|Home | About | Journals | Submit | Contact Us | Français|
How do contemporary fertility ideals, desires, and intentions relate to contemporary low fertility? At the empirical level the answer is straightforward: observed fertility is well below the levels of ideal family size and also usually well below respondents' desires and intentions. In fact, below-replacement fertility in many countries would disappear if respondents' fertility intentions were realized (Bongaarts 2001, 2002; Goldstein, Lutz, and Testa 2003).
At the conceptual level, all behavioral models of contemporary low fertility feature choice: individuals choose to have children (Thomson and Brandreth 1995). Thus fertility intentions, if not ideals, hold the promise that actual fertility could mirror intentions over the long run, thus bringing observed fertility closer to replacement. Consistent with this view, Bongaarts (2002) argues that a substantial part of the discrepancy between aggregate fertility intentions and contemporary fertility levels can be accounted for by timing shifts, specifically the postponement of childbearing to later ages. Once this postponement abates, as it eventually must,1 intentions and current behavior will be more similar. In addition, frustrated demand for children could encourage institutional adjustments that would allow childbearing to correspond to levels desired by women and couples, again raising fertility to levels that approximate intentions.
Expressing an alternative view, Goldstein, Lutz, and Testa (2003) report that German-speaking areas of Europe now show levels of ideal and expected family size well below replacement levels (i.e., less than two children per woman), a pattern they suggest portends a permanent shift to sub-replacement fertility. They argue that the previous disjuncture between intentions and behavior in German-speaking counties, now evident in many other countries as well, represents “cultural lag.” Specifically, women and couples act in accordance with contemporary antinatalist constraints and have few children, but continue for a period (a time lag) to express the prevailing cultural imperatives to be parents and to have at least two children. Younger cohorts, once they have lived much of their lives in a low-fertility setting, adopt the status quo as an expectation and thus express intentions (and ideals) that more closely correspond to contemporary behavior. Goldstein et al. (2003) hypothesize that fertility intentions and ideals in Italy, Spain, and elsewhere will soon decline further, creating greater symmetry between subreplacement fertility and subreplacement ideals and expectations. The suggestion that aggregate shifts in intentions can follow rather than precede behavioral shifts is not new (see Lee 1980; Westoff and Ryder 1977; Morgan 1981, 1982).
Finally, Demeny (2003) argues that a wide gap between expressed preferences and behavior can persist indefinitely. Even if such preferences for a particular number of children are genuine, these preferences compete with others, and the result need not be reconciled by having children. Demeny characterizes the deficit of actual-to-preferred number of births as those that occur only “in the minds of their would-be parents.” Current social and economic conditions, he argues, encourage a persistent shortfall of the actual number of births compared to what women say they would wish to have. Hence, this shortfall provides little evidence against long-term subreplacement fertility even when expressed fertility preferences are at replacement level.
Thus while we juxtapose these positions as starkly as possible and acknowledge that they need not be mutually exclusive, reasonable arguments can be made that: i) fertility will rise soon to more closely approximate intentions, ii) intentions will fall soon to approximate subreplacement levels of fertility, or iii) a shortfall of births relative to fertility preferences will persist. Given such a range of expectations, do we conclude, as Demeny's argument suggests, that data on preferences are irrelevant to debates about “how low will fertility go”? We argue no. First, we note the inconsistency in taking seriously the disjuncture between intended and realized fertility when the observed outcomes are greater than the former (e.g., unwanted fertility) but not the reverse (when fertility falls short of intentions or expectations). In both cases, the intention–behavior disjuncture puts the interest of the individual and the collective in agreement, that is, realizing individuals' stated intention or preference increases the likelihood of a desired aggregate outcome (i.e., replacement-level fertility). Such an aligning of individual and collective goals is opportune for policy interventions.
Second, a large theoretical and empirical literature focuses on the predictive validity of reproductive intentions. Morgan's (2001) review of this literature concludes that intended parity is not a consistently accurate predictor of completed fertility for individuals or of aggregate fertility for cohorts. However, a model proposed by Bongaarts (2002) that we use below provides a framework for explaining intention–behavior inconsistency. Further, examining cross-national variation in the extent to which intentions match behavior allows us to assess societal features that promote correspondence between intention and behavior.
