In recent years, data suggesting that 49-57% of pregnancies to American women are mistimed or unwanted at conception have focused increased attention on the concept of pregnancy intendedness.1 This statistic implies that contraception may not be as widely accepted as one might think (given national data on the proportion of women who have ever used a method) and is certainly not practiced with the greatest possible efficiency. Yet it also raises questions about the definition of intendedness, a concept that has been explored largely from a demographic perspective. Because demographers seek measures for populations, it is perhaps not surprising that the constructs formulated to measure intendedness are not appropriate in every context. Although such measures are somewhat predictive at the aggregate level,2 and although individual desires and intentions are often related to births, there are large minorities for whom behavior and intention are poorly correlated. Our understanding of these constructs is, therefore, often questionable at the individual level—the level at which decisions about contraception and pregnancy resolution are made. To comprehend the relationship between reproductive attitudes and intentions, and in turn the link between intentions and behavior, a broader understanding and a more subtle definition of intention are probably needed.
One might well ask what the components of such a definition should be. In previous work, we showed that ambivalence plays an important role in an adolescent's attitude toward pregnancy;3 it may well play a major role for older women as well. Here, we suggest that a definition focusing more explicitly on individual women and acknowledging the role of ambivalence would necessarily include partnership variables. Indeed, in modern societies, where unions are often unstable and sexual contact frequently takes place outside of unions, women's pregnancy intentions may well be defined in terms of the changing dyads within which they occur.
Early formulations of intendedness depended largely on two assumptions: that couples have, at some level, a concept of "ideal family size" that they see as appropriate to their social context or as optimal for their own family situation; and that all pregnancies occurring before that ideal is reached are "wanted" and all those happening afterward are "unwanted." This basic formulation required the assumption that births take place in stable marriages; long ago, this method of assessing excess fertility was described as assuming "idealized conditions...namely, no divorce or separation and fixed family goals...."4 The assumptions underlying such models—e.g., that all women marry and remain married—have been described as "simplistic."5
The formulation now widely used distinguishes conceptions to women wanting no (or no more) children, described as unwanted, from those categorized as mistimed; conceptions in either of these categories are considered "unintended."6 Even this definition implies that women believe their present pregnancy intentions will extend into the future because it asks a woman to discriminate between the desire to postpone childbearing and the desire never to conceive at all. That dichotomy assumes either that a woman has a childbearing goal that is independent of her partner's desires, or that she knows what is in her future—a future that may include partnerships as yet unformed, indeed often unanticipated, at the time of an index pregnancy. At the least, such formulations, which are clearly not relevant to adolescents,7 may not apply to adult women in unstable relationships, or even to cohabiting or married women in unstable unions. The notion that a woman "knows" (i.e., beyond some vague abstraction), at a given point in her life, that at some future time she will or will not want a (or another) child is, for many women, questionable.
The applicability of many of our definitions of intention to modern societies, then, is limited. For example, the desire to postpone pregnancy has generally been seen as weaker than the desire to have no (or no more) pregnancies at all. But the relationship between intention and behavior—as reflected in rates of contraceptive failure—is significantly affected by marital status and age:8 In the current societal context, a young or single woman's motivation to postpone pregnancy may be stronger than an older married wom- an's desire to avoid pregnancy. The body of literature that, over the last several decades, has addressed the subject of couple preferences in childbearing has generally done so in the context of marriage,9 even while the proportion of child-bearing that takes place within that context declines.
Indeed, the concept that fertility intentions are sequential, that they change over time and parity, has long been a part of the literature. Although most of the hypotheses summarized in a review of a parity-specific formulation10 do, indeed, suggest the presence of a married couple, one proposes a role for "marital disruption." One other study also addresses the relationship of marital disruption and childbearing.11 But when a third of all childbearing and an even larger proportion of conceptions occur out of wedlock, when premarital or nonmarital sexual contact is frequent and many pregnancies occur in the context of sequential or even simultaneous partnerships, and when marriage itself is often transient, neither the notion of a fixed union nor the idea of a fixed desired family size (to which both members of a conception dyad may or may not subscribe) appears to describe the context of pregnancy and childbearing.
