Abstract

Dynamic theories of family size preferences posit that they are not a fixed and stable goal but rather are akin to a moving target that changes within individuals over time. Nonetheless, in high-fertility contexts, changes in family size preferences tend to be attributed to low construct validity and measurement error instead of genuine revisions in preferences. To address the appropriateness of this incongruity, the present study examines evidence for the sequential model of fertility among a sample of young Malawian women living in a context of transitioning fertility. Using eight waves of closely spaced data and fixed-effects models, we find that these women frequently change their reported family size preferences and that these changes are often associated with changes in their relationship and reproductive circumstances. The predictability of change gives credence to the argument that ideal family size is a meaningful construct, even in this higher-fertility setting. Changes are not equally predictable across all women, however, and gamma regression results demonstrate that women for whom reproduction is a more distant goal change their fertility preferences in less-predictable ways.

Introduction

Demographers have historically studied the trends in and predictors of population-level family size preferences. There has been comparatively less research into the determinants of individual-level preferences. Although family size preferences are powerfully shaped by societal norms (Ryder 1976/2010), considerable heterogeneity remains within communities and, importantly, within individuals themselves over their reproductive lives. Through understanding how individual-level preferences evolve dynamically over time, demographers can better appreciate the meaning of such preferences and their life course determinants.

Most studies of the determinants of family size preferences assume either explicitly or implicitly that they are relatively stable over time. Conceptually, though, it seems more likely that family size preferences develop over time as one’s life circumstances unfold in often unanticipated ways. A small number of studies have demonstrated relationships between individuals’ family size preferences and sociodemographic characteristics, attitudes, and values (Bankole 1995; Ezeh et al. 1996; Hayford and Morgan 2008; McCarthy and Oni 1987). The largely cross-sectional nature of most studies, however, stands in contrast to the potentially dynamic quality of preferences. Although men and women may develop family size preferences at young ages, these desires are mutable. Indeed, as early as the 1960s, studies documented significant changes in family size preferences and expectations over time (Freedman et al. 1965). A recent proliferation of longitudinal data has inspired a new generation of research, which takes seriously the claim that changes in family size preferences are legitimate. Many of these studies have investigated how individuals’ fertility preferences change over time and the predictors of change in low-fertility contexts (e.g., Hayford 2009; Heiland et al. 2008; Iacovou and Tavares 2011; Liefbroer 2009). In higher-fertility contexts, however, there is a history of treating such change as evidence of poor reliability and validity of the construct of an ideal family size (Bankole and Singh 1998; Bankole and Westoff 1998).

In this article, we examine the dynamic nature of family size preferences among young women in southern Malawi, a high-fertility context where desired family size has been declining slowly for two decades. We posit that the process of fertility transition—whereby having fewer children becomes an increasingly acceptable family formation strategy, and eventually a societal norm—implicitly allows women new latitude in developing and altering their family size preferences. In doing so, we take seriously the possibility that changes in preferences are real and not simply data artifacts. We begin by describing the prevalence of change in ideal family size reports across eight waves of data, each spaced four months apart. We then test a series of hypotheses derived from variants of the sequential decision model of fertility to answer whether—and for which women—changes in ideal family size are empirically predictable in this setting.

Background

Preferences, Intentions, and Expectations

There is a lack of consensus about the best way to measure reproductive goals and a variety of concepts are used: most notably, preferences, intentions, and expectations. Fertility preferences are feelings or desires related to having children, whereas fertility intentions involve planning or action, such as actively trying to conceive (Thomson 1997). Different still, fertility expectations are realistic projections of future fertility that incorporate desires for children, beliefs about fecundity, and access to contraception (Ryder and Westoff 1967). Despite their conceptual differences, each of these concepts is useful for predicting reproductive behaviors and completed fertility (Freedman et al. 1980; Thomson 1997; Westoff and Ryder 1977). In the present study, we are interested in the ways family size preferences are influenced by changing life circumstances during the transition to adulthood. However, because much of the literature conflates these concepts, and because each is useful for understanding fertility goals, we draw on literature that investigates preferences, intentions, and expectations in the subsequent review.