Figure 1 provides data that are consistent with key claims above and that situate and motivate our research. Data from European countries (for countries of the EU-15, with data for Germany shown separately for West and East) come from the 2001 Eurobarometer Study (see Goldstein, Lutz, and Testa 2003). The US data come from sources examined in detail in this article and described fully below. Note that US average ideal family size is among the highest, but it is not unique. As in many other countries, mean ideal family size in the United States is clearly above 2.0. What is striking and unique about the US estimates is the coupling of the mean levels of intended fertility with correspondingly high levels of period fertility (TFR). The contrasting and dominant pattern is for fertility intentions to substantially exceed levels of contemporary fertility.
In this article, we 1) discuss the import of fertility ideals and intentions for understanding fertility levels, 2) propose a model that can account for variable inconsistency between attitude and behavior, and 3) use this model as a framework to examine trends in American women's fertility ideals, intentions, and actual fertility.
Family size norms are relevant because they narrow the range of likely choices by endorsing particular choices. Questions asking “ideal family size” exploit this normative dimension. At the extreme, some choices (to be childless, for instance) could lie beyond the realm of conscious choice—never actively considered. More commonly, various family sizes are apportioned relatively more positive (or negative) endorsements than others. Given a range of choices weighted by acceptability and the exigencies faced by individual women, asking about a woman's or couple's intention for additional children provides an indicator of demand for children that is proximate to the key fertility behaviors (sexual activity, use of contraception, and decisions to carry a pregnancy to term). In fact, stated intentions for additional children are strong predictors of subsequent fertility—consistently among the strongest predictors examined by social demographers (see Schoen et al. 1999). Further, these arguments and evidence are consistent with claims that intentions or expectations operate as a key mediating or proximate variable in predicting fertility behaviors.
On the other hand, as noted above, the literature on the predictive validity of reproductive intentions clearly shows that many individuals do not realize their fertility goals and that these discrepancies can cumulate so that mean aggregate intentions depart substantially from mean aggregate levels of observed fertility (for a recent empirical example see Quesnel-Vallée and Morgan 2003).
To reconcile these claims and following Bongaarts (2001, 2002), we place fertility intentions at the heart of a model including other factors that condition the extent to which these intentions are realized (see Morgan 2003; Morgan and Hagewen 2005). We do not argue that intentions consistently play a dynamic role in contemporary fertility change. We do, however, assert that fertility intentions take on a central role in understanding fertility trends. In fact, Bongaarts does not attribute developed-country fertility differences to dissimilarities in intentions. Rather, he argues that cross-country and cross-time variation must be explained by changes in ages at childbearing and couples' ability and determination to realize intentions. Like Bongaarts, we argue—and, for the United States, empirically demonstrate—that there is a remarkably pervasive desire (and supporting norms) for a family size of two children.2
Specifically, Bongaarts (2001, 2002) offers a conceptual model to account for differences between fertility intentions and behavior like those shown in Figure 1 (also see Morgan 2003; Morgan and Hagewen 2005). The framework is:
In this conceptualization, the level of current fertility (TFR) equals the intended parity (IP) of women increased or decreased by a set of model parameters that reflect forces not incorporated into women's reports of their childbearing intentions. The foundation of this framework is the concept of intended parity. If all women realize their parity intention, then TFR = IP. The model parameters that can augment completed parity vis-à-vis intended fertility include: unwanted fertility (Fu) and, of far less consequence, replacement of children who have died (Fr) and additional children needed to satisfy strong gender preferences (Fg). One would expect these effects to be greater than 1.0 and thus to augment observed fertility relative to intentions. Other factors represent parameters that (at least in recent periods) would be expected to take on values less than 1.0 and thus reduce fertility relative to intentions. These factors include changes in the timing of fertility (Ft), subfecundity and infecundity (Fi), and competition with other energy- and time-intensive activities that may lead people to revise downward their intentions (Fc).3 This model provides a useful beginning schema to systematically examine the discrepancies between intended fertility and observed behavior in Figure 1.4
This model and these comments are intended to dissuade simplistic accounts of the relevance of ideals and intentions for predicting future levels of fertility. What is needed is an understanding of these intentions and the set of factors that modify or frustrate them. Further, this model places intended parity at the core of a conceptual model of contemporary fertility. Below we examine this concept.
