1Amongst the different methods that could be used to make midrange forecasts of fertility, surveys on fertility intentions seem at first sight to offer the advantage of being simple: to estimate the number of children likely to be born, why not ask the people principally interested – women likely to have children – what their intentions are?
2A longitudinal study, undertaken by INED, makes it possible to judge the validity of these methods and, more specifically, to assess whether it is possible to forecast fertility behaviour from statements about intentions.
3During the first phase of this survey, in 1974, a representative sample of women with a young child were asked questions about their fertility intentions [1]. By a "follow-up" of this sample, it was possible to re-interview these women, mainly by postal questionnaire, in 1976 and 1979. 1,749 women, or 82% of the original sample of 2,135, were traced in 1976 and replied, and 1,906 or 89% in 1979. 1,439 women replied to all the stages of this longitudinal survey [2].
4After a period of roughly five years, these data made it possible to assess the consistency between intentions and behaviour more accurately than was possible in a previous article [3], in which information about the first two stages (1974 and 1976), extending over less than three years, was used. Furthermore, these data have suggested some reflections on the nature of the decisions which motivate women to reproduce or not.
From intentions to behaviour patterns: different possibilities
5To compare intended and actual behaviour accurately, it is necessary to look at the replies of individual respondents at different stages of the survey. Such a procedure would result in the sample being broken down, even if intentions were to be grouped systematically and behaviour defined by a limited number of criteria. Women interviewed about their fertility intentions in 1974 could reply:
- I intend to have a(nother) child;
- I do not intend having any more children;
- I have no specific intentions.
6Women in each of these groups could have produced 0, 1 or 2 children between 1974 and 1976, and could have been pregnant or not in 1976. Thus, information about the period 1974-1976 could result in 18 (3 × 3 × 2) different outcomes. The same is true for the period between 1976 and 1979, so that each individual life history can be placed into one of 18 × 18 = 324 different cells.
7Obviously, most of these outcomes are purely theoretical and many of the cells will be empty. To avoid fragmentation of the sample, particularly because agreement between actual and intended behaviour needs to be judged for groups with different sizes of family in 1974, which could make the results impossible to assess, two sub-divisions have been made, which do not result in any significant loss of information:
- when the woman was pregnant, her pregnancy was considered to have already resulted in a live birth, when intention and behaviour were compared: in this case, conception is more relevant than the birth itself [4];
- it seemed unnecessary to distinguish between cases where two children had been born between two successive stages of the survey or between the first and the last survey.
8Applying these rules results in a dichotomy of behaviour patterns: at least one child was born (or expected) / no child was born (or expected). In Diagram 1, we present the range of possibilities thus defined.
General presentation of the results
9In 1974, 70% of mothers with one child desired to have at least one more, whereas between 70 and 90% of women with two or more children did not desire an additional birth (Table 1) [5].
Fertility intentions in 1974, by number of children already born (sample observed in 1974, 1976 and 1979)
Fertility intentions in 1974, by number of children already born (sample observed in 1974, 1976 and 1979)
10These results, calculated for the sub-sample of women interviewed during the three stages of the survey, are very similar to those observed in 1974 amongst respondents to the original sample. This leads us to believe that the attrition in numbers between the various stages of the survey has not resulted in any significant bias in the distribution by intention in 1974.
11Between 1974 and 1979, 42% of the women interviewed produced at least one child (or were expecting one). This proportion obviously varied considerably for women who had expressed different intentions and, to a lesser degree, for those with different numbers of children already born for women with the same intentions (Table 2). Amongst the women who hoped to have a child, 72.8% produced at least one, and the proportion was slightly higher (76.1%) when the desired child was a second child. At the other end of the scale, 80.3% of the women who did not want another child did not produce any, but this proportion dropped to 65.3% in the case of women who had had only one child by 1974. Finally, the women who had no specific intentions could be divided into two separate groups of roughly the same size: 46.1% produced a child, and 53.9% did not.