Dynamic Models of Fertility

Early microeconomic studies of fertility were structured around a static, “one-decision” model that assumed family size preferences were formed early in life and stayed relatively fixed from marriage onward (Ryder 1973; Udry 1983:117). Decades ago, a number of demographers challenged the assumption of an a priori set of fertility preferences and proposed more dynamic models of reproduction. They argued that childbearing decisions were more likely to be made sequentially, child by child, on the basis of factors such as childrearing experience, the sex composition of children, or other context-specific determinants (Bulatao 1981; Hout 1978; Namboodiri 1972; Udry 1983; Yamaguchi and Ferguson 1995). Whereas the theory initially described decision-making that drove fertility behaviors, the theory is equally applicable to sequential evolutions in reproductive goals. Indeed, Lee (1980) described reproductive goals as akin to a “moving-target” that changed over time. Although the vast majority of demographic scholars would agree with some variant of the sequential decisions thesis, a number of our practices—such as the measurement of unwanted fertility—still rely on the assumption that preferences are relatively stable (Casterline and el-Zeini 2007). For example, one concern when measuring unwanted fertility is that a birth that has already occurred but may not have been desired at the time of conception may be subject to ex-post rationalization of it as a wanted birth. Indeed, the wantedness of a child can change even over the course of a pregnancy (Joyce et al. 2002). Not all changes in fertility preferences, however, reflect ex-post rationalizations of past behavior: they can also reflect evolving life circumstances and values (Bulatao 1981; Lee 1980).

Studies of Change

Recently, the availability of high-quality longitudinal data has led to a resurgence of interest in how and why people revise their family size preferences over time. Studies from Western Europe and the United States have found that people change their fertility preferences over their reproductive life course and that these changes are frequently patterned and predictable (Hayford 2009; Heiland et al. 2008; Iacovou and Tavares 2011; Liefbroer 2009; Rocca et al. 2010). Longitudinal studies of change in fertility preferences have also been conducted in higher-fertility contexts in Africa. Two studies, one from Ghana and one from Morocco, found that approximately two-thirds of women provided different responses to ideal family size questions over their respective two- and three-year reference periods (Bankole and Westoff 1998; Debpuur and Bawah 2002). Other studies from Africa have found that changes in the desire for more children and the desired timing of children are common (Kodzi et al. 2010; Sennott and Yeatman 2012).

A recurring question in research on family size preferences in developing contexts is what survey reports of preferences really mean (Agadjanian 2005; Bankole and Westoff 1998; Debpuur and Bawah 2002; Johnson-Hanks 2007; Knodel and Prachuabmoh 1973). Most studies from sub-Saharan Africa treat changes in family size preferences as evidence of either poor construct validity or poor reliability of the measure (Bankole and Singh 1998), rather than an indication of genuinely evolving desires. This approach stands in stark contrast to that adopted by studies from Europe and the United States that are more inclined to interpret changes in fertility preferences as meaningful (e.g., Hayford 2009; Heiland et al. 2008; Iacovou and Tavares 2011; Liefbroer 2009).

Fertility Transition and the Context of Malawi

Some researchers have argued that the concept of a particular family size target does not translate well to the way many African cultures think about fertility. First, some have questioned the ability of less-educated women in high-fertility contexts to reliably report a consistent family size over time or to even think numerically about their family size (Hauser 1967; van de Walle 1992). Although this may have been the case in pretransition fertility contexts, in 2010, only 2 % of women in Malawi (where our data are from) gave a nonnumeric response to the Demographic and Health Survey (DHS) question on ideal family size, demonstrating that Malawian women do think numerically about their family size desires (Measure DHS and ICF Macro 2011). Second, others have argued persuasively that African reproductive regimes are less focused on absolute numbers of children than on adequate spacing and maintaining the possibility of future reproduction (Bledsoe et al. 1998; Caldwell and Caldwell 1987; Johnson-Hanks 2007). However, in an environment of declining fertility, where norms around family size are changing, women may become more committed to limiting their family size and more focused on numbers of children, even as these numbers evolve over their reproductive years. Indeed, in Malawi, the mean age at which the demand for limiting births exceeds the demand for spacing births is 29 (Van Lith et al. forthcoming). Malawian women’s interest in limiting their absolute number of children is additionally evidenced by the fact that more than 25 % of women older than 40 have had a tubal ligation, up from just 3 % in 1992 (Measure DHS and ICF Macro 1994, 2011).