Should we expect reports of ideal family size and intended parity to be changing? To answer this question, we first define and discuss these concepts in some detail. We then review arguments that suggest either change or stability in these indicators.
Ideal family size is conceptualized as the preferred number of children for some hypothetical family.5 The question about ideal size is problematic because the characteristics of the hypothetical family are not specified. Clearly, normative family size varies by circumstance,6 and one can only assume that respondents adopt a “typical” family as the referent. Other criticisms of the question have focused on the answers given. First, answers of “no children” are relatively rare but are essentially illogical since “0 children” cannot be an ideal number for a “typical family.” A one-generation population “implosion” is the obvious result. One-child answers present a similar problem. Given these answers, we might assume that some persons are actually reporting their own “ideal,” “expected,” or “intended” fertility. Finally, ideal fertility is criticized because it is “uninteresting” in the sense that it does not seem to change. This last criticism is the easiest to dismiss. There are changes in responses to the ideal family size question, as we will show (also see Goldstein, Lutz, and Testa 2003), and the stability that exists may be real and substantively important. In sum, despite its flaws, many demographers consider ideal family size to be a useful general indicator of the degree of societal pronatalism (Trent 1980). Ideals reflect the degree of the normative context within which fertility intentions are formed and expressed. In the United States, midsized families (two or three children) are preferred over both larger families (four or more children) and smaller families (no children or one child), with larger families being preferable to not having children at all (Sensibaugh and Yarab 1997).
Intended parity is an individual's or couple's report of their intention and thus is a proximate determinant of their actual behavior. Conceptually, the major concern about reports of number of intended children revolves around the sequential nature of births. Since births generally occur one at a time and not in lots (Namboodiri 1972), operationally relevant behavior (e.g., having sex, using contraception) is parity specific. Further, logically speaking, family plans can be reassessed after each birth. In subsequent sections we demonstrate that intentions for a certain number of children, while changeable, have substantial stability. Below we assess parity-specific responses as well as the intended number of children and we assess uncertainty and nonresponse to see whether they provide clues regarding the stability of reproductive intent.
With reference to the United States, the literature we review below suggests the following: 1) strong norms against childlessness persist but some evidence suggests that these may be weakening; 2) norms against one-child families persist, but they too may be weakening; 3) norms against large families were historically weak, but plausible arguments suggest that they may be strengthening. In general, norms should be reflected in reports of ideal family size, and norms also constrain intentions.
Regarding norms about childlessness, virtually all societies have endorsed parenthood (Blake 1979; Ganong, Coleman, and Mapes 1990) and have stigmatized voluntary childlessness (Jamison, Franzini, and Kaplan 1979; Magarick and Brown 1981; Mueller and Yoder 1999; Park 2002; Pohlman 1970; Shields and Cooper 1983; Somers 1993; Veevers 1972). In American society, having a child is a key marker of the transition from adolescence to adulthood (Hogan and Astone 1986; Rossi 1968); and, for women in particular, motherhood is central to overall identity (Russo 1979). Thus, voluntary childlessness is considered by many as a form of deviance, not because such people do not have children, but because they want none and therefore reject the parenting role (Veevers 1980). This deviance of the intentionally childfree is thought to reflect on individuals' overall personalities (Houseknecht 1987), and individuals often express a desire for more social distance from those who eschew parenthood (Polit 1978).