Perhaps some of the problem in understanding the intention to conceive stems from a lack of distinction among the various psychological underpinnings of the construct of intention—hence of the measurement of the complex process by which such abstract notions are translated into intention. An exploration of this process using couples rather than women alone describes a sequence moving from childbearing motivation to desire to intention to behavior.12 (Unfortunately, the distinctions between desires and intentions can rarely be captured in current population samples.) In viewing that sequence as a dynamic involving both partners, this research finds that the relative influences of the husband and the wife differ in magnitude in different segments of the sequence. Its approach, drawing more on the social-psychological tradition than on the purely demographic one, suggests that "child-number desires cannot validly be used as a proxy for childbearing motivation,"13 and that relationships between these constructs may vary over different parities14—and therefore, presumably, over the life cycle.
Attitudes and behaviors that result in unintended births have generally been defined as attributes of a woman15 rather than of a relationship. Many studies focus only on the woman, although the literature frequently calls for study of the partner's role.16 Analyses based on the National Survey of Family Growth (NSFG), even when they address the desires of both, must rely on the woman's perception of her partner's attitude.17 Considerable research, however, has addressed the childbearing plans of husbands and wives, the relative strength of each partner's intentions in relation to future births or the effects of agreement or disagreement between them;18 recently, cohabiting couples have been studied in a setting where cohabitation is a common form of union.19
It is well demonstrated that a marital or cohabiting partner exerts an important influence on a woman's intentions, but the role of partners in less formal relationships is not as clear. A partner may influence the social and economic circumstances of a, woman's life at the time of conception; in addition, he can affect her reproductive decisions in many different ways. For example, a woman's sexual partners, relationships, acts, drives, and sexual and gender roles may be related to her childbearing intentions and decisions.20 In models that posit a role for ambient norms,21 the partner can be seen as a part of the setting, which has normative influences on the woman's own desires. The partner may actually control the decision, especially in settings where a woman's power to negotiate is weak. Or, a woman's perceptions of future as well as present partnerships may influence her childbearing decisions. For instance, limited access to acceptable life partners may cause a woman to settle on choices different from those she would make were her expectations of a future union more optimistic.
Finally, a woman may also be influenced by her knowledge—or perception—of her partner's reproductive desires; she may be more likely to believe he wants her to conceive than to wish to do so herself.22 Here, we must propose the following question: Is the reason a woman does not want a pregnancy at any specific moment in time because she does not want to conceive with her current partner?23 Our focus, then, is not on the effects of the partner's desires or intentions, but rather on a woman's own desires in the context of an existing partnership.
In a study population in which women with multiple partnerships were oversampled, we explore demographic, attitudinal and other correlates of intention, including the nature and number of a woman's partnerships. We then examine the "not with this partner" variable in the context of these several correlates of intention. We cannot elucidate here the dynamic process through which the partner's influence is translated; we merely propose here how a woman's partnerships can contribute to her expressed intentions relative to a particular conception. Thus, we argue for the inclusion of variables on individual partnerships—or at least on a woman's perception of her partnerships—in future models of childbearing intention.
Data and Methods
Data were collected from 263 women enrolled in a study at the Bayview Medical Center in Baltimore between February and September 1993. All were urban, and most were low-income—63% received social services, or food stamps, or both. Some 44% were recruited in the hospital's Center for Addiction and Pregnancy, 29% were patients in a clinic for long-acting contraception (the implant or the injectable) and 27% were enrolled in the hospital's regular family planning, obstetric or gynecologic services. The women, of whom 38% were multiethnic white and 62% were black, were aged 18-41. Fewer than 5% had never been pregnant, and 23% had never borne a child. Almost half (47%) were pregnant at interview. Only 35% had ever been married; at interview, 16% were married and 23% were living with a partner to whom they were not married.
Fewer than 14% said they had had only one or two partners over the course of their lives; the median number reported was five. The maximum number of serious partners was 14, and the median was more than two. (We defined serious partners as those with whom the respondents said they had had more than a single or casual sexual experience.) In addition, half of the women had had more than two casual partners, and one-quarter reported more than seven; the highest number reported was 78.
Thus, we oversampled women who demonstrated a range of behaviors that would appear to place them at high risk of unintended conception. Although the study population is in many ways an excellent one in which to identify and explore new explanatory variables that may relate to women in nonmarital partnerships, our findings should not be generalized to more representative populations. Furthermore, some caution must always be exercised in using retrospective attitudinal data; these women were being asked to recall their intentions with respect to pregnancies that may have occurred many years before. That they recalled such complex and changing patterns—patterns that appear to relate well to the other life events they report—is noteworthy. Nonetheless, we should understand that some of their recollections must have been tempered by subsequent events.
