Low Fertility at the Turn of the Twenty-First Century

Abstract

In the past few decades, demographic concerns have shifted from rapid population growth fueled by high fertility to concerns of population decline produced by very low, sub-replacement fertility levels. Once considered a problem unique to Europe or developed nations, concerns now center on the global spread of low fertility. Nearly half of the world's population now lives in countries with fertility at or below replacement levels. Further, by the mid-twenty-first century three of four countries now described as developing are projected to reach or slip below replacement fertility. We review the research on low fertility through the predominant frameworks and theories used to explain it. These explanations range from decomposition and proximate determinant frameworks to grand theories on the fundamental causes underlying the pervasiveness and spread of low fertility. We focus on the ability of theory to situate previous and future findings and conclude with directions for furthur research.

Introduction

Sub-replacement fertility and concerns about its consequences are not new phenomena. In fact, one could argue that such poor reproductive fitness is the oldest and most general threat to a population or species. Many births, 5 or 6 per woman, do not guarantee population replacement if mortality is very high. Throughout much of human history, population growth was slow and erratic, with an approximate balance of high birth and death rates. In contrast, the past few centuries have witnessed fertility rates well above mortality rates owing to the rapid decrease in mortality vis-à-vis fertility. This rate discrepancy produced rapidly growing populations, especially in the post-1950 period.

In most contemporary developed countries, mortality below age 50 is now very low, and thus the replacement level of fertility is slightly above 2 births per woman (i.e., usually taken as 2.1) (see Espenshade et al. 2003). Contemporary concerns about low fertility have resulted from period fertility rates well below this level. Such low fertility is not unprecedented. In the 1930s, demographers noted fertility levels in Europe well below 2 births per woman (Keyfitz & Flieger 1968; also see Kohler et al. 2002), along with the inevitable consequences. Specifically, persistent sub-replacement fertility produces rapidly aging populations, declining workforce size and smaller overall population size. These changes may have cascading negative consequences (see Teitelbaum & Winter 1985, Wilson 2001).

These previously observed low fertility rates were not sustained. Instead, birth rates increased during the baby boom of the 1940s, 1950s, and 1960s and assuaged previous fears of population decline (Teitelbaum & Winter 1985). In fact, coupled with high fertility in the developing world, the baby boom pushed concerns of a population explosion to the forefront of public attention in the 1960s, 1970s, and 1980s (e.g., Ehrlich 1974). But this situation has again reversed. By the late 1980s and early 1990s, countries such as Germany, Spain, and Italy reported fertility rates well below replacement level (<1.5), refueling concerns about very low levels of fertility.

Figure 1 shows fertility levels for selected world regions over the past 50 years. Our measure of fertility, the total fertility rate (TFR), is the most commonly used measure of fertility; it is an age-standardized period measure that can be interpreted as the number of births a woman would have if she experienced current age-specific rates throughout her lifetime (and she did not die prior to the end of reproduction). In the 1950–1955 period, Figure 1 shows that fertility ranged from 2.3 in Northern Europe to 5.7 in East Asia. Over the next two decades, the TFR dipped below 2.1 in all these regions. The pace of decline in East Asia is especially striking and signals that low fertility is not a uniquely Western phenomenon.

Figure 1.

Regional fertility (TFR), 1950–2005.

Less immediately striking than the pace of declines is the variation in fertility levels across low fertility countries. Table 1 illustrates this variation, showing recent TFRs for selected low fertility countries. TFR estimates range from 1.25 (Greece) to 2.04 (United States). To illustrate the population dynamics of such low fertility, we calculate and present estimated negative growth rates implied by these TFRs and existing mortality conditions.¹ Again the range is set by Greece and the United States, −1.7% to −0.1% per year, respectively. Table 1 also shows the number of years required for these 2000–2005 rates to produce a population one half the current size: 41 years to over 1000 years. Clearly, the observed variation in TFRs is important. The implications of a TFR of 1.25 are vastly different than those of a TFR of 1.75, as are the potential policy responses to ameliorate the consequences (Caldwell et al. 2002, Demeny 2003). Thus, the concerns facing Greece, Italy, and Spain are very different from those facing France and Australia. For the latter two countries, moderate levels of immigration could offset sub-replacement fertility, producing population stability. For Greece, Italy, and Spain, only massive immigration could offset their very low fertility.

TABLE 1.

Variation in low total fertility rates

Country	TFR 2000–2005a	Implied growth ratec	Implied years to halved	Period when TFR fell below 2.1a	Years TRF<2.1e
Europe
Germany	1.32	−1.5%	46	1970–1975	30
France	1.87	−0.4	196	1975–1980	25
Russian Federation	1.33	−1.6	43	1965–1970	35
Spain	1.27	−1.6	42	1980–1985	20
Italy	1.28	−1.6	42	1975–1908	25
Greece	1.25	−1.7	41	1980–1985	20
Sweden	1.64	−0.8	88	1970–1975	30
Asia
China	1.70	−0.9	75	1990–1995	10
Japan	1.33	−1.5	46	1955–1960	45
Australia/New Zealand
Australia	1.75	−0.6	119	1975–1980	25
North America
United States	2.04	−0.1	1025	1970–1975	25
United States: white non-Hispanic	1.84b	−0.4	196	N/A	N/A

^aSource: UN Population Division (2005).

^bU.S. data for 2001 from Ventura et al. (2003).

^cSource: UN Population Division (2005). Calculated from net reproduction rate (NRR). Mean age at childbearing set at 30 for all countries.

^dTime to halve = ln(2)/growth rate.

^eCalculated from beginning of period TFR fell below 2.1 to 2000.

Finally, Table 1 presents the approximate duration in years of below-replacement fertility for these selected countries. No country has transitioned back to replacement-level fertility (for any five-year period) once falling below it. Also note that although much of the low fertility literature has focused on countries such as Italy and Spain, the Russian Federation and Japan have actually sustained sub-replacement TFRs for longer periods.