Negative views toward voluntary childlessness, however, may be changing. In fact, over the past decade several books celebrate the choice to be childfree (Burkett 2000; Cain 2002; Casey 1998; Lisle 1999; Morell 1994; Safer 1996). These authors make the obvious point that women can lead healthy, productive lives whether or not they decide to reproduce (Safer 1996) and can draw satisfaction from their own contributions and accomplishments (Casey 1998). They further argue that American society has become a “culture of parental privilege,” with the childfree unable to take advantage of benefits, such as tax credits, childcare subsidies, school vouchers, flextime, and parental leaves (Burkett 2000), and that American society's unwavering pronatalism ultimately disadvantages all women (Morell 1994). Indeed, Gillespie (2003) argues that women no longer view motherhood as evidence of their femininity.
In addition, the romanticized version of the “traditional family” persists even while behavior has changed dramatically. Cohabitation, nonmarital childbearing, and union dissolution are contemporary realities. Moreover, about 20 percent of individuals aged 65 or older report that they have never had children (Wu and Hart 2002), and the percentage of women still childless at ages 40 to 44 has increased from 10 percent in 1980 to 19 percent in the late 1990s (Coltrane 1998). Therefore, despite the purported stigma attached to voluntary childlessness, more individuals and couples have this status. Raising a child to adulthood is, and has been, an exceptionally expensive undertaking (Espenshade 1984). In 2002, the annual cost of a child less than two years old for middle-class parents is estimated at about $9,000, and for a 15–17-year-old the cost is slightly over $10,000 (Statistical Abstract of the United States: 2003, Table 675). In addition to these direct costs are the indirect costs of lost wages, typically of the mother, due to leaving the work force to care for the child (Crittenden 2001). Also, substantial research indicates that children have a negative effect on marital happiness (Belsky and Kelly 1994; Heaton et al. 1996). Becoming a parent almost always involves major changes in lifestyle, including the loss of freedom of activity and flexibility in one's schedule.
These strong antinatalist forces shape declining ideals and intentions and have led some to ask: net of “outmoded norms,” why do people choose to have any children (Miller 2003; Morgan and King 2001; Schoen et al. 1997)? Clearly, beyond genetic dispositions shaped by evolution, there are social influences motivating individuals to have children. Our experience of being reared in families encourages us to have a family of our own, and our friends who have children encourage us to do likewise. Children symbolize a kind of immortality and establish a link to both past and future. They can also give the parent an enhanced status, create social capital (Schoen et al. 1997), and provide opportunities to become better integrated in the community (Ambert 1992).
Similar conflicting expectations focus on “only children.” Historically, few considered a one-child family to be the ideal, and it is generally deemed disadvantageous to be an only child (as is documented by Blake 1981 and Veenhoven and Verkuyten 1989). Despite the argument that only children do not form a singular, homogeneous group, but instead are representative of different types (Rosenberg and Hyde 1993), there are many negative stereotypes of only children (Blake 1981; Thompson 1974; and Westoff 1978 among others). Only children are often seen as socially unskilled, self-centered, dependent, anxious, and generally maladjusted (Terhune 1974; Thompson 1974). These stereotypes are supported by some evidence. Research suggests that only children are less social (Claudy 1984) and less able to make friends (Miller and Maruyama 1976), more likely to suffer mental distress (Belmont 1977; Howe and Madgett 1975), and more self-centered and less cooperative (Jiao, Ji, and Jing 1986; Thompson 1974). On other dimensions, especially those relating to socioeconomic status, singletons perform well (Blake 1981). In fact, only children have been found to be superior to children with siblings on several measures of intelligence, achievement, maturity, leadership, health, and life satisfaction (Blake 1981). For our purposes, however, we need not reconcile these conflicting findings. As Blake (1981) points out, if prospective parents believe an only child is disadvantaged, they will be motivated to have at least two children. Clearly, many contemporary Americans believe that only children suffer social or psychological disadvantage. It is worth noting, though, that research in this area is quite dated, and it is certainly possible that feelings toward only children have shifted over time. Yet recent research still indicates that, compared to those who have one child or are childfree, those who have two children perceive greater approval from parents, in-laws, other relatives, friends, and acquaintances regarding their family size (Mueller and Yoder 1999).