Our extensive data on these women include sexual, marital, residential and fertility histories—the beginning and end of each serious relationship and additional information on sexual behaviors, contraceptive use and childbearing with each of several key partners. The women were asked what they considered the best family size for themselves, and were also asked, in a different context, to report the wantedness of each of their lifetime conceptions.
If a woman reported that her intention at the time of conception was to have a child, she was asked if she had wanted a child with the man with whom she had conceived. Similarly, if she did not intend to conceive at that time but wished to do so in the future, she was asked if her reason for not wishing to get pregnant was that she did not want to conceive with her current partner. When a particular conception was unwanted, the woman was not asked "Why not now?" Among the 250 ever-pregnant women in the sample, only 8% reported that they had wanted no (or no more) children at the time of each conception, but almost half gave that response in connection with at least one pregnancy.*
We conducted analyses on several levels that required somewhat different methods and are therefore presented separately: We explored demographic, attitudinal and behavioral correlates of intention and of the "not with this partner" variable relative to each pregnancy these women had experienced (n=839), and relative to each ever-pregnant woman and her lifetime pattern of intention (n=250). We then examined intention relative to a woman's changing partnerships over time. Finally, to explore relationships between intention and ideal family size or ambivalence, we focused on the most recent pregnancy because these constructs were measurable in our data only with respect to that conception.
For analyses on the level of the individual pregnancy or woman, we began with bivariate relationships and then constructed multivariate models using either intention to conceive or "not with this partner" as the dependent variable. When we examined the intention outcome relative to all pregnancies, we included as independent variables not only characteristics of those pregnancies but also characteristics of individual women; we incorporated these characteristics separately and then combined them in a single model. To allow us to include both levels in a single model, we used Zeger and Liang's generalized estimation equation approach to the modeling of correlated data.24
We used two intention categories. The first was the outcome as it is treated in most studies, as intended or unintended pregnancy; the latter category included both unwanted and mistimed conceptions.25 Then, because we had not asked about partner relationships in regard to pregnancies occurring when the woman wanted no more children (or none at all), we included only pregnancies that are intended now and those that are mistimed—that is, not intended now—in models with that response as the outcome.
Analysis by Individual Pregnancy
The 250 ever-pregnant women in the sample had experienced a total of 839 pregnancies; they reported intention status for 825 (98%). Unwanted or unintended conception and childbearing were frequent: Some 57% of their pregnancies were not intended at the time of conception, and more than half of these were to women who reportedly wanted no (or no more) children. Among the 215 mistimed pregnancies, 18% were not wanted because the woman did not want to have a child with her conception partner. Even among pregnancies described as wanted "now," 7% were not wanted with the conception partner. (For the 253 pregnancies that occurred when women wanted no more children, or none at all, some might also have cited partner relationships as relevant to their intentions if they had been asked.)
We developed three models examining pregnancy intentions (Table 1). In the first, in which the intention to conceive a specific pregnancy is the dependent variable, we first incorporated independent variables related to individual pregnancies; correlates of intended conception include being older at conception (odds ratio, 1.04) and being married at conception (2.61).
In the second model, which examines variables specific to individual women, having had a higher cumulative number of serious partners significantly reduced the probability that any particular conception was intended. Having graduated from high school also lessened that probability, but the effect was only marginally significant. As one might expect, when the specific pregnancy represented a higher parity than the number the woman cited as best for her, she was less likely to have intended to conceive. In addition, black women were more likely than white women to report a specific pregnancy as unwanted. In the combined model, all of the woman-level variables except race had highly significant effects, as did two pregnancy-level variables, having been married at the time of conception and age at conception.
Many of these same variables were significantly related to the response "because [I] didn't wish to get pregnant with that particular partner" (Table 2). Not surprisingly, women were less likely to cite this reason in connection with a pregnancy occurring in a cohabiting relationship than with one occurring in a noncohabiting relationship. Being in a serious partnership at conception worked in the same direction (although it was significant only in the combined model), but women who reported more serious partners over their lifetime were significantly more likely to cite a current partner as a reason to wish to postpone a particular conception.
