certain that fertility at birth orders three and above is higher than that of France.
-
[14]
Greece comes out as the country where two-child families are most common: more than half (52%) of the women born in 1960 produced precisely two children. The number of two-child families is also very high in Eastern Europe, though this model is in decline due to rapidly decreasing fertility.
-
[15]
For a comparison of family policies in Europe, see Gauthier (2002) and Math (2005).
-
[16]
For the sake of clarity, we did not represent here the annual indicators resulting from the combination of duration-specific probabilities; these annual indicators (which have already been published by Toulemon and Mazuy, 2001, Figure 10) vary in exactly the same way as the sum of rates, though the amplitude of variation is slightly smaller, especially for a1.
-
[17]
The interval between the second and third birth was 3.3 years among cohorts of women who had a second child between 1965 and 1968. It then rose very quickly, varying between 4 and 4.3 years from the 1973 cohort. For the interval between the first and second birth, the pattern is very similar, though the mean interval has risen from 3.2 years to 3.7-3.9 years.
-
[18]
In theory, this relation is only fully verified if timing evolves in a linear manner and if intensity does not change or also evolves in a linear manner.
-
[19]
The estimate of non-observed births is based on stable third-birth probabilities for long birth intervals.
-
[20]
The introduction of the APE for children of birth order two has clearly shortened the interval between the first and second births (a1 rose well above A1 from 1996), and also appears to favour parity-two fertility intensity since, according to our projections, A1 tends to rise slightly from the 1992 cohort onwards. The impact of the APE on fertility has been analysed by Piketty (2005).
-
[21]
In the cohorts of women who had their second child between 1970 and 1983 (last cohort observed during the 15 years prior to the survey), only around 3% of the third children were born 12 to 15 years after the second, and the contribution of these births to the PPR is below 1.5 for 100 women. This explains why the values of A2 in Table 2 are slightly below those of Figure 1.
-
[22]
Due to family recomposition, children do not always have the same birth order in the paternal and maternal lineages. However, the differences between men and women may also be due to variations in the reliability of responses given by men and women in the survey, or to a difference in representativeness of male and female samples.
-
[23]
Toulemon and Mazuy (2001) also made the same observation (see Figure in Appendix II of their article).
-
[24]
The standard deviations are respectively 5.22 and 5.62 years in the 1975 and 1995 cohorts for men versus 4.05 and 4.2 years for women.
-
[25]
For women aged 35-39 at the second birth, this slight decrease in the PPR is accompanied by an increase in the mean interval between second and third birth. However, for all cohorts, beyond age 25-29 this interval decreases in length as the age of the mother (or father) at second birth increases. Hence fewer women and men who have a second child after age 30 go on to have more children, but for those who do, the interval between births is shorter.
-
[26]
The link between increased foetal malformation risk and mother’s age is now widely recognized.
-
[27]
Some of their children may have been born in the home country: persons present in France in 1999 are asked in the EHF survey about all the children they have had, whatever their place of birth.
-
[28]
In the 1962 Family survey, the PPR in “marriages of completed fertility” (women whose first marriage was not dissolved before her 45th birthday) was 57.6% in the first case and 60% in the second (Calot, Deville, 1971, p. 37).
-
[29]
Most third births reported by persons whose union had been dissolved without repartnering occurred prior to dissolution. Likewise, in the case of union dissolution followed by repartnering, some of the children are born before separation (one-third of the 1985 female cohort). For this descriptive analysis, it is not possible to calculate more precise indices.
-
[30]
For men however, the selection effect does not concern the same categories of qualification as for women, since it is the most highly qualified men who are most likely to form a family, while the less qualified have more difficulty finding a partner (de Singly, 1982).
-
[31]
For example, for the entire 1975 cohort, 11.7% of women never worked before the second birth. Among non-European women, the proportion is 65.7%. Among women whose partner is a blue-collar worker it is 44.2% for Europeans and 76.5% for non-Europeans.
-
[32]
Note that with the extension of the APE to the second child from 1994, this category of mothers who have not stopped working drops back to 67% in the 1990 cohort and to 58% in the 1995 cohort. The date of the 1999 EHF survey is too close to 1994 to measure the effect on third births (Piketty, 2005).
-
[33]
Current partner or most recent partner if the person is no longer in union.
-
[34]
The number of possible couple types is too large to be presented here.
-
[35]
We built four models: one for each sex and parity cohort. We began by building a global model incorporating the cohort and sex as independent variables, though these two variables were the least significant of the model. In addition, the female and male models have their own constraints, such as the “occupational history” variable which must be excluded from the male model, given the infrequency of economic inactivity and interruption of activity among men.
-
[36]
The odds ratio (OR) is the ratio between two quantities: first, the probability of an event occurring, in a given group, divided by the probability of this event not occurring (p1/(1 – p1); second, the same fraction in the reference group (p0/(1 – p0). The result is interpreted “all other things being equal” when the OR is calculated for a set of factors. In practice, an odds ratio significantly different from 1 indicates that the probability of the event occurring in the group being studied is signficantly higher (or lower if the OR is less than 1) than in the reference group. The more the OR is significantly different from 1, the greater the difference between the two groups.