Thus, at the turn of the twenty-first century, fertility decline and sub-replacement fertility have become widespread. Both the current prevalence of low fertility and its persistence are unprecedented (see Frejka & Ross 2001, Population Reference Bureau 2004, Stark & Kohler 2002). Note that Wilson (2004) has estimated that more than half of the world's population now lives in countries with sub-replacement fertility. Further, note that only 3% of the world's population now lives in countries that have not begun fertility decline (Morgan 2003), and that past research suggests that once a fertility transition has begun, it does not stop until birth rates of 2 or below have been achieved (Bongaarts & Bulatao 2000, Bongaarts & Watkins 1996). Consistent with this claim, UN projections suggest that by the mid-twenty-first century, three of four countries now described as developing will have fertility rates at or below 2.1 births per woman (UN Popul. Div. 2003). These developments portend an end to global population increase shortly after the mid-twenty-first century (Bongaarts & Bulatao 2000, UN Popul. Div. 2004a). This forecast is remarkable, removed by only a decade or two from serious concerns about a global population explosion. Widespread fertility at or below 2 births per woman is an important and remarkable development whether one sees it as a necessary transition to a sustainable human population or as an ominous movement toward a continual crisis of low fertility.

Decomposition, Proximate Determinants, and What we Know about Low Fertility

In exploring the spread and variation of low fertility, some demographers have returned to “bread-and-butter” decomposition approaches as a way to describe low fertility regimes and their variations, approaches highly useful in explaining variation in, and declines from, high levels of fertility. Although these approaches have been described as nothing more than “elaborate description” (Hobcraft 2002, p. 131), they have provided valuable information about the proximate determinants of fertility variation that need to be addressed in building, evaluating, and revising theories. Specifically, these approaches have yielded much of the information that is agreed upon (i.e., “what we know”) and thus must be accommodated in plausible explanations of low fertility.

Decomposition approaches are important in revealing the behavioral components that require explanation. Useful decompositions rest on straightforward observations linked to the biological structure of reproduction: Fertility is a sequential, time-limited, nonreversible process. That is, women usually have babies one at a time (sequentially), between menarche and menopause (biologically time-limited) with experiences prior to and during each birth affecting subsequent decisions or actions (the process is path-dependent and not reversible). Especially important to low fertility is the fact that timing of births (i.e., the ages at which women have their babies) can vary dramatically. With intentions for only two children, women could reasonably expect to bear two children if they began at age 20, 30, or 35. Thus, a key question is whether aggregate fertility change (in a given period) reflects fertility postponement (timing) or a change in the number of births women will have (quantum).

Timing and Quantum

As exemplars of decomposition, consider Ryder's (1980) important explanation of the American baby boom and Bongaarts & Feeney's (1998) more recent and general examination of fertility timing. Both classic articles decompose fertility along two dimensions: parity (e.g., the number of prior births of the woman) and timing (the woman's age at giving birth). Our discussion here focuses on the TFR defined above.

If one decomposes fertility by parity, then one can see whether aggregate, current trends are produced by changes in the birth propensity of childless women, of those with one child, or of those at higher parities. For example, the transition from high to low fertility (the first demographic transition) has resulted primarily from the declining incidence of higher parity births (i.e., births to women who already have two or more children). Large families (or high parity births) are generally seen as diluting the parental resources that are available to each child and thus threatening children's full potential. This pressure to reduce family size is heightened by claims (probably of a more recent origin) that each child deserves individualized treatment and focused parental attention. Some researchers argue that informal sanctions against large families now accompany the high monetary and opportunity costs of higher parity births (i.e., large families) in many developed countries (see Morgan 2003).

In contrast, fertility trends in low fertility populations hinge primarily on the behavior of women with no or few births. For instance, Ryder (1980) shows that the fertility behavior of childless women and those with one child accounts for substantial parts of the American baby boom and bust (the 1945–1980 period). Bongaarts & Feeney's (1998, also see Sobotka 2004) work illustrates the crucial importance of first births for understanding cross-national fertility trends and differentials.

But the age of childbearing at each parity is also important. Shifts across birth cohorts are common and can be quite dramatic. To understand better the influence of timing shifts on the TFR, let us consider the TFR for first births only. Let us assume that 80% of women in each birth cohort will always have at least one child (i.e., quantum is fixed). If we allow for increases in the ages of first birth and assume that this trend toward later childbearing continues, the period TFR measure will dip below 0.8. In effect, births that would have occurred in the current year (under the existing timing regime) are postponed into subsequent years under the later timing regime, thus depressing the birth rate in the current year. Bongaarts & Feeney (1998) show that a rough estimate of this effect is indexed by the change in the mean age at childbearing.² For example, if the mean age of first births increases by 0.15 years each year, then the first-birth TFR is reduced by a factor of 0.85 (1.0 − .15 = .85). Thus, it follows that given an underlying quantum of fertility (say 0.80 having a first birth), postponement of fertility depresses TFR relative to underlying quantum (in this example to .68 = .80 * .85). Declines in mean age at childbearing have the opposite effect on the TFR.

These timing dynamics are important because fertility timing does change and can have prolonged and substantial effects on period fertility rates. Part of the reason for low rates observed during the 1980s and 1990s was a pervasive shift upwards in the ages at childbearing (especially controlling on the woman's parity). Bongaarts & Feeney show that timing shifts can reduce period rates by 10%–20% (i.e., by factors of .9 to .8) for up to two to three decades. Thus, a country with a TFR of 1.7 may not have any birth cohort of women that actually averages less than 2.0 births (2.0/.85 = 1.7). For many cases of “lowest-low” fertility (i.e., TFRs less than 1.5), up to one third of the fertility deficit relative to replacement levels can be attributed to timing shifts (see Sobotka 2004). Other countries, such as the United States, can attribute all years with below-replacement fertility to postponement (see Billari & Kohler 2004, Bongaarts & Feeney 1998).