Children born or conceived between 1976 and 1979(1)
Children born or conceived between 1976 and 1979(1)
(1) Results not calculated because of the small numbers.12Diagrams 1.1, 1.2 and 1.3 represent the various paths which resulted in a woman producing or not producing a child, taking into account their intentions in 1974 and 1976, irrespective of the family size already achieved in 1974. 72.7% of the women who did not wish to have more children (Diagram 1.2) are in only one of the twelve possible available cells: by 1976, their behaviour coincided with their intentions in 1974, and they produced no further children. The dispersion is greater for women who wished to have another child (Diagram 1.1), in part because they could have the additional birth they desired during either the first (1974-1976) or the second period (1976-1979). However, there was less consistency between their original intention and their actual behaviour. Finally, Diagram 1.3 confirms the heterogeneous nature of the group of women who had no well-defined intentions in 1974.
Intended and actual fertility (1974-1979). All birth orders
Intended and actual fertility (1974-1979). All birth orders
Intended and actual fertility (1974-1979). All birth orders
Intended and actual fertility (1974-1979). All birth orders
13A more detailed analysis of the survey results, which takes into consideration the number of children already born in 1974, will make it possible to verify these preliminary observations.
Intended and actual fertility (1974-1979). All birth orders
Intended and actual fertility (1974-1979). All birth orders
Consistency between intentions and behaviour
14Consistency between intention and behaviour may be measured (as shown on Diagrams 1.1 and 1.2) as follows:
15— If the woman intended to have another child, we could use the ratio of the number of those who had at least one child born between 1974 and 1979 or who were pregnant (A1 + A’1 + A2 + A’2 + A3 + A’3 + A4 + A5+ A6) to the original number of women in 1974 (T1).
16— If the woman did not intend to have any more children, the ratio of those who had remained childless and were not pregnant (B’4 + B’5 + B’6) to the original number of women (T2).
17The results of these calculations are shown in Table 3 and illustrated in Figure 1.
Consistency between intended and actual behaviour (%) (1974-1979)(1)
Consistency between intended and actual behaviour (%) (1974-1979)(1)
(1) Results not calculated because of the small numbers.18These show:
19— Consistency between intended and actual behaviour was around 80% for women who were already mothers of one child and intended to have a second (76.2%) and for the women who already had at least two children and did not desire any more (83.3%, 80.9% and 82.3% for women with two, three and four or more children respectively).
20— Consistency was significantly lower for women who intended to have a third child (63.7%) and for mothers of one child who did not intend to have a second (62.7%). In 1974, only a very small proportion of the population fell into these categories (17.8% and 14.3% respectively) (cf. Table 1). With the passing of time, such women have, therefore, tended to adopt a behavioural pattern which was more in keeping with the norms relating to family size, centered around two children.
21Up to now, we have compared actual fertility between 1974 and 1979 with intentions in 1974. Some of the differences observed between intended and actual behaviour could have been due to changes in intentions relating to a further birth. Moreover, consistency between intended and actual behaviour is not necessarily constant over time: like intentions, it can change. These two aspects can be studied by using the data collected at the intermediate stage of the survey in 1976.
Consistency between intended and actual behaviour
Consistency between intended and actual behaviour
Changes in women’s intentions not to have any children between 1974 and 1979
22Between different stages of the survey, women can have confirmed their intentions or modified them upwards or downwards (cf. Diagram 2). The three replies by each woman, therefore, yield 27 different cells, and we have defined nine different trends.
23For each of the two periods (1974-1976 and 1976-1979), we have calculated the distribution of the women interviewed, and compared their stated intentions in 1976 with those of 1974 and their stated intentions in 1979 with those of 1976.
24These proportions, which were calculated for women who had borne no children since 1974, can be interpreted as the probability that a woman who had expressed a particular intention at the beginning of the period, would state the same or one of the two possible different intentions at the end. For the two periods together (1974 to 1979), these probabilities will be equal to the product of the probabilities relating to each of the two periods. The results are shown in Table 4 (the numbers in bold type will be commented on in the text).
Changes in intention
Changes in intention
25Changes in intention varied considerably and depended on the nature of the original intention and, to a lesser extent, on the size of the family that the woman had already achieved.
Changes in intention(1),(2)
Changes in intention(1),(2)
(1) +: Intends to have another child; –: Does not intend to have another child;?: Undecided.(2) Results not calculated because of small numbers.
The results in brackets are based on fewer than 15 cases.