Malawi is slowly embarking on a fertility transition: the total fertility rate declined from 6.7 in 1992 to 5.7 in 2010 and is as low as 4.0 in urban settings. During the same 1992–2010 period, ideal family size fell from 5.1 to 4.0 (Measure DHS and ICF Macro 1994, 2011). Most of the change was due to the aging-out of older women with high family size ideals and not change within cohort (see Fig. 1 in the  appendix).1 In 2010, ideal family size was as low as 3.2 among 15- to 19-year-olds (compared with 5.3 among women aged 45–49). As fertility declines, having fewer children becomes an increasingly acceptable family formation strategy—and eventually, a societal norm, which may allow individuals (and perhaps women in particular) more latitude in developing and altering their family size preferences. Thus, young women in transitioning contexts who are planning to have children may be more willing and able to state lower preferences and to subsequently adjust them because there are less-rigid normative expectations about completed family size as fertility declines (Agadjanian 2005).

Our analyses focus on women at the beginning of their reproductive careers because fertility preferences are likely to be particularly flexible during this period (Iacovou and Tavares 2011). Additionally, it allows us to move beyond the ex-post rationalization of births as an explanation for changes in preferences and to examine more broadly how relationship and reproductive events influence ideal family size over time. Such events mark salient moments in the transition to adulthood for our young sample and have been shown to be particularly meaningful in other settings (Iacovou and Tavares 2011; Liefbroer 2009). They also provide a narrow lens through which to assess the applicability of the sequential model of fertility preferences to the Malawian context. In Malawi, a tremendous amount of change happens in a woman’s life between the ages of 15 and 25. At age 15, most Malawian women are unmarried and childless; the majority will be married with at least two children by age 25 (Measure DHS and ICF Macro 2011). The median age at first sex for Malawian women is 17, followed by first marriage at age 18, and first birth at age 19 (Measure DHS and ICF Macro 2011). Divorce is common, particularly in southern Malawi where our study is based, as is remarriage following divorce or widowhood (Reniers 2003).

Hypotheses

This article seeks to answer three questions. First, how stable are the family size preferences of young women living in and around a large town in Malawi? Second, how do important life events influence desired family size? Third, for which women are changes in family size preferences empirically predictable and for whom are they more volatile?

Building on the existing literature on fertility and fertility preferences in sub-Saharan Africa and recent longitudinal analyses of fertility preferences from the United States and Europe, we advance a set of hypotheses. First, we hypothesize that young Malawian women’s family size preferences will fluctuate, even over short periods of time. Studies from the West have found that young people are less vulnerable to social pressure around childbearing norms and less committed to a particular reproductive trajectory (Hayford 2009; Iacovou and Tavares 2011; Rocca et al. 2010). We further hypothesize that women with more established life course trajectories will have more predictable fertility preferences. More specifically, women who already have children or expect to begin childbearing soon (i.e., older and married women) will be both more likely to hold consistent family size preferences and to experience changes in ideal family size that are better predicted by the sequential theory of fertility. In contrast, the fertility preferences of younger and unmarried women depend on future contingencies that are uncertain (Hayford 2009), which may lead to more volatile preferences.

Relationship and Reproductive Events and Changes in Family Size Preferences

Women may alter their fertility preferences when they enter or exit a relationship. Unless there is perfect assortative mating with regard to preferences, partners’ desires will often differ (Voas 2003). Some studies have found that a partner’s preference influences fertility desires in the direction of that preference, and others have found that simply acquiring a partner exerts upward pressure on fertility desires (DeRose and Ezeh 2005; Iacovou and Tavares 2011; Joyce et al. 2000; Voas 2003). Partners’ preferences may have already changed by the time they marry (Voas 2003), however, so the formation of a serious relationship may have a more substantial influence on preferences than marriage. The 2010 Malawi DHS found that men and women of the same ages report similar ideal family size preferences and that ideal family size increases with age (Measure DHS and ICF Macro 2011). Because women tend to partner with older men (Barbieri et al. 2005), their ideal family size preferences going into a relationship are likely to be lower than their male partner’s preferences. We therefore hypothesize that, on average, the ideal family size preferences of young Malawian women will increase following the formation of a new serious partnership.