In an analysis that included all pregnancies experienced by the women in our sample (not shown), we explored the relationship between the "not with this partner" response and pregnancy outcome. (Among these pregnancies, 84% were carried to term.) We asked: Are women who do not want to conceive a specific pregnancy by their current partner more likely to terminate it than those who do not cite this response? The probability that a pregnancy was terminated was significantly greater when the response was cited than when it was not (33% vs. 14%). Nonetheless, of the pregnancies that occurred despite the mother's desire not to conceive with her current partner, two-thirds (67%) were carried to term.
Analyses by Individual Woman
In these analyses, we explored the lifetime patterns of intention reported by the respondents. The majority of these women (80% of those who had ever been pregnant) had had at least one unintended conception, and 51% had had more than one. Only 21% claimed that all their pregnancies had been intended at the time of conception (Table 3). Some 54% of the women responded that at the time of at least one of their conceptions, they had wanted a child in the future but not at the time they conceived; 48% had had at least one conception when they had wanted none (or no more). More than one in five (22%) reported both unwanted and mistimed conceptions.
On a bivariate level, women who had ever had an intended conception were older at interview and were more likely to have been married or to have cohabited with a partner (or both) than were those whose pregnancies were all unintended. Similarly, women who were using or had ever used drugs, those who had traded sex for drugs those who had had more casual partners—or any combination of these— were significantly more likely to report at least one intended conception.
In a multivariate model including factors significant at the bivariate level, age and drug use remained significantly and positively associated with intended conception (Table 4). Lower economic status (as indicated by the receipt of social services and/or food stamps) also had a significant and positive effect in this model. However, having had a greater number of serious partners lowered the probability of ever having intended to conceive; the number of casual partners no longer had a significant effect. (This finding results from a high correlation between numbers of casual partners and the use of drugs; when drug use was left out of the model, having more casual partners was significantly and positively associated with intention.)
When the dependent variable was defined as women whose conceptions were all wanted (even if some were mistimed), age had the opposite effect: The longer a woman was exposed to the risk of pregnancy, the less likely she was to have experienced only wanted conceptions. The number of casual partners was only marginally significant, and no other variables except race had a significant effect. Race was included in this model because—consistent with the models described above—black women were more likely than white women to have said, at some time, that they wanted no more pregnancies or none at all.
When considering the "not with this partner" response on the level of the individual woman, we excluded the 20 women who never wanted to conceive and limited the sample to 227 women who had had at least one wanted pregnancy, whether or not it was mistimed.† Among these women (the majority of whom had had both wanted and unwanted pregnancies), having had more serious partners was significantly and positively associated with the "not with this partner" response. Having had at least some wanted conceptions was negatively related and significant. Overall, 21% cited the "not with this partner" response in connection with at least one of their mistimed or intended pregnancies.
Among women who had had both intended and unintended conceptions, 19% attributed their desire to postpone at least one pregnancy to not wanting a child with a particular partner. The women most likely to cite that reason were those who reported that none of their pregnancies had been intended; 33% of these women gave that response in connection with at least one conception. More surprising, however, is that even among the subgroup who described every pregnancy as wanted and appropriately timed, almost one woman in five (18%) claimed that, at least once in her life, she had not wanted to conceive by the man who impregnated her.
Changes in Intention by Partner
We then asked whether a woman's responses changed over time, by partner, or both. Was a woman likely to change her childbearing intentions? Was she likely to make the "not with this partner" statement about one partner and not another? Among the 201 women with more than one pregnancy, 65% changed their intentions between pregnancies, many of them more than once (Table 5). For some women, an intended pregnancy was followed by an unintended conception, while for others, the sequence was reversed; in many cases, further changes occurred.‡
Using this simplest formulation of intention, we asked whether a change in intention was likely to be related to a change in partnership. (Changes in partnerships were counted only if they occurred during an interval in which intention changed between one pregnancy and the next; there might actually be more than one partner change during the interval.) Among women who changed intention only once between adjacent pregnancies, 66% experienced at least one partner change in the same interval. Among those with two changes in intention, 81% experienced a change in partner during one or both of these pregnancy intervals, more than 44% during both (not shown). Among women whose intentions changed three or more times, a partner change was associated with at least one switch in intention for 94% of the women and with two or more such changes for 37%. This finding suggests that a woman's changing partnerships may play some role in changed pregnancy intentions, whether or not she explicitly cites that reason.