26Slightly more than one-third (36.9%) of the women who wished to have another child but who had not done so (Part I of the table) confirmed their original intention in 1976 and 1979. If we also count the women who had revised their original intention in 1976 but who had reverted to their original intention in 1979, the proportion of women who, in 1979, stuck to their intention in 1974 is 43.1% (36.9 + 3.2 + 3.0). Changes in intention were more rapid amongst mothers of two children (when their intention related to their third child) than amongst those who had only had one child: as early as 1976, 37.3% of the former group stated that they did not want any more children, as against only 16% of the latter. These downward revisions of original intentions were confirmed by similar proportions of women (68.8% and 65%) at the end of the period 1976-1979. Finally, roughly one-quarter of mothers of one child were undecided in 1979.
27This proportion was slightly higher than in 1976 (20.5%). In general, it would seem that the women who did not have the child they had planned tended to give up the idea of having another child after five years.
28In comparison, the "robustness" of the intentions by women who did not want any more children (Part II of the table) was quite remarkable, in particular from the second child on: 79.5%, 87.9% and 90.3% of mothers of two, three or four children respectively, who had no further births, confirmed their original intentions in 1976 and 1979. A small minority had changed their mind in 1976 – e.g., 9.7% of the mothers of two children then declared themselves as "hesitating" – but, in 1979, most of these women, provided they had not had a child in the meantime, reverted to their original intention. Finally, only a very small percentage of mothers of two or more children, revised their original intention upward (between 2 and 4%).
29On the other hand, women who, in 1974, did not wish to have a second child tended to revise their intention upwards much more frequently (in 1976, 14.3% wished to have a second child), but one out of two (49.6%) confirmed their original intention, in 1976 and in 1979.
30Finally, among women who did not have any specific intention (Part III of the table) in 1974, a majority stated both in 1976 and 1979 that they did not want any further children.
Variations over time in consistency between intentions and actual behaviour
31Consistency between intentions and behaviour can change from one period to the next, in the same way that intentions can.
32Since the length of the two periods is not the same – slightly over two years for the first, almost three years for the second – the proportions of women who had at least one child (P) were recalculated on an annual basis using the formula 1 – (1 – P)12/n, where n was respectively equal to 26 and 33 for the two periods. The frequency of realized intentions (or "consistency ratio") is equal to the proportion of women who had produced a child in the case of the women who had intended to have another birth or were undecided, and to the difference between this proportion and unity in the case of women who did not wish to have any more children. These ratios have been calculated in relation to the latest statement of intention and have been uniformly transformed to an annual basis.
33The results (Table 5) measure the consistency between intentions and actual behaviour during each period, according to intentions in 1974 and, for the second period, depending on whether these intentions had or had not been revised in 1976.
Frequency of agreement between intended and actual behaviour (on an annual basis), 1974-1976 and 1976-1979(1)
Frequency of agreement between intended and actual behaviour (on an annual basis), 1974-1976 and 1976-1979(1)
(1) Results not calculated because of small numbers.34These results suggest three comments:
35— Consistency is significantly higher, during the second period, for women whose original intention, no matter what it was, remained unchanged in 1976. Amongst the women who, in 1974, desired another child, roughly one-quarter (24.8%) produced one, during each year of the period 1974-1976, and almost one-third (31.4%) did so during the second period, if only those women who, in 1976, still intended to have another child are considered. In the same way, 95.3% of the women who, in 1974, did not want more children did not, in fact, have a birth in each year between 1974 and 1976. This proportion reached 97.6% between 1976 and 1979, amongst those whose intentions had remained the same in 1976.
36— Consistency between intentions and behaviour, between 1976 and 1979, was significantly lower amongst women who had revised their original intentions than amongst those whose intentions remained unchanged. Amongst women who, in 1976, wished to have a(nother) child, almost one-third of those who had already wanted a child in 1974 produced one (31.4%), compared with only one-fifth (21.1%) of those who in 1974 stated that they did not want any more. Among women who, in 1976, declared that they did not intend to have any children and whose intention was the same in 1974, 97.6% each year did not, in fact, produce a child. Amongst the women who had only one child in 1974, the corresponding proportions were respectively 93.6% and 87.3%.
37— Finally, women who, in 1976, were undecided had, on average, about the same number of children between 1976 and 1979, whatever their previous intention had been. However, those who were in this group and already had a child tended to have a second child slightly more often, if, in 1974, their intention had been to have another child (17.7%) than those without such an intention (13.7%).