The Transition to Parenthood

Consistent with above decompositions, many researchers have focused on the key transition in the fertility process, the transition to parenthood. This work is key for two reasons. First, it posits a framework, the life course perspective, for conceptualizing the effects of cultural, structural, and other factors on first-birth timing. First-birth timing is highly responsive to temporal and structural factors and thus is a major part of the story of timing changes discussed above. Second, although timing is important, the end of postponement would still leave many countries with fertility well below replacement (see Frejka & Sardon 2004, Lesthaeghe & Willems 1999). Low fertility explanations thus must account for both fertility postponements and for fewer births (lower quantum). Work focusing on the timing of parenthood shows the many ways that timing and quantum are interrelated. Specifically, in the aggregate, fertility postponed implies some fertility foregone (see Kohler et al. 2002, Morgan & Rindfuss 1999, Quesnel-Vallée & Morgan 2003). Three mechanisms are at work: (a) Later childbearing leaves fewer years at risk of an unintended pregnancy/birth; (b) later childbearing increases the risk of sub/infecundity; and (c) postponement allows women/couples to revise intentions, and these revisions tend to be disproportionately reductions (owing to the development of competing interests). The strength of the link between timing and quantum both at present and in the future is in dispute.³ At stake is the important question of the “degree of recuperation” or how much delayed fertility can be made up at older ages (see Frejka & Sardon 2004, Lesthaeghe & Willems 1999).

Hobcraft & Kiernan's (1995) “Becoming a Parent in Europe” has been cited by a number of authors in the recent debate about low fertility (e.g., the recent, provocative debate on low fertility in the journal Population Studies; see Caldwell & Schindlmayr 2003, 2004; Billari et al. 2004). The Hobcraft & Kiernan framework has many similarities with one proposed a decade earlier by Rindfuss et al. (1988) to explain “first births in America” and with one proposed a decade later by Sobotka (2004) to account for postponement of childbearing and low fertility in Europe. The comparable features of these frameworks can be taken as some long-standing and accepted features of a model of parenthood.

Specifically, all three frameworks propose the life course as an analytic frame, with the focus being a set of sequential decisions to continue to postpone or to have children now. The fertility decision is reached by balancing pro- and antinatalist forces that are rooted in the intersection of the unfolding life course and the social context in which the individual is embedded. Each of the frameworks acknowledges norms and attitudes held by relevant reference groups. Each also discusses constraints to responding to these macrolevel social pressures, but the factors stressed vary across authors.

Kohler and colleagues (see Kohler 2001, Kohler et al. 2002; also see Montgomery & Casterline 1996) offer an appealing model that builds on those discussed above. Consistent with these models, Kohler et al. (2002) conceptualize first-birth postponement as a rational response to socioeconomic incentives, insecurity, etc. Kohler et al. point out that these responses are insufficient and that one must consider the dynamics of social interaction. Specifically, Kohler and colleagues stress the importance of social multiplier effects: social interactions that can give rise to multiple equilibriums and status quo enforcement. For example, Kohler et al. (2002) suggest that increased female human capital and work opportunities encourage marriage and fertility postponement (a rational response to increasing opportunity costs of childbearing). But because these effects are pervasive (i.e., also influence the behavior of many others), they also lower the costs of postponement by making the pool of eligible partners at older ages both larger and of higher quality; this second indirect or multiplier effect results from the behavior of/interaction with others.

Status quo enforcement, in contrast, impedes change by maintaining positive sanctions for normative behavior and negative sanctions for new behavior. But norms weaken with pervasive violation because the imposition of sanctions becomes unfeasible. Thus, social interaction and collective movements can provide an incentive not just for delay but for shifts from one fertility equilibrium to another with much lower mean fertility.⁴

A Proximate Determinant Model of Low Fertility

Reminiscent of the very important conceptual and analytic contribution of the Bongaarts & Potter's (1983) proximate determinants framework, Bongaarts (2001, 2002) suggests an overarching aggregate framework for low fertility that integrates the work discussed above on the decomposition of timing/tempo and the life course theories focused on first-birth timing. Specifically, this approach decomposes the TFR into a set of multiplicative factors that represents a current (period) fertility regime. We develop the model a bit differently than does Bongaarts. Specifically,

TFR = F_t × (IP) × (F_u × F_g) × (F_i × F_c).

The level of current fertility (i.e., the TFR) is first adjusted by the factor F_t for the accounting effects of shifting births toward younger or older ages at childbearing (i.e., tempo), as suggested by Bongaarts & Feeney (1998) and discussed above. In recent decades, pervasive postponement implies F_t values well below 1.0 (i.e., F_t < 1.0). Thus, TFR/F_t equals the quantum of period fertility (i.e., period fertility corrected for the effects of shifts in fertility timing).

Key to explaining TFR/F_t (quantum) is intended parity (IP, the number of births intended) of young women (e.g., those aged 21–25) increased or decreased by a set of model parameters that reflects forces not incorporated into their reports of childbearing intentions. These factors are not incorporated into stated intentions because they cannot be well anticipated. Illustrative of factors that increase fertility relative to earlier stated intentions are:

• F_u: Unwanted fertility. Unwanted births measured through women's reports (at the time they become pregnant) that they had intended not to have any more children. Thus, these pregnancies (and births) would not have occurred in a “perfect contraceptive” society. Unwanted fertility increases TFR relative to IP (and F_u > 1.0).