Finally, what about norms regarding large families (those of three, four, or more children)? Some claim that large family sizes are viewed in a positive light overall (Mueller and Yoder 1997; Polit 1978). One can argue that the pronatalist twist in American mass media actually supports large families (Faludi 1991). In the mid-1970s Blake (1974) showed that few men or women were willing to characterize family sizes of three or four as “too large.” She concluded that, in contrast to strong norms against having fewer than two, norms against having more than two were quite weak. However, more recent evidence finds pressure to limit childbearing after the third or fourth child. Women with four or more children feel that others view them in a negative light and assume they cannot give as much to their children, that they are too busy to tend to their house, and that at least some of the pregnancies must have been accidental (Mueller and Yoder 1999). With the addition of each child to a family, the economic resources per child diminish. Even if economic resources were theoretically unlimited, time and energy are not. Parental interaction with children and correlates, such as knowing their friends, are negatively affected by additional children (Downey 1995). Increasingly, many claim that having a large family is inconsistent with being a good parent (as noted, e.g., in Morgan 2003).
This study uses data from two US sources: the General Social Surveys and the Current Population Surveys. The General Social Survey (GSS) is a national area probability sample of noninstitutionalized adults collected annually from 1972 to 1994 and biennially beginning in 1994. The survey netted a sample size of about 1,500 in-person interviews reporting ideal and intended family size for the first 19 surveys and 3,000 in-person interviews when the survey became biennial in 1994. We used surveys from the past three decades for this study.
The Current Population Survey (CPS) has asked questions on fertility intentions since the 1970s. Small changes over time in mean intentions and general challenges to the usefulness of this measure have discouraged recent data collection. The CPS has asked about fertility intentions with decreasing frequency, more recently only in 1992 and 1998. We use selected June Current Population Surveys over the past two decades: 1980, 1985, 1990, and 1998. CPS sample sizes are large, ranging from over 30,000 women aged 18–39 years in 1980 to slightly less than 20,000 in 1998.
Substantial numbers of both GSS and CPS respondents do not answer the question on intentions or they report uncertainty. Our tabulations include only those answering yes or no to the question on intending to have an additional child. Only these respondents were asked the additional question, “how many more children do you intend?” We return to the issue of nonresponse in a subsequent section of the article.
Figure 2 shows the trends in ideal family size for the US population aged 18–46 over the past 30 years. The popularity of the different family sizes remains roughly the same throughout. Two children is always the most popular choice, followed by three children, four or more children, and one child or no children. Chi-square tests, however, indicate a significant difference in responses over time.7 Changes are primarily seen in the decrease in the popularity of four or more children over time.8 In the 1970–74 period, 19.7 percent of the sample indicated that four or more children would be ideal. This number plummets to 9.5 percent in 2000–02. This change is consistent with arguments that large family sizes are viewed increasingly less positively (and perhaps negatively) by a substantial segment of the population.
Although not shown here, we have also examined the trends in ideal family size for the older segment of the US population (those aged 47 or older).9 While these respondents are unlikely to become parents at these ages, they are in a position (e.g., as parents, relatives, etc.) to influence younger persons. We find that the popularity of the different family sizes has changed among older respondents in ways that parallel changes in Figure 2, providing additional evidence for the emerging pervasiveness of a two- or three-child ideal. The primary evidence is the increase over time in the popularity of the “two children” response and corresponding decreases in those responding “four or more.” In 1970–74, 37.8 percent of the sample indicated that two children would be ideal. This number climbs to 50.9 percent in 2000–02.
Finally, we examined trends in ideal family size reported by women aged 20–26, those who will have the highest rates of childbearing in the five-year period following the survey. We chose young women for special focus because this is the group for whose fertility decisions the normative environment is of greatest immediate consequence. We do not show these data here because the results are hardly distinguishable from those in Figure 2.10 Again, two children is the most popular response in every year, followed by three children, four or more, and one child or none. As in Figure 2, there is a significant decline, from 16.6 percent to 10.6 percent, in the proportion of those indicating an ideal family size of four or more children over the period 1970–2002.