38In short, consistency between intended and actual behaviour depends, on one hand, on how long the intention has been held (the longer, the more "robust" it is) and, on the other, on the nature of previous intentions: the behaviour of women who express the same intention, at a given moment in time, will differ between those who had previously expressed a different intention and those who had not. Women who had revised their original intentions downwards had slightly more children than those who, at the outset, did not want any more children and, conversely, those who revised their original plans upwards will have slightly fewer children than those who wanted another child at the outset.
39Furthermore, it should be noted that time has opposite effects on intentions and behaviour of women who wished to have a child. With the passing of the years, their intentions tend to change, but, at the same time, the degree of consistency between intention and behaviour becomes stronger. This contradiction is a reflection of the underlying selection procedure: women who, in 1976 or 1979, had still not produced the child they desired in 1974 undoubtedly experienced greater difficulty in conceiving or, more generally, were subject to conditions which were less favourable to a birth. Revising their original intentions downwards can, therefore, be interpreted as an acknowledgement of their difficulty or even inability to conceive.
40Conversely, amongst women who did not want any more children, the passage of time tends to harden their original intentions and to increase the degree of consistency between intention and behaviour.
Overall view and comments
Fertility intentions and fertility forecasts
41One of the objects of this longitudinal study was to assess the possibility of forecasting the fertility of a group of women, from a knowledge of their fertility intentions.
42To begin with, we could use aggregate data. Amongst the 1,351 women whose replies were analysed in 1974, 484 intended to have another child; now, between 1974 and 1979, 496 children were born [6]. But this almost perfect consistency is an illusion: it results from various significant discrepancies according to birth order, which compensate each other.
43The study of data for individual women, which is undertaken in this article, tends to show up significant differences between intentions and behaviour, especially for the women who intended to have a second or a third birth. In total, between 1974 and 1979, more than one-quarter of women did not have the child they originally intended, and one-fifth did have a child, even though they had originally reported that they did not wish to have any more.
44The overall predictive value of statements about intentions can be judged by weighting the consistency rates in Table 3 by the proportions in the various groups concerned (without taking into account the women whose intentions were indefinite (Table 6).
45Considering all birth orders together, the index thus calculated is 77.3%, with the variations by birth order being relatively low. Given this result, what can we say about the predictive value of fertility intentions? Two aspects must be clarified before offering a reply:
46— On one hand, the fact that, for 77% of the women, actual behaviour corresponded to their intentions would suggest that these intentions were useful in predicting individual behaviour;
47— However, at the same time, this figure of 77% shows the limitations of fertility intentions for estimating future births: the intrinsic predictive quality of the statement about intentions, measured after allowing for compensations between births which had been desired, but not realized, and between births which were not intended, was relatively mediocre [7].
Overall predictive value of fertility intentions (1974-1979)
Overall predictive value of fertility intentions (1974-1979)
48These results may appear to be disappointing from the point of view of the forecaster. However, they throw some light on the very notion of fertility intentions, on certain aspects of fertility behaviour and, more generally, on what is usually called family planning.
What is intended fertility?
49The notion "intended fertility" presents certain advantages compared with others used for judging attitudes to family size ("ideal number of children", for the community or for oneself, "desired number of children", under present conditions or under certain other assumptions):
- it does not require the respondents to interpret notions which are inevitably subjective ("ideal", "desired");
- the respondent is personally involved: it is her intention which is in question, not a "generalized" intention.
50However, this notion implies a number of assumptions.
51Fertility intention can refer either to completed family size (including the children already born) or, as in the present study, to intended additional births. In the two cases, the existence of a fertility strategy is assumed: to a given intention, there corresponds a coherent pattern of behaviour, in particular with regard to the practice of contraception.
52Our results lead us to question whether such rational behaviour is generally followed. Differences between intentions and behaviour may come about, for two reasons:
53— Intentions are at risk of being revised downwards, particularly for reasons not foreseen at the time. The fact that the proportion who confirmed their intentions as stated in 1974 both in 1976 and 1979, was more than twice as high among those who did not want more children as among those who intended to have an additional child (80% and 37% respectively, without distinction of family size, cf. Table 5), leads us to believe that the degree of realism implied in a statement of intention depends on the nature of the intention. In general, it is easier to revise an intention downward than upward, as regards fertility as well as in other matters. We conclude that some of the women who declare their intention to have another child are not expressing a well thought-out strategy, but a simple possibility, which results in too high an estimate of future births.