• F_g: Gender preferences. Some couples who intend, say, two children, will have strong preferences for at least one son and one daughter (see Pollard & Morgan 2002). But given that sex of children is not easily controlled or anticipated, roughly half of such couples will reach the two-child goal but not have an acceptable gender composition. These couples are likely to revise their IP upwards. Such preferences and behavior would increase TFR relative to IP (and F_g > 1.0).⁵

Illustrative of factors that attenuate the impact of intended parity on observed fertility are:

• F_i: Impaired fecundity. In general, women (and their partners) will not know if they are, or when they will become, sub- or infecund. Thus, women cannot factor impaired fecundity into their IP reports. A few women are infecund at young ages (1% or 2% at ages 15–19; see Bongaarts & Potter 1983), but the proportion increases with age (especially after age 35; see Bongaarts & Potter 1983, Menken 1985) and with the prevalence of some diseases/infections. Male fecundity declines with age also, but its onset is much later than for females and its pattern of decline less certain. Clearly, impaired couple fecundity decreases TFR relative to IP (and F_i < 1.0).

• F_c: Competition. Women/couples can also revise their IP upwards or downwards depending on their experiences, opportunities, and constraints (that encourage or compete with childbearing). Following Bongaarts, we refer to this broad class of constraints and competing opportunities as competition. Some of this competition may be anticipated and incorporated into IP. What is not, e.g., the unanticipated difficulty of combining career and family responsibilities, of finding a suitable partner, etc., is reflected in this parameter. In contemporary settings, F_c is expected to be < 1.0.

In this framework, observed fertility reflects the balance of these forces. Morgan and colleagues (Hagewen & Morgan 2005, Morgan 2003, Morgan & Hagewen 2005, Quesnel-Vallée & Morgan 2003) have applied this framework in several different contexts as a conceptual and as an analytic model.

What We Know

The elaborate description captured in the decompositions and frameworks above outlines what we know. We extract from these frameworks a set of low fertility axioms.⁶

1. Fertility postponement lowers current period measures of fertility relative to cohort measures or to period measures not adjusted for these timing shifts. Thus, timing shifts, which are time-limited phenomena,⁷ are a significant part of contemporary low fertility.

2. Fertility timing (the age at which women have children) may have consequences for the number of children that individual women will bear. Specifically, in the aggregate, fertility postponed implies some fertility foregone.

3. Women's (men's and couple's) fertility histories unfold in tandem and in interaction with human capital formation, mental and physical health trajectories, and other key aspects of their life course and that of their partners. The life course perspective is the unquestioned, appropriate analytic frame for contextualizing fertility intentions and behavior.

4. Parents incur high direct and indirect costs in having and rearing children in most contemporary contexts. Indirect costs are substantial and reflect primarily the mother's foregone earnings due to pregnancy, childbirth, and childrearing (see DiPrete et al. 2003). Direct costs are also substantial and are more easily and regularly calculated (see Lino 2004).

5. Dominant norms and cultural schemas legitimate active birth control (i.e., it is widely accepted that the quantum and timing of children impact individual and family well-being and that strategizing and acting on this belief is legitimate).

6. Low parity births (especially the first and having a sibling for the first child) remain strongly normative and fulfill women's/couples' desire to be parents and have a family. Ideal and intended family sizes of young women are usually two, with three being the second most popular choice. Childlessness and one-child families are not seen as ideal and are not commonly intended (see Bongaarts 2001, Hagewen & Morgan 2005).⁸

7. High parity births are increasingly rare and non-normative; given the other axioms discussed and especially the high costs of children (see axiom 4 above), high parity births may be viewed as disadvantageous for parents, siblings, and society more generally.

Fundamental Causes of Low Fertility and Its Variations

Low fertility frameworks must also allow for more fundamental (or distal) causes; the axioms above are proximate and beg more fundamental questions such as, for example, why are monetary and opportunity costs of children very high? At this level of abstraction, explanations for low fertility are varied, and the number of theoretical contributions is large. Moreover, explanations for low fertility are not fully divorced from the empirical and theoretical work on fertility declines from high levels or “fertility transitions” (see Bumpass 1990, Hirschman 1994, Mason 1997). To situate these varied contributions, we offer a two-dimensional conceptual space created by theoretical dimensions of scope and content. Scope refers to the geographical and temporal applicability of the theory, and content refers to the primary putative causal factors. Table 2 displays these dimensions and locates various contributions in this space.

TABLE 2.

Typology of low fertility theories/schemas with illustrative examples

	Scope
	Global	Interactive	Idiosyncratic
Content
Economic change	Notestein 1945, 1953 Davis 1937 Caldwell & Schindlmayr 2003	Caldwell 2001	Sobotka et al. 2003 Kreyenfeld 2003 Witte & Wagner 1995
Ideological change	Lesthaeghe & van de Kaa 1986 Thornton 2005
Institutional		Esping-Andersen 1999 McDonald 2000 Rindfuss et al. 2003	Morgan 2003
Technological change	Potts 1997	Galor & Weil 1996, 2000	Goldin & Katz 2000
Synthetic/path dependent	Goode 1963 Bumpass 1990 Mason 1997		Ginsborg 2003 Atoh 2001 Prachuabmoh & Mithranon 2003

The Conceptual Range of Theories

Before discussing particular contributions, let us justify the importance of these two orienting dimensions. First, the horizontal dimension—scope—captures a theory's reach across time and space (see Table 2). At one extreme we have global theory, presumably relevant to all contemporary countries/groups and one useful for anticipating future change. As one moves from left to right, one shifts from grand global theories to interactive ones, and finally to idiosyncratic ones. The distinction between interactive and idiosyncratic is important. Consistent with its meanings in quantitative statistical frameworks, an interactive model implies that the effect of a specific change is dependent on the level of a second (third, and/or fourth) variable. Interactive explanations can still be highly structured with broad scope. But some scope conditions are incorporated, by definition, as part of these models. Idiosyncratic explanations are historical in nature and stress the uniqueness or path dependence of each case. As one moves from left to right, the models lose parsimony but typically gain substantive plausibility and improved fit to key observations.