Examination of Figure 2 shows that the percent giving the “two children” response declined in the most recent data (GSS 2000–02). However, the shift is not toward smaller ideal family size; it is a small shift toward more diversity. Specifically, all categories increased vis-à-vis “two children.” This change was not expected and is confined to one survey. Rather than offer post hoc explanations, we await additional data.
CPS and GSS female respondents were asked, “Do you expect to have any (more) children? If yes, how many (more)?” By definition, when one adds the number of additional children expected to the number the woman has (i.e., her current parity), one obtains the variable intended parity. We attach no substantive import to the distinction between intended, expected, or desired children. While demographers sometimes distinguish between these terms, evidence suggests that respondents generally do not (Ryder and Westoff 1971).
Using CPS data, Figure 3 shows the intended parity of the 1959–60 birth cohorts by age. Specifically, these women were aged 20–21 in 1980, and ages 25–26, 30–31, and 38–39 in subsequent survey years (1985, 1990, and 1998). The mean fertility intentions for these cohorts remain stable over time. At no point does intended parity fall below two children and at no point does it rise above 2.25 children. Also shown at each time point is current parity. Logically, these two curves must converge as women reach the end of the childbearing years. However, the pattern shown here need not hold. In this figure, intended parity stays largely fixed and actual parity converges to it—that is, in the aggregate women's stated intentions are met.
At the individual level, we know that these intentions change and that they can change disproportionately upward or downward, challenging the value of intentions for projection purposes (see Westoff and Ryder 1977). But Quesnel-Vallée and Morgan (2003) show for these same cohorts that downward revisions in intentions were largely offset by upward revisions. In fact, in the CPS data for the 1959–60 cohorts (see Figure 3) intentions are realized almost exactly. Specifically, when members of this cohort were 20–21 years old (1980), the average number of children intended was 2.05. By 1998, when members of this cohort were 38–39 years old, the average number of children intended was 2.11. Of course, these data are not longitudinal, and compositional changes (e.g., larger numbers of Hispanics and other immigrants) may have increased this value slightly; however, the impact of immigration is minor. With the 1998 CPS data it is possible to exclude immigrants from the tabulations. We show these points in Figure 3, and, while lower, the impact of excluding the foreign-born on estimated levels is small.
With the series of cross-sectional surveys examined here, we cannot follow all cohorts for the 18-year period shown in Figure 3. However, the data we examine produce very similar results. For instance, Figure 4 displays in period perspective the current parity, additional children intended, and intended parity of women aged 20–26. The top line in the figure shows little change in intended parity, which is the sum of current parity and additional children intended. In the 1970–74 and 1990–94 periods of the GSS, the intended parity was 2.19 and 2.10, respectively. Similar stability is seen in the CPS data, where intended parity was 2.03 in 1980 and 1.98 in 1998. Beneath this stability, one can see clear declines in current parity, particularly in the GSS data, an indicator of fertility delay. We interpret this decline as postponement because of the observed mirror-image increase of additional children intended. At the aggregate level, this postponement did not lead to substantial declines in intended parity for young women (as was shown in Figure 3 as well).
A final way to display these results follows a strategy used by O'Connell (2002) and addresses a criticism of intended parity. Specifically, intended parity combines three variables: current parity, the parity-specific intent for at least one additional child, and a report of how many more children are expected. One could argue that information and understanding are sacrificed by the intended parity measure that we employed above. Table 1 shows the first two variables: intent for at least one additional child by current parity. We include here data presented by O'Connell (2002) as well as from the CPS time periods we examine. The repeated numbers in parentheses for 1998 indicate that the CPS data yield exactly the results of O'Connell (2002).
For the youngest age group (18–24), intentions have not varied across the 1978–92 period for all women in this age group (column 1, “All parities”) or for groups disaggregated by current parity (columns 2–4). There is evidence, however, of a decline in intent for at least one additional child between 1992 and 1998. This decline is pervasive with respect to current parity and is not evident for any of the other age groups in Table 1. This result is potentially important as it may signal that the youngest cohorts have intentions for having additional children that are significantly lower than preceding ones. We hesitate, though, to interpret this finding substantively on the basis of a result from a single age group in a single survey.