Similarly, the number of partnerships was a factor: Women who said that they had ever been impregnated by a partner with whom they did not want to conceive reported a significantly greater average number of serious partners over time than those who had not (4.1 vs. 3.3). As we indicated earlier, this variable remained significant in some of our multivariate models. In addition, women who had ever had a pregnancy that was not desired with their current partner reported higher numbers of casual partners (13.2 vs. 8.6); because of this variable's large range, however, the difference did not attain statistical significance.
The Most Recent Pregnancy
It is often assumed that women can give clear and unequivocal responses to questions about their intention to conceive. (These questions are generally presented with only yes or no responses, and few women respond "I don't know.") In our work with adolescents, we found a remarkable degree of ambivalence in that response.26 In this study, we used the same methodology—in which a construct consisting of two or three items was used instead of a single question to describe intention—to test the level of ambivalence toward the current or most recent conception. When the responses reflected the same attitude relative to childbearing across items, we considered that the woman unequivocally intended or did not intend to conceive; if they did not, we considered her ambivalent.
When we used several constructs,§ many women appeared ambivalent about childbearing (30-56%, depending on the model). In turn, ambivalent women were very likely to say they had not intended to conceive their current or most recent pregnancy because they did not want a child with their current partner. In fact, they were more likely to give that response than were women who unequivocally did not want to conceive.
One might hypothesize that a woman's intention to conceive would be closely associated with the relationship between her current and her preferred family size. Indeed, that variable remained significant in the combined model in Table 1. A woman's most recent (or current) pregnancy was the one occurring closest to the interview at which she reported the "best" family size for her. Nonetheless, the parity of the child resulting from that pregnancy rarely exceeded the family size she reported as "best." Despite all of the mistimed and unwanted pregnancies they reported, only 6% of the women in our sample (12% if all current pregnancies are counted as births) reported more children than their preferred family size.
Similarly, the relationship between women's reported "best" and actual family sizes does not necessarily accord with their stated intention regarding their last (or current) pregnancy. Among those whose families were still smaller than or equal to their stated preference at that conception, 48% reported their last pregnancy as intended and 26% reported it as mistimed, not as unwanted. However, among the few who had already exceeded their preferred family size, 36% nonetheless said the last conception was intended and 21% reported it as mistimed, rather than saying it was totally unwanted.
In this study population, we note that an individual woman's childbearing intentions and behaviors vary across partners, events and time periods. Three findings suggest that women's childbearing goals are not fixed. First, women who wanted no children at one time in their lives did want them later. Second, very few women, no matter how many children they had, reported that they had more than their preferred family size. Finally, many women's current statements about intention had little relationship to the ratio of their actual to desired family size.
The data, then, call into question the formulation of a family size "goal"; they suggest that, if such a goal exists at all, it is a moving target. However, they also call into question the notion, on which most current research depends, that a woman can distinguish between unwanted and mistimed conceptions. She may be quite certain that a pregnancy is unwanted when it occurs, but have no clear idea whether she wants to conceive in the future. And even if she does, her desires may change.
As we noted, we oversampled women whose sexual behaviors and low economic status—and, in many cases, other behaviors such as substance use and trading sex for drugs or cash—would suggest, based on the literature, that they are at high risk of unintended conception. (It is interesting to note that the proportion of all pregnancies reported as intended—43%—is exactly the same as that reported in national data collected at about the same time.27)
The data suggest a different interpretation: In some of our multivariate models, so-called "high-risk" variables emerged as significantly related to a woman's statement that at least some pregnancies were intended. Similarly, in one model, women who had ever graduated from high school were more, not less, likely to report an unintended conception. These findings are not conclusive, as they vary according to the definition of the pattern of intention; they suggest, however, that the traditional way of looking at "intention" may be flawed, and even judgmental. Perhaps the belief that women in certain circumstances should not choose to become parents has led to the assumption that they are at risk of unintended pregnancy, when that is not the way they would describe their own conceptions. These findings suggest yet another reason to rethink the very definition of intendedness.