The vertical dimension—content—does not represent a continuum but rather a set of categories of putative causes of fertility change. The last category, synthetic/path-dependent explanations, represents contributions that stress the importance of multiple factors. The importance of multiple factors does not necessarily imply an interactive model because the multiple factors can have additive effects. Because no contribution we discuss denies the existence of multiple causes, row entries in Table 2 represent the primary putative causal factor lying at the heart of the scholar's argument.

Contributions Stressing Economic Change

The classic work on demographic transition theory and its antecedents falls in the upper left (NW) of this conceptual space (see Notestein 1945, 1953). This theory has grand scope and views economic development as the driving causal force. Specifically, demographic transition theory posits that economic development will lower mortality; in turn, both of these factors will lead to a subsequent fertility decline. Eventually, fertility levels will match low mortality, but the homeostatic mechanisms underlying this balance are not specified.

We position the classic work of Davis (1937, 1997) in this area of Table 2. As in demographic transition theory, Davis argues that economic development (i.e., industrialization) drove a process of social change that undermined incentives for childbearing (see also Thompson 1929). Correlated institutional changes served as proximate causes, but Davis further argues that “the declining birth rate has resulted from a ripening incongruity between our reproductive system (the family) and the rest of modern social organization.” This incongruity was not transitional (or the result of a time lag) but was fundamental to modern societies. The modern urban-industrial-mobile society was essentially at odds with the social organization characterized by a kin-based, familistic society. The new fertility regime was one of small families (also see Livi-Bacci 1999). The discrepancy was at least partially the result of the changing roles, expectations, and opportunities of women as economic development occurred (Davis & van den Oever 1982). This theory is grand in scope; it predicts global fertility decline with continued economic globalization and low fertility as a looming crisis for all economically advanced countries.

In addition, the rationales to have higher parity births were increasingly and specifically undermined by economic development. Value-of-children studies suggest that key rationales for childbearing (becoming a parent, having a child to love, carrying on the family name, etc.) were satisfied by a first and second child. Higher parity births had rationales (e.g., help around the house or with chores, look after younger children, etc.) that were quickly undermined by economic development (see Bulatao 1981). In some contexts the direct costs of education were salient, but indirect costs were likely more consequential (see Becker 1981, Becker & Lewis 1973, Caldwell 1982). The increase in these costs does not seem to be a fact in dispute and must be an antinatalist force (see Blau & Robins 1989, Mason & Kuhlthau 1992).

Caldwell & Schindlmayr's (2003) recent, provocative review shares this conceptual space; they cite Davis approvingly at several points. Caldwell's earlier work (1982) notes a set of achievement ladders constructed by industrial society, placing demands specifically on women and extending education for both sexes and for children. Caldwell & Schindlmayr (2003) are impressed with the global spread of low fertility, e.g., its emergence in a broad range of cultural contexts. They argue for a theory with a grand scope to account for the emerging global pattern of low fertility. Economic development and globalization, and accompanying greater wealth and consumerism, fit their search requirements for a fundamental, powerful, and pervasive causal force.

Adjacent to this space and to the right, we place a Caldwell (2001) elaboration. Specifically, in attempting to understand the nearly simultaneous decline of fertility in both developed and developing countries in the 1960s and 1970s, Caldwell (2001) continues to stress economic development and increasing incomes as the ultimate cause of low fertility. But a more compelling explanation of the simultaneous timing required acknowledgment of new birth control technologies (e.g., the contraceptive pill) and new rationales for controlling fertility (e.g., concern about rapid population growth). These new technologies and new rationales provided an environment that accentuated and hastened fertility decline.

Next, further right (NE corner) are idiosyncratic explanations that focus on markets/economic change as a driving force for fertility declines in specific nations/regions. Notable among these idiosyncratic explanations is the contrast of East and West Germany postunification (Kreyenfeld 2003) and whether fertility decline in East Germany was fueled mainly by the economic crisis or whether it was more of an adaptation in evolving toward West Germany (Conrad et al. 1996, Eberstadt 1994, Witte & Wagner 1995). According to the crisis argument, the upheavals caused by the fall of communism had social consequences (Caldwell 2004), including an insecurity about the future or personal disorientation in addition to economic hardship (Philipov 2002). Thus, the fundamental causes are overarching, but nations sharing these experiences (e.g., nations in Eastern Europe) may fare uniquely on fertility outcomes after such upheavals owing to their unique political and social histories (see Sobotka et al. 2003). Although the crisis explanation is used for many countries undergoing the shift from communism, it has not been substantiated at the microlevel in Russia (Kohler & Kohler 2002). In addition, the argument that upheaval and uncertainty promote fertility has also been forwarded (see Friedman et al. 1994).

Contributions Stressing the Import of Ideological Change

In the next row of Table 2, we place the arguments of scholars who stress ideological changes, as opposed to objective, economic ones. As Mason (1997, p. 450) argues, “theories of fertility change must recognize that changing perceptions ultimately drive fertility change, and that perceptions may change more slowly or more quickly than the reality with which they are concerned.” A useful way to conceptualize changing perceptions or ideological change is as a package, an integrated set of ideas that provides an orienting and motivating schema for life. Behavioral change, including fertility change, results from the adoption of new cultural schema that interpret contemporary contexts in ways that produce low fertility. One can conceive of rational decisions once this interpretative schema has been set (see Hammel 1990).

The work of van de Kaa and Lesthaeghe is illustrative. Their causal interpretation of low fertility (Lesthaeghe & van de Kaa 1986, van de Kaa 1987) hinges on a cultural shift in the dominant mental/cultural schema (see also Lesthaeghe & Willems 1999). Specifically, following Ariès (1962, 1980), van de Kaa and Lesthaeghe (see van de Kaa 2003, p. 78) argue that there are two successive motivations for declining fertility. The first, associated with the (first) demographic transition and especially with declining family size, assumes that all who could have children would bear them, and that parents' dominant orienting goals were to provide substantial resources to their children. Van de Kaa (2003, p. 78) says that altruism toward children defines this schema, one that Ariès claimed produced an enormous sentimental and financial investment in children. In contrast, low fertility and associated demographic change (i.e., the second demographic transition, including the emergence of sustained low fertility and associated demographic change such as increased rates of divorce and cohabitation) are motivated by new ideas that place the individual and individual choice at the core of the unfolding life course; the contemporary challenge, they argue, is for individuals to construct a meaningful life in the absence of a clear normative life course (one not necessarily including parenthood).