Table 2 shows expected parity by age, current parity, and CPS survey year. This presentation allows one to disaggregate intended parity into two components: current parity and additional children intended. These data are not independent of those shown in Table 1; all those saying yes in Table 1 intend at least one additional child. The new information added here consists of reports of number of additional children intended. Consistent with estimates in Table 1, there is evidence of a shift in intentions between 1990 and 1998. Specifically, young women with no children and with one child intend fewer children in the later survey. Such a decline is not visible for those with two or more children or for other age groups.
A weakness of both the GSS and CPS data is the large proportion of women who are “uncertain” of their intention or do not answer the question on fertility intentions.11 Here we have excluded women who did not give “yes” or “no” answers to the question “Do you intend to have a (another) child?” These uncertain responses and nonresponses may have a substantive interpretation. We have examined these data closely for young women and have found additional evidence that the decline in intent is substantial. For instance, between 1992 and 1998 we have shown that the percent of young childless women who intend to have no children has increased. Additional analyses (not shown here) indicate that the percent uncertain has also declined. Morgan (1981, 1982) has argued that respondents may reduce intentions by first becoming uncertain and then, at an older age, acknowledging that they will have no more children. The decline in uncertainty among young women here requires a different interpretation. Specifically, firm choices to have no children may signal an increasing proportion of women who see the costs of childbearing as too high and an accompanying acknowledgment of increasing preference for lower fertility.
Figure 1 showed that many low-fertility societies have childbearing intentions well above current fertility levels. In this article, we have focused on the United States, an important and interesting exception. Reported fertility intentions of American women approximate the country's contemporary period levels of fertility, and cohorts recently reaching the end of their childbearing years showed, in the aggregate, both stable intentions across time and an ability to realize those intentions.
Our results have implications for future levels of US fertility and for a more general understanding of low fertility and for discrepancies between fertility intentions and behavior in contemporary contexts. We return to the general question we raised in the introduction, which was usefully framed by Bongaarts (and formally expressed by equation 1): why do we observe large differences between stated intentions and period fertility in many countries? The Bongaarts model suggests that intentions are central to understanding levels of fertility, while not completely determining these levels. To repeat, the framework above posits that observed fertility (the TFR) is augmented relative to intentions by factors such as unwanted fertility. Conversely, observed fertility (relative to intended parity) is attenuated by fertility postponement, infecundity, and competition with other activities. This framework makes explicit that congruence between aggregate intention and observed fertility depends on the first set of factors compensating for the later ones. This appears to be the case for the United States, but these factors need not be fully compensating. Morgan (2003; also see Morgan and Hagewen 2005) provides a hypothetical example of how these two types of factors can cumulate differently for two countries to produce strikingly different levels of observed fertility and substantial differences between intended and observed fertility. Morgan and Hagewen (2005) present evidence consistent with these illustrative examples. Thus, intentions as the fundamental parameter of the model deserve careful study, as do the factors that augment or attenuate actual fertility relative to intentions.
Our focus here has been on US ideal family size and fertility intentions. The primary question has been whether there is evidence that the strong preference and intent for moderate family sizes have eroded with time. The evidence is remarkably clear: the dominant American ideals and intentions are for two or three children; these preferences have persisted across the last three decades. As noted, the concept of ideal family size is problematic in several ways. However, we interpret evidence of relative constancy of response as consistent with the presence of strong norms that support having two or three children. Figure 2 and other analyses discussed show that “two children” is the modal response, with “three children” as the clear second choice among respondents. This stability does not mean the item is insensitive to change. Rather, there has been relatively little change, and broad norms about family size have remained relatively stable. The major change observed was a decline in the percent saying four or more children; as Morgan (2003: 593) has argued about this shift in perception: “Being a good parent is now largely inconsistent with having more than a small number of children.”