Few studies have examined how union dissolution influences fertility preferences. A UK study found that a marital breakup was associated with a reduction in expected family size among men but not women (Iacovou and Tavares 2011). Other research from the West has found that actual fertility falls following a marital breakup (Coppola and Di Cesare 2008; Lillard and Waite 1993; Schoen et al. 1999; Thomson et al. 2012). At the same time, research has demonstrated the importance of having a shared biological child within a union and higher-order unions leading to births that were not originally intended (Bledsoe 1990; Griffith et al. 1985; Thomson et al. 2002; Vikat et al. 1999). In southern Malawi, where divorce is common and usually closely followed by remarriage (Reniers 2003), similar pressures exist as each new union is expected to produce children. However, we expect union dissolution to be most likely to reduce family size preferences in the near term (our four- and eight-month reference periods).

Childbearing may lead to changes in family size preferences in two ways. First, an unexpected pregnancy may prompt an ex-post facto upward revision of preferences if it occurs after a woman has already achieved her ideal family size (Casterline and el-Zeini 2007; Freedman et al. 1965). Second, the experience of becoming a parent or becoming a parent of a certain number of children can affect preferences (Freedman et al. 1965; Hayford 2009; Heiland et al. 2008; Miller and Pasta 1995; Udry 1983). The literature focuses on the transition to parenthood as a key determinant of future preferences, but the model can also be applied to higher-order births. We hypothesize that young Malawian women will revise their family size preferences upward in response to a new pregnancy if they have already achieved their ideal family size. On the other hand, if they have not reached their desired family size, we expect that young women will shift their preferences downward in response to a new birth as the reality of parenting in a context increasingly accepting of smaller family sizes will exert downward pressure on a woman’s ideal family size.

Child mortality is thought to increase fertility through both physiological and volitional mechanisms. The two volitional mechanisms have been termed “replacement” and “insurance” effects (Rahman 1998). With replacement, the death of a child could lead to more rapid and ultimately higher fertility via a desire to replace the child who passed away. With the insurance effect, people who have personally experienced the death of a child, or who perceive high child mortality within their community (see Sandberg 2006), will have more children in order to guarantee a minimum number of surviving children in the face of future mortality. It is difficult to separate the physiological (i.e., shortened postpartum infecundability) from volitional influences of child mortality (Lindstrom and Kiros 2007); however, our focus on preferences allows us some leverage on the topic. Despite demographers’ interest in live births, people are interested in children and ideal family size is more of a sociological than biological construct (Ryder 1973). Therefore, even in the presence of genuine replacement motivation, desired family size need not increase. However, the death of a child may still lead to an increase in ideal family size if it serves as a form of insurance against the possibility of future child mortality—a possibility that may seem particularly likely based on recent experience.

Lastly, we expect that increases in educational attainment will be associated with reductions in family size preferences in accordance with economic theories of fertility and particularly the quality-quantity tradeoff (Becker and Lewis 1973; Kirk and Pillet 1998).

Data and Methods

Data

We use data from Tsogolo la Thanzi (TLT),2 a panel study of young adults in Balaka, a growing market town and district capital in southern Malawi. TLT was designed to study how young people navigate reproduction and union formation in an AIDS epidemic. The TLT sample was drawn from a complete household listing of people living within a 7 km radius of Balaka town center, which includes the rural villages that surround the more-urban town center. Approximately 1,500 women and 600 men between the ages of 15 and 25 were randomly selected from the household listing and recruited into the study. The present study is limited to female respondents because the smaller number of randomly sampled men and the lower prevalence of childbearing and marriage among men within the study’s age range limit analytic power.