This new schema is consistent with Giddens's (1991) description of the deinstitutionalized modern life course. The choices to have children and when become ones that women/couples make as they construct stimulating and meaningful lives. Caring for children remains important but in a context in which the decision to have children is optional and parenthood should contribute to individual self-actualization. Note that this reflexive or postmodern cultural schema is consistent with the behavioral components of the second demographic transition: fertility delay, decisions to forego childbearing, cohabitation, union instability, etc. This emphasis on changed cultural schema does not deny structural changes, including economic globalization. But the fundamental change, according to van de Kaa and Lesthaeghe, was a shift in the dominant interpretative frame through which these changes were viewed.

Van de Kaa's and Lesthaeghe's argument was developed to account for pervasive changes that occurred in the West, so the initial implied scope of their argument is limited. However, low fertility is not just a Western phenomenon, as we have noted. Thornton (2005) argues that a powerful cultural frame or schema, developmental idealism, conflates Western wealth and power with Western family forms and Western cultural forms. Thus, these ideas have power by association, and Western conceptions of the family follow or precede economic globalization. These sets of ideas can have powerful effects on individuals. Combining the arguments of van de Kaa, Lesthaeghe, and Thornton, one can imagine shifts in cultural schema that could help explain patterns of low fertility on a global scale.

Contributions Stressing Institutional Change/Differences

Moving downward again we reach a category, institutional influences, that emphasizes structural aspects of society that can constrain and channel behavior. This conceptual contrast with cultural frames/schema (discussed above) is, of course, only an abstraction. Culture and structure continually interact; in fact, one might define institutions as embodied culture. Nevertheless what we discuss here is frequently called structure and includes gender, family structure, child care provision, consumer markets, government policies, etc.

Gender has attracted much attention in the low fertility literature. McDonald (2000) explains how gender equity can be antinatalist in high fertility contexts but pronatalist in low fertility settings. This question arises because of the anomalous positive association appearing in the 1990s between aggregate fertility levels and levels of female labor force participation. This reversal of an assumed, structural antinatal relation between women's work and fertility elicited great interest (Ahn & Mira 2002, Billari & Kohler 2004, Brewster & Rindfuss 2000, Rindfuss et al. 2003). McDonald (2000) argues that the transition from high fertility to fertility around replacement level is accompanied by, and encouraged by, increasing gender equity within the family. Gender equity promotes lower fertility by increasing the likelihood that women's fertility intentions will be consequential and by increasing alternative avenues to satisfaction, status, and prestige. Thus, McDonald's arguments are consistent with those of social demographers (Calhoun & Espenshade 1988, DiPrete et al. 2003, Kravdal 1992, Rindfuss&Brewster 1996) and economists (see Easterlin 1980, 1987; Engelhardt & Prskawetz 2002; Willis 1973) that gender change increases the opportunity costs of mothers' foregone labor market opportunities. Empirical evidence on this point is substantial and supportive (Mincer 1985, Weller 1977, Adsera 2004).

Once fertility is low, gender equity may rise further in individual-oriented institutions (e.g., education, the economy, politics) while remaining relatively low in family-oriented institutions. As a result, fertility can fall to very low levels, e.g., among societies usually characterized as more patriarchal (e.g., Italy, Spain, and Japan) but remain closer to replacement levels in more egalitarian contexts (e.g., Scandinavia and the United States) (see Gauthier & Hatzius 1997, Rindfuss et al. 2003). Fertility closer to replacement level (with pervasive gender equity) occurs because women bear fewer of the time, energy, and financial costs of rearing children (greater equity has spread these costs more broadly than in settings with stronger familial patriarchy). Note that McDonald's theory of gender and fertility change is clearly interactive: Very high gender equity in nonfamilial institutions produces low fertility only when gender equity in familial institutions lag behind.

Esping-Anderson (1999) offers another important institutional and interactive theory focused on Western welfare state regimes. Esping-Anderson identifies institutional clusters that combine different types of labor markets, the state, and the family. Thus, the fertility rate under a social democratic regime (e.g., Norway) may vary from a conservative regime (e.g., Italy) owing to the balance of the family, state, and market in managing social risks.

The relationship between defamilialization (or a state's willingness to absorb the responsibilities traditionally relegated to the family) and fertility is increasingly positive and supported by the now positive relationship between female employment and fertility (see Brewster & Rindfuss 2000). Low fertility could become a “low fertility equilibrium trap that springs from the incapacity of women to harmonize careers and family obligations” (Esping-Anderson 1999, p. 5). However, other regimes constitute contexts in which near-replacement fertility remain (e.g., a liberal regime such as the United States)

The McDonald and Esping-Anderson genre of explanation has been commonly used; that is, the origin of fertility variation and change is often sought in institutional variation and change (see Ryder 1980). For instance, Rindfuss et al. (2003; also see Morgan 2003) stress institutional variation that impacts (a) availability, acceptability, accessibility, quality, and cost of childcare; (b) market substitutes for goods/services formerly produced in the home; (c) labor market accommodations (e.g., flex time); (d) public policy interventions (e.g., family leave); and (e) gender role flexibility and men's contributions to housework and childcare.

Contributions Stressing Technological Change

New technologies are not exogenous. They emerge within contexts that allow or encourage their development and use, or they diffuse or are imported from other settings. But their precise use, their meaning, and the extent of their diffusion depend on cultural understandings and existing social institutions of the receiving population. Thus, although we acknowledge the fuzziness of this conceptual category (technology that is separate from culture and institutions), it identifies a class of arguments in which technology plays a key role (see Table 2). More specifically, fertility and technological change are intertwined on several levels: contraception/assisted reproduction, household technologies, labor market and production technologies that make work more women-friendly, and increased “child quality” owing to health and education inputs increasingly coming from nonfamily institutions (Becker 1981, chapter 11).