Using the common definition, these women reported almost three in five of their pregnancies as unintended; furthermore, among both pregnancies that were mistimed and those that were not mistimed, a considerable number were not intended with the current partner. It may be facile to suggest that the desire not to conceive with a current partner is more frequent among women whose lives reflect the instability associated with multiple partnerships than it would be in a population of women with fewer partners and more stable living arrangements. That interpretation would be consistent with the popular concept of intention noted above—that "high risk" pregnancies are frequently "unwanted."
However, an opposite hypothesis suggests itself: Women who have little reason to anticipate marriage at any time in their lives may have less, not more, reason to postpone childbearing to some future partnership. Perhaps women who look forward to a stable family in the future, finding themselves temporarily in nonmarital or unstable partnerships, would cite this reason even more often than the women in our present sample. It would require a very different population to test this hypothesis adequately.
To understand the intention to conceive more fully, it will be necessary to explore some substantive factors other than timing that might influence a woman's desire (or lack of desire) for a child, or, in the context of timing, to explore reasons for her desire not to have one "now." Our findings suggest that, especially among women with multiple sequential partners, the desire to postpone childbearing may have little to do with an absolute notion of ideal family size and much to do with the nature of the conception dyad.
On the level of the individual woman, we found that the number of changes in intention between pregnancies increased with number of partner changes; thus, women appear to change not merely over time but over partners. When we focused on individual pregnancies, the nature of a woman's current partnerships and living arrangements had a highly significant effect on her childbearing intentions. Not surprisingly, a serious partnership was associated with a lower probability that the "not with this partner" response would be cited in relation to a specific pregnancy and with a higher probability that the pregnancy was intended. However, women with a pattern of many serious partners over time—and thus more changes over time—were more likely to report a pregnancy as unintended and more likely to cite not having wanted to conceive with a given partner as the reason.
Many more subtle questions can be raised: For example, why were women with cohabiting partners less, not more, likely than those with marital partners to cite that response? In this population, is cohabitation expressive of commitment—perhaps because, when undesirable, it is more easily terminated than an undesirable marriage? Such refinements would have to be explored in larger, more representative samples, but there is evidence here that the partnership component of pregnancy intention requires further study.
In a recent study of adolescents, we identified a high level of ambivalence toward childbearing and proposed a method for assessing ambivalence in future surveys.28 One of our items—a scale measuring how happy a woman was to be pregnant—was adapted for Cycle 5 of the NSFG; testing it against the traditional measure suggested that, in the aggregate, they measured the same concept. However, although ambivalence (as defined by a response between four and seven on a 10-point scale) is most common among women aged 15-19 (31%), it reaches almost 20% among women aged 20-29.29 In our sample, ambivalence clearly plays a major role, affecting a substantial proportion of women and—although we do not know the direction of causality—increasing the likelihood of a "not with this partner" response.
A potentially important reason for ambivalence is reflected in our finding that, even when only pregnancies reported as wanted and appropriately timed are considered, 7% were conceived when the mother did not want a pregnancy with her current partner. Further, even among women whose pregnancies had all been intended, 18% had experienced at least one pregnancy with the wrong partner. Yet, women who had not wanted to conceive with the partner who impregnated them were twice as likely to carry the pregnancy to term as to abort it. If, indeed, one of the important reasons for concern about the intentionality of pregnancy is concern for the consequences of unintended childbearing,30 future research might well address the implications of this variable for the subsequent support and well-being of the child. For implications such as these, and many others not enumerated here, a definition of intention is needed that is explanatory at the individual as well as the aggregate level.
It could be argued that partner relationships may help explain intention but do not increase our ability to identify unintended conceptions. Indeed, most of the women who said they had not wanted a pregnancy with their current partner had already told us that their pregnancies were mistimed. However, even a woman's statement that she intended to conceive at the time she did does not guarantee that she wished to do so with her actual partner at conception. Without a direct question about intention relative to that partner, a number of births fathered by men with whom the mother did not wish to conceive would be counted merely as intended. While her own statement of intention cannot be dismissed, the meaning of intention under such circumstances is, at best, ambiguous. At the very least, then, these data illustrate the importance of new, partner-related items in investigating the issue of pregnancy intention, and suggest how limited classical approaches to the issue may be when the conception does not occur in the context of a stable, monogamous and life-long partnership.