54— Actual behaviour may contradict intentions as described in reply to the questionnaire, particularly among women who said that they did not want any further children. Thus, only one-quarter of the women who had a child, having stated that they did not intend to have another one (and one-tenth of those who had only one child), had given a positive reply to the question (put in 1979): "when your last pregnancy occurred, were you doing anything to avoid becoming pregnant?".
55In short, the difference between intended and actual behaviour, and the successive revisions of intentions, lead us to believe that there is a degree of uncertainty about fertility intentions, the strength of which depends on the nature of the intentions, and on the way in which these have developed over time, and also on the number of children already born. This uncertainty is least for women who have already had at least two children and do not wish to have any more, whose intentions change less, and whose behaviour more often accords with their stated intention. Uncertainty is considerably greater for women with two children who wish to have a third, and for those who do not wish to have a second child. Overall, this uncertainty leads to an overestimation of the number of future births.
56This conclusion has been reached in all surveys in which fertility intentions and effective fertility have been compared. Westoff and Ryder, who studied a cohort of women in 1970 and 1975 [8], postulated that the difference between intentions and behaviour reflects the influence of contemporary conditions on intentions (in that respect, intentions would not differ from any other index of attitude or opinion). Freedman et al. confirmed this hypothesis over a longer period (1962-1977) [9]. However, both these studies were undertaken when fertility was decreasing. To verify the conclusion, we would need a similar study undertaken during periods when fertility was increasing: do intentions underestimate future numbers of births in such cases?
57Fertility trends in France, during the late 1970s, were such that an answer to this question could be given.
58In effect, the overall drop in fertility, which was very marked during the ten years preceding the first stage, began to slow down in 1976, and there was even a slight increase in fertility around 1980. This movement was due to a recovery of legitimate births of second and third orders in particular (cf. Table 7).
Period parity progression ratios, per 1,000
Period parity progression ratios, per 1,000
59As we have shown previously, the actual number of births has, generally speaking, been lower than the intended number and no noteworthy upward revision of intentions was observed in either 1976 or 1979.
60But this does not invalidate Westoff and Ryder’s hypothesis, it merely leads us to question whether fertility intentions, though admittedly reflecting, to some extent, contemporary conditions, may not be subjected to systematic bias of overestimation, which could be sufficiently large to conceal variations caused by changes in conditions.
61In short, the inherent uncertainty of intentions, even when limited to the immediate intention of having an additional birth, casts doubt on the likelihood that the majority of women or couples have a specific completed family size in mind. It would appear that decisions are made sequentially, with intentions being revised with each new birth and reflecting the conditions prevailing at the time, that is, in the short term. Each birth would thus be more or less planned, but this does not mean that we can speak of family planning.
Enquête 1974
Enquête 1979
Enquête 1976
Notes
-
[*]
Translated by Bobbie Le-Texier.
-
[**]
INED.
-
[1]
A. Monnier. – La naissance d’un enfant, Travaux et Documents de l’INED, cahier n° 81, Presses Universitaires de France.
-
[2]
The questions relating to intentions, asked in the original survey, and the questionnaires used in the second and third stages, are published in the Annex.
-
[3]
A. Monnier. – "Projets de maternité et comportements réels", Population, 4-5, 1978.
-
[4]
It would, moreover, have been interesting to know the number of women who had had either an abortion or a miscarriage. It is, however, difficult to judge the reliability of replies to questions about these events in a self-administered questionnaire.
-
[5]
These results, and those which follow, take no account of women who were already pregnant in 1974, during the first stage of the survey (of whom 88 were also observed in 1976 and 1979).
-
[6]
Not taking into account the women who, in 1973, were undecided, or the children they produced.
-
[7]
The calculations presented here, and the conclusions they sustain, relate to births of second and higher orders. For all births, the overall predictive value of the intentions would depend, to a great extent, on the predictive value of the intentions of childless women.
-
[8]
Charles F. Westoff and Norman B. Ryder: "The predictive validity of reproduction intention", Demography, 14 (4), November 1977.
-
[9]
Ronald Freedman, Deborah S. Freedman, Arland D. Thornton: "Changes in fertility expectations and preferences between 1962 and 1977. Their relations to final parity". Demography, 17 (4), November 1980.