A key area of technological change involves contraceptive technology and infertility treatments. These technologies play a secondary role in many explanations of fertility change. But they can be seen as playing a fundamental role. An extreme example is Potts's (1997, p. 5) sweeping evolutionary argument: He claims that humans have a set of biological predispositions that, in our hunting and gathering past, would have jointly produced high reproductive fitness (i.e., above replacement-level fertility). He claims that humans are genetically predisposed to seek sexual relations, to love and support our own children once they are born, and to be socially and sexually competitive. In the absence of contraception and abortion, these predispositions assure that babies are born, that they are cared for after birth, and that the most able (in terms of accumulating wealth and status) will have the most surviving offspring. In contrast, in contemporary contexts the link between sex and reproduction can be severed by contraception and abortion, which are key technological/social innovations. Potts argues that evolution has provided us with no gene for a certain number of children. Once the link between sex and reproduction is broken, what will provide the motivation for children? From this perspective, effective contraception plays a key role in fertility decline and sub-replacement fertility.

Goldin & Katz (2000) make a very different argument about contraception and, specifically, the oral contraceptive, or the pill. Theirs is an idiosyncratic argument focusing on the United States; they present time series data consistent with claims that the pill altered women's career decisions through both direct and indirect routes. The direct route was the pill's increased reliability and ease of use compared with other methods. Thus, the pill allowed a larger group of women to invest in expensive, long-duration training without paying as high a price in terms of abstinence or postponement of unions as women did before the pill. The indirect effect is the one mentioned above, termed by Kohler social multiplier effects (see Kohler 2001, Kohler et al. 2002). Goldin & Katz conclude that the pill had a large effect on career and marriage (and by inference fertility timing), even though isolating this effect with great certainty is not possible. (A weakness of this argument is the low fertility in some countries where the pill is far less important, e.g., Japan, Italy.)

On technological change more generally, a number of arguments stress the role of technology in interaction with economic development and increased productivity (e.g., Galor & Weil 1996, 2000). Specifically Galor & Weil (1996) argue that technology (increased capital per worker) raised women's wages relative to men's. These higher wages reduced fertility owing to the higher opportunity costs of childbearing. This lower fertility, in turn, led to greater investments in technology (i.e., increases in level of capital per worker).

Theories Stressing Multiple Domains and/or Their Interactions

Another important conceptual space is the bottom left (SW corner) (see Table 2). This space reflects global and synthetic explanations of social change, including low fertility. It differs from those above in that the putative causal forces are not derivative solely of economic, ideological, institutional, or technological change. Rather, it is the combination of these factors. Goode's (1963) classic, World Revolution and Family Patterns, defines this conceptual space. Goode argues that industrialization, family change, and ideological change are sets of mutually reinforcing factors that were sweeping the globe because of their joint attractiveness: industrialization because of the greater wealth and standard of living it brought; Western ideology because of the widespread desire for greater freedom and choice in life decisions; and the Western family because of its free mate choice, companionate marriage, and low fertility. Bumpass (1990) supported this approach to theorizing about family change in his presidential address to the Population Association of America, as did Mason (1997) in her presidential address focusing on the demographic transition.

Path Dependence and Idiosyncratic Explanations

At the bottom right (SE corner) of Table 2, we place explanations that stress the unique details of particular contexts and their unique intertwining. For instance, the historian Ginsborg (2003, p. 74) explains the very low Italian fertility as the distinctive “intertwining of the old and the new in family strategies.” More specifically, Ginsborg does not dispute some transnational influences. “[T]here were strong forces pushing toward a European model of modernity, among which were the cultural revolution of the 1960s and 1970s, leading to greater individual choice, the spread of contraception and legalized abortion, the partial emancipation of women and their entry in to the labor market.” But Ginsborg also stressed Italian-specific factors that combined with these transnational ones in a distinctive intertwining. He writes, “tradition weighed heavily in both the public and the private spheres; the felt obligation to have children within marriage, the power of the family as an intergenerational collective, the state's disinterest, after the unfortunate Fascist experience, in reproduction politics” (Ginsborg 2003, p. 74). Although we have used the work of the historian Ginsborg as an illustration, many demographers and sociologists are receptive to these kinds of arguments. Hobcraft (see Billari et al. 2004, p. 81), in responding to Caldwell & Schindlmayr (2003), writes, “I have no difficulty with accepting differing explanations for different regions and time periods, though these explanations need to be drawn from a common overarching framework.” Consistent with Hobcraft's view, Morgan (2003) argues that the exceptionalism of American replacement-level fertility is best accounted for by a host of distinguishing factors; he illustrates this point by contrasting Italy and the United States, where a set of modest differences (some idiosyncratic) adds up to a substantial overall difference.

Recent study of East and Southeast Asia highlights the path-dependent nature of fertility transitions within the regions and their differing causes (Caldwell & Caldwell 2003, Gubhaju & Moriki-Durand 2003). The experience of low fertility in Japan and its causes (Atoh 2001) are distinctly different from those in China (Yi 1996). Countries such as Thailand (Prachuabmoh & Mithranon 2003), Singapore (Yap 2003), and Korea (see Cho 2000) are currently discussed in terms of their specific contexts more often than situated in a broader global theory (see Atoh et al. 2004).

Conclusion: Where do we go from here?