As is well known, US fertility has also been quite stable over the past three decades. Modest observed declines below replacement fertility in the 1970s and 1980s can easily be attributed to postponed childbearing (see Bongaarts and Feeney 1998). Changes in other parameters must be offsetting, given the model expressed in equation 1 and the trends that we documented in fertility and intended parity.
Thus, we conclude that the “exceptionalism” of American fertility is not fleeting; it has been evident now for over a decade. Further, continued replacement-level fertility is a reasonable forecast in the United States. This forecast rests on two observations: 1) current levels of fertility intentions are consistent with replacement-level fertility, and 2) past evidence indicates that factors augmenting fertility versus factors attenuating fertility relative to intentions have roughly balanced. Further, we think it reasonable that this circumstance will persist (see Morgan and Hagewen 2005). Of course, this forecast would be stronger if we understood better all factors in the model above. Bongaarts's framework provides, therefore, an important research agenda.
One strand of evidence that runs counter to these statements and this forecast warrants attention. The most recent intention data (1998) suggest a decline in intended parity among young women ages 18–24. These data (in Tables 1 and and2)2) suggest a possible crack in the normative expectation of two or more children. We urge caution in interpreting results from the most recent survey while stressing their potential importance. Only additional data can determine whether recent data are an aberration or the beginning of a trend that portends appreciably lower fertility in the United States. Note that the latter explanation is consistent with arguments and data presented on Europe by Goldstein, Lutz, and Testa (2003). Data now available from the 2002 National Survey of Family Growth will give a more recent view of trends in intent for additional children. It is also crucial for the CPS and GSS to again ask questions on fertility intentions in the near future. Fertility intentions remain the central concept for understanding contemporary fertility trends and differences. Constancy in this social indicator for a substantial period of time suggests that while yearly monitoring may not be crucial, periodic (say on a two- or three-year cycle) monitoring of fertility intentions should be given high priority.
We thank Nan Astone, John Bongaarts, and Ronald Rindfuss for helpful comments. This research was supported by Duke University and a grant from the National Institutes of Health, National Institute of Child Health and Human Development, R01 HD41042. The authors' names are listed alphabetically and first authorship is shared.
An earlier version of this article was presented at the 2004 Annual Meeting of the Population Association of America.
1Postponement, measured as increases in mean age at childbearing, cannot continue forever given a fixed age-range of childbearing. The timing effect disappears once age at child-bearing stabilizes.
2We acknowledge evidence that voluntarily childfree couples show greater levels of cohesion, dyadic satisfaction, and life satisfaction than do parents (Somers 1993), and that, while marriage has been found to increase global happiness, the presence of children, particularly within married couples, has been found to significantly decrease global happiness (Campbell 1975; Campbell, Converse, and Rodgers 1976; Glenn and Weaver 1979). We also acknowledge substantial gender change over the period of study. Yet, despite these facts and changes, a persistent average desire for having children remains.
3Decisionmakers incorporate “competition,” i.e., costs of childbearing, into their reports of intended parity. The attenuation represented by this parameter results from the unanticipated intensity of competition that becomes apparent as the life course unfolds. Measurement of this component would certainly be challenging, but this is not our focus here.
4Equation 1 is used here as a conceptual model, a first approximation of a set of proximate determinants of low fertility. The model assumes independent effects, a representation we adopt for parsimony. In addition, we defer model elaboration (interactions among model components) until warranted by additional evidence and conceptualization.
5Question wording in the General Social Survey: “What do you think is the ideal number of children for a family to have?”
6For instance, Ryder (1973: 61) characterized norms of the 1960s as prescribing that persons should get married and have two children as soon as possible, provided that they could afford them. The economic referent, provided that they could afford them, illustrates the conditional nature of childbearing norms.
7χ2 for model of independence of response and year=229.8, df=12, p<.001.
8χ2 for model of independence of 4+ versus all other responses by year =144.6, df=4, p<.001.
9Interested readers may request these data from the second author.
10Interested readers may request these data from the second author.
11Interested readers may request from the second author tables showing the proportion uncertain and details of the analysis leading to the conclusion reported here.