Our analyses use eight waves of TLT data, each spaced four months apart. The first wave was collected between June and August 2009, and the last was collected between October and December 2011. At the end of each interview, respondents were asked to set up an appointment for their next interview approximately four months later. All interviews were conducted in private rooms at the TLT research center so that sensitive information could not be overheard. Respondents were provided with a 500-kwacha incentive (equivalent to approximately $3.50 USD) at each interview for their time and travel costs. Nearly all (97 %) contacted and eligible women completed a baseline interview (11 women completed a baseline interview at Wave 2). To ensure that the same respondents were reinterviewed at each wave, a digital photograph was taken of respondents at their first interview, and a receptionist used the photographs to confirm respondents’ identities at subsequent interviews.

The baseline TLT interview collected detailed demographic and socioeconomic information, pregnancy and marriage histories, and fertility preferences. Subsequent surveys were similar to the baseline survey but also focused on measuring events that occurred since the last interview. Our dependent variable is ideal family size (IFS), measured at each wave using the following question translated into Chichewa: “People often do not have exactly the same number of children they want to have. If you could have exactly the number of children you want, how many children would you want to have?” Because our sample is early in their reproductive lives and unlikely to have more children than their ideal at baseline, we interpret IFS as their desired number of children and use the terms “IFS” and “desired family size” interchangeably. Numeric responses range from 0 to 12. Respondents who gave nonnumeric responses, such as “up to God” or “don’t know,” or who had missing data on IFS at any wave are excluded from the analysis (n = 7).

The unit of analysis is a person-segment. We define a segment as the period of time between two successive waves of data in which a respondent participated. Respondents can contribute up to seven segments and are included in the analytic sample if they are interviewed at least twice. For example, a respondent interviewed at Wave 1 (baseline), Wave 2, Wave 3, Wave 4, and Wave 6 would contribute four segments of data, the last of which occurred between Waves 4 and 6. Three-quarters (75 %) of the individuals in the analytic sample were interviewed at all eight waves and contribute seven person-segments of data.

Methods

To address the first research question, we use descriptive statistics to examine the prevalence and direction of changes in IFS across segments. To address the second research question, we use a fixed-effects analytic framework. Fixed-effects models are appropriate to account for the structure of the panel data and the lack of independence of multiple observations from the same individual. We use a fixed-effects model rather than a random-effects model because Hausman tests indicated that the data violate the assumption that individual-level error is not correlated with observed covariates (Hausman 1978). Fixed-effects models yield consistent estimators by focusing on intraindividual variation and controlling for all measured and unmeasured time-constant variables (Petersen 2004). The approach allows for arbitrary correlation between individual-specific effects and our measures of relationship and reproductive events. This is important for studies of change in fertility preferences because it is extremely likely that there exist individual-level characteristics that affect both ideal family size and the life events of interest, and these would bias conventional estimates.

The fixed-effects model can be expressed as follows:
formula
(1)
where represents subject fixed effects; and are J time-varying covariates, such as age and the incidence of a new birth, marriage, or serious partnership. Because life events might not influence fertility preferences only over a four-month period, an additional model adds a one-segment lagged effect for all relationship and reproductive predictor variables.
The third research question centers on evaluating whether certain subject characteristics affect the volatility, or predictability, of IFS over time. Stated differently, the fixed-effects model in Eq. (1) provides a method for assessing how time-varying covariates relate to changes in IFS over time. In contrast, the third research question is concerned with understanding factors that relate to the residual variance. A significant body of research has developed frameworks and models for understanding factors that relate to error variance (Culpepper 2010; Smyth 2002). This research question is assessed with the following model:
formula
(2)
where is an intercept; and is the effect of the kth predictor, Zitk, on the log of the absolute value of the residuals. This article employs a two-stage estimation strategy using the statistical package R, where Eq. (1) is first estimated and the corresponding absolute valued residuals are used in Eq. (2). The second stage uses a gamma regression model with a log link function (Culpepper 2010). The slope for in Eq. (2) is interpreted as the percentage change in absolute errors for a one-unit change or difference in Zitk.