Reviewers often conclude by stating that we need to know much more and that huge gaps exist in our understanding. We begin by stressing the opposite point: We know a great deal! We agree with Mason (1997, p. 452) that fertility researchers are inhibited more by “erroneous thinking than by any fundamental lack of knowledge.” We challenge claims that fertility researchers “really don't know why fertility fell from high levels (6–8 children per woman) to two children” or that we “really don't understand the postwar baby boom and bust” or that “we don't really know why very low fertility exists in some societies and why fertility levels approximate replacement elsewhere.” In fact, at certain levels of abstraction (at the level of decomposition and the proximate determinants), we have frameworks to answer all these questions, and we can provide the specific details that make the explanations compelling (e.g., see Hobcraft & Kiernan 1995, pp. 23–27; Cherlin 1992, pp. 31–65). Much of what critics mean when they say we don't understand is probably unknowable—both now and in the future—because the answers they seek imply a level of determinism that does not exist.

To explain, we know a great deal about fertility change and variation. Consider the axiomatic claims listed above. These claims are embodied in at least one of the frameworks discussed in the section on Decomposition, Proximate Determinants, and What We Know About Low Fertility. These statements and models reflect one level at which we can know things. The models provide general frameworks within which we can construct social histories of particular cases. These frameworks can accommodate the more distal causes stressed by various authors and shown in Table 2. Note that interactive effects and case-specific idiosyncrasies do not render general claims useless or false. Rather, they illustrate that our models are simpler than reality itself and that our models capture what is systematic and known about/across cases. As Lieberson & Lynn (2002, p. 4) note “incomplete is not the same as erroneous.” Models are useful to the extent that they capture key aspects of the cross-case variation we seek to explain.

If these proximate/descriptive frameworks had been available in 1945, they would not have predicted the baby boom or bust. Further, these models will not predict well future fertility levels (on a decadal timescale). Critics will say the poor predictive capacity of these frameworks is due to their omission of true, fundamental, or distal causes of social change or that the more proximate variables discussed above are not linked in consistent ways to these fundamental and distal causes. Both criticisms are true. But we argue that such precise, mechanistic theories of social change should not be sought (see Lieberson & Lynn 2002, p. 10). Such attempts assume powerful, persistent, and pervasive drivers of social change, forces that make irrelevant existing cultures and institutions, a level of determinism that does not exist.

Thus, our final question is, What do we need to know or do? Each of the frameworks discussed previously imply important basic science research agendas. Decomposition analyses must update our understanding of current trends and differences so that our social histories focus on the dynamic demographic components of change. To understand the transition to parenthood in particular times and places, the general perspectives outlined above need to be applied with a thorough understanding of the particular case. We need more sophisticated modeling and detailed longitudinal data to understand how life course development occurs for particular individuals in different contexts. We need to map the ever-changing cultural and structural landscape and how persons understand it if we are to craft compelling social histories of changing family and fertility behavior.

We especially recommend the conceptual framework of Bongaarts (2001, 2002) as a frame for organizing subsequent low fertility work; it suggests a set of important unfinished research agendas. At the heart of the model is the intended parity of young women, i.e., how many children do women/couples want? All other parameters reflect factors that impinge on translating these intentions into behavior. A first step is to further explore fertility intentions and the individual experience and normative and structural constraints that they reflect. Are these intentions general normative statements, or do they reflect more concrete plans for the individual? How are fertility intentions weighed vis-à-vis other goals as the life course unfolds? Are fertility intentions primarily about the timing of parenthood and subsequent births? Or, are number-of-children concerns paramount?

The next question implicit in the Bongaarts framework is: Why don't women/couples realize intentions? This is key because all current data suggest that below-replacement fertility reflects a shortfall of births relative to intentions, i.e., fertility intentions are significantly higher than are current TFRs (see Hagewen & Morgan 2005). Rather than reflecting any inherent problem with stated intentions, this intent-behavior disjuncture may reflect the agency, constraints, and happenstance that define the individual life course. So a second agenda item is to understand better how, for instance, individuals' family trajectories are interrelated with human capital formation, health trajectories, or other key life cycle domains across contexts.

Third, the Bongaarts framework decomposes the intent-behavior disjuncture into that due to unwanted fertility, timing shifts, infecundity, competition with other goals, etc. The import of these constraints will vary across contexts, providing important questions for cultural and structural analysis. Why, for instance, is unwanted fertility much higher in the United States than in most other low fertility contexts? The residual competition parameter will vary in magnitude across populations: Does this variability reflect availability of partners/spouses or variation in work-family conflict traceable to various institutional structures and interactions?

For the more distal causes (featured in Table 2), new integrative work is needed. We need models of change that are conciliant with what we know from other disciplines—from cognitive and brain sciences, social anthropology, history, economics, psychology, and political science. We want a model that corresponds to the (biological, psychological, social, and cultural) processes at work in the world. These models will highlight the range of possible outcomes and stress the unpredictability of social change. In this regard, linking biology, genetics, life course development, and social context stands as a major challenge to a fuller understanding of fertility change and variation, and more generally, family change and variation. On this score, our knowledge is rudimentary; the most promising sign is the increasing willingness of some scientists to take up the agenda. Cross-species, cross-cultural, and historical observations provide some relevant data (e.g., Hrdy 1999). Measurement of genetic markers and heritability studies provide additional clues (see National Research Council 2003) for the needed theoretical development. But the greatest need is for a fuller understanding of the dynamics of cultural and institutional change, an effort requiring tools and concepts often associated with cultural anthropology and history. We must move beyond debates of economic change versus ideology or structure versus culture to new formulations that do justice to the dynamics of social change (e.g., Sewell 1992, 2005).⁹

In addition, the role of policy requires further attention. Although noted as a substantial factor in the work of Esping-Anderson and others (see McDonald 2002), the role of policy as a fundamental cause of fertility variation remains uncertain while variation in the policy histories of low fertility regimes is great (see UN Popul. Div. 2004b). If low fertility persists and spreads, the interest in effective pronatalist policies will intensify and proliferate. Thus, an increased knowledge base on effective policies and attention to the ethical and practical implications of policy action/inaction are needed.