Independent Variables

At each wave following the baseline interview, respondents were asked whether they had experienced a series of events since their last interview. Using these questions, we develop a series of dichotomous indicators of relationship and reproductive life events that occur within segments. Relationship events include two indicators of union formation—getting married and gaining a new serious partner (i.e., a previously unreported boyfriend)—and two indicators of union dissolution—the end of a marriage (i.e., divorce, separation, or widowhood) and the end of a serious relationship. Reproductive events include a new pregnancy, a new birth, and the death of a child. The TLT protocol included pregnancy testing after each survey in addition to self-reported pregnancy status. Respondents are considered to have a new pregnancy if they report a new pregnancy during the segment or if they had an unexpected positive pregnancy test at the wave beginning the segment. A woman is considered to have achieved her ideal family size if her parity at the wave beginning a segment equals her ideal family size at that wave. The fixed-effects models additionally include time-varying measures of age, years of education, and school enrollment. Gamma regression models include time-varying measures of age, years of education, school enrollment, marital status (married, never married, and formerly married), and parity.

Panel Conditioning

Panel conditioning—bias introduced by participating in repeated waves of a longitudinal study—is a potential concern in a study design like that of TLT3 (Warren and Halpern-Manners 2012). We identify two potential types of panel conditioning that could affect the present study. The first is that repeated participation in the survey may lead to differences in the reporting of IFS, toward either more or less normative responses. Second, repeated participation could lead to more stable responses among survey respondents who are repeatedly encouraged to think about their desired number of children through survey participation (Warren and Halpern-Manners 2012). We take two approaches to assess the occurrence of these events. First, between February and April 2012, TLT conducted interviews with 315 women who were randomly sampled from the original 2009 sampling frame but were not part of the TLT panel study. These women were not statistically different from the analytic sample at Wave 8 on age, years of education, or school enrollment but were somewhat more likely to be married (65.4 % compared with 60.2 %; p = .095) and had slightly more children (1.5 compared with 1.4 children; p = .088). When we contrast the comparison sample’s reported IFS to that of the analytic sample at Wave 8, we find no statistically significant difference (3.5 in comparison compared with 3.4 in sample; p = .130), even after controlling for sociodemographic characteristics that might also be associated with panel attrition (p = .273) (Toepoel et al. 2009). This comparison is bolstered by the fact that attrition was relatively low in TLT: 80 % of women ever interviewed were reinterviewed in Wave 8, and 72 % of respondents lost to follow up were migrants, a reason for attrition that women in the comparison sample (drawn at the same time as the main sample) were also subject to. This offers some evidence that participation in TLT has not led respondents to report particularly high or low family size preferences.

To assess the second type of potential panel conditioning, we add a time-varying variable to the gamma regression model that captures the number of times a respondent was interviewed. This result is discussed at the end of the next section.

Results

The analytic sample consists of 1,449 women who contribute a total of 9,043 person-segments of data from 10,492 completed interviews. Table 1 presents descriptive statistics of the female analytic sample at their baseline interview. The mean age of respondents is 20 years, and 60 % are partnered: 42 % are married, and 18 % are in serious relationships. The mean years of education completed in the sample is just under 8, or just below “completed primary school,” and 39 % of respondents are enrolled in school at baseline. Parity is low (mean: 0.8 children); approximately one-half have not yet had children, and only 6 % have had as many as three (not shown). The vast majority of the sample wants between two and four children (89 %), which is similar to the national average for women in this age group (Measure DHS and ICF Macro 2011).

There is considerable instability in respondents’ family size preferences across segments (Table 2). Approximately one-quarter of women alter their reported ideal family size at each segment, and 66 % of women change their preference at some point over the 30-month period (Table 3). Although a few women have particularly unstable preferences that change across every segment (<1 %), the modal report is zero changes, and the median is to change in one in seven segments or fewer. Slightly more than one-quarter (28 %) of person-segments contain a change in ideal family size, 77 % of which are changes by one child in either direction.

The most common life events experienced by women in the sample are a change in years of education, acquisition of a new serious nonmarital partner, and a new pregnancy (Table 3).

Table 4 presents the parameter estimates for the fixed-effects regressions investigating the association between key life events and changes in IFS. As shown in Model 1, entering into a serious partnership is associated with a small increase in IFS; however, there is no relationship between getting married and a change in IFS. In terms of union dissolution, the relationship between the end of a marriage and a reduction in IFS is of marginal statistical significance (p = .063), but the dissolution of a serious partnership is not associated with a change in family size preferences. Despite our hypothesis, there is no significant relationship between a new birth and a change in IFS; however, women who learned of a new pregnancy between waves reduce their IFS, on average. IFS increases as women age, and there is a marginally significant negative association between an increase in years of education and family size preferences (p = .053).

Model 2 adds an interaction between women who achieved their reported IFS at the start of the segment and experienced a new pregnancy during the segment. This allows us to test whether the relationship between a new pregnancy and changes in ideal family size differs among women who have and have not yet had their preferred number of children. Respondents had achieved their IFS before 8 % of person-segments. The interaction is significant and indicates that women who have achieved their IFS yet go on to experience a pregnancy increase their family size preferences. In contrast, women who have yet to reach their desired family size tend to decrease their IFS after getting pregnant.

Model 3 builds on Model 1 by adding measures of relationship and reproductive events lagged by one segment. The inclusion of these lagged variables allows us to explore the sensitivity of family size preferences to the timing of life events by examining whether IFS changes are associated with life events in the short (one segment; approximately four months) and/or medium term (two segments; approximately eight months). All relationships identified in Model 1 remain significant in Model 3, and a few become stronger. The lagged variants of all significant associations between relationship and reproductive life events and changes in IFS are also statistically significant (at least at the p < .10 level). In other words, relationship and reproductive events do not just influence family size preferences in the short-term but can also have a somewhat delayed effect. Additionally, there emerges a relationship between a child death at the prior segment and an increase in IFS—a relationship that was not statistically significant in the short term.

Our third research question focuses on whether some women report more or less volatile family size preferences over time. Stated differently, we assess whether there are characteristics that are related to reporting family size preferences that are more or less predictable over time. Some of this residual variance will reflect noise resulting from measurement error or poorer construct validity among certain subpopulations. Indeed, 31 % of all changes in IFS could be considered corrective in that respondents’ reports change across two consecutive segments, ending up back at the original number (not shown). The gamma regression analysis models the residual variance from Model 1 in Table 4 and enables us to identify the individual-level characteristics that contribute to the unexplained variance. Table 5 shows that women who are younger, less educated, formerly married, and of low parity have the greatest variability in IFS. These women are the most likely to have changed their IFS in ways that are left unexplained by the relationship and reproductive events included in the fixed-effects models. For example, women with an additional year of education have errors that are 4 % smaller in absolute value. Similarly, a one-year increase in age is associated with a 2 % reduction in errors; each additional child a woman has is associated with an 11 % reduction, and being enrolled in school is associated with a 13 % reduction. Being divorced, widowed, or separated is associated with a 19 % increase in errors compared with being married. The results in Table 5 coupled with the negative relationship between marital dissolution and IFS suggest that dissolving a marriage, on average, tends to decrease IFS; and women who were formerly married fluctuate more around predicted values than married women. Lastly, there is no relationship between the number of surveys a woman had completed at the time she reported her IFS and unexplained variance, which helps to allay additional concerns about panel conditioning.

Discussion

Overall, we find a mix of support for our hypotheses. The first hypothesis about the prevalence of change in ideal family size is fully supported: young Malawian women frequently change their ideal family size preferences. Approximately one-quarter of women alter their ideal family size report across interviews, and just under two-thirds revise their preferences at some point during the study. Such instability, however, does not mean that family size preferences are “meaningless” to young women in Malawi and that such changes are necessarily random error (Hauser 1967:404). During a period of transitioning fertility, when norms around family size are evolving, IFS may be tentative but is still meaningful (Agadjanian 2005; Knodel and Prachuabmoh 1973). Although we interpret a change of approximately 25 % across waves as an indication of considerable movement, it is equally valid to acknowledge that 75 % of respondents report the same ideal family size across sequential waves, a remarkably consistent level of reporting if women do not ascribe to these goals.

Young women in Malawi are experiencing frequent changes in life circumstances, many of which manifest in genuine revisions in preferences. Consistent with sequential models of fertility, young Malawian women change their family size preferences following changes in their relationship and reproductive circumstances. They are more likely to revise their preferences when they enter a relationship than when a relationship progresses to marriage. This may reflect women adjusting their preferences to better accord with their new partner’s preferences, which are likely to be higher (at baseline, male partners were an average of 3.4 years older), or women may simply be altering their preferences as the prospect of starting a family becomes more tangible. In contrast, the dissolution of a marriage is associated with a reduction in preferences. Divorce and widowhood often leave women primarily responsible for the children they already have. Such an economic burden may encourage women to reduce their future fertility desires. Future studies should examine how the fertility preferences of both female and male partners change over the course of courtship, marriage, and divorce.

A number of reproductive events are also associated with changes in family size preferences. Young Malawian women’s family size preferences are more responsive to a new pregnancy than the birth that might follow it. Women who have yet to achieve their IFS tend to reduce their family size preferences following a pregnancy. In contrast, women who have already achieved their IFS at the time they get pregnant increase their IFS in response to a new pregnancy. In other words, the only women to reduce their family size preferences after a new pregnancy are those who can do so without it identifying their current pregnancy as unwanted. Many of these young women are likely to be new or recent parents, and the effects of pregnancy on women’s bodies and the financial costs associated with planning for a new baby might prompt them to reconsider their family size goals and shift them downward. In the context of fertility transition, in which fertility norms are in flux and a broader range of preferences are deemed socially acceptable, a pregnancy may prompt a woman to think more carefully about her realistic future fertility goals. This might result in a woman reducing her IFS to a level that is below her natal family size yet remains socially permissible in the context of lower fertility. Although some women who have achieved their ideal family size before getting pregnant may have changed their minds and wanted another child before getting pregnant, the short time periods used in the present study lead us to conclude that most of these cases are ex-post rationalizations of pregnancies that were unwanted at the time of conception. Such rationalizations reflect a small proportion of the overall changes in preferences among this young sample.

Child deaths over our study period are relatively rare events, and we are careful in interpreting the results from our analyses. Following the death of a child, women tend to increase their IFS, a finding that is statistically significant only for those deaths that occurred roughly between four and eight months earlier. Both replacement and insurance theories (as discussed earlier) posit that women will have additional children in response to the death of a child (Rahman 1998); however, women are more likely to report wanting more children as a mechanism to safeguard against child deaths that may occur in the future. We cautiously interpret this finding as evidence of an insurance effect, whereby women increase their desired family size to guard against the possibility of future child mortality.

Respondents’ family size preferences increase with age. This corresponds with cohort analyses of DHS data: even as average IFS has fallen among reproductive-age women, IFS actually increased slightly within birth cohorts (particularly younger ones) between 2000 and 2010 (Measure DHS and ICF Macro 2011; National Statistical Office and ORC Macro 2001). This may be indicative of young and less-experienced women stating lower family size preferences in order to present themselves as more “modern” during a period of life when they are constructing their personal identities and may perceive marriage and childbearing to be in the distant future. As women age, marry, and gain life experience, IFS may say less about the kind of person a woman wants to be and more about how she sees her now-tangible family life in the future. As hypothesized, IFS decreases as women gain more education, offering some evidence of a quality-quantity tradeoff.

Our study is subject to a number of limitations. First, in dynamically analyzing fertility preferences over time, we make the assumption that women’s preferences are particularly sensitive to new life events and change in response to them within a period of four to eight months. Some events that we examine, however, may not have such a prompt influence on a woman’s fertility preferences; rather, it may take some time for the effect of the event to be reflected in family size preferences. We address this to some extent by additionally modeling the lagged effect of all relationship and reproductive events and find evidence that most events that influence family size preferences over the short term (less than four months), also influence them with a slight delay (about four to eight months). Other events, however, might influence preferences over an even longer time frame. For example, we examine changes in IFS following a new birth; however, the sequential decision model of fertility suggests that the childrearing experience and not only the childbearing experience or sex of the child (i.e., things that are known as soon as the child arrives) can influence fertility preferences. Perhaps these processes take longer than eight months to unfold. Nonetheless, we believe that the innovation of using closely spaced waves provides a valuable opportunity for examining the dynamic development of fertility preferences and assessing questions of timing and sequencing that earlier studies cannot address with longer reference periods. Second, in demonstrating that certain life events are associated with changes in fertility preferences, we do not separate short-term reactions from long-term readjustments in family size goals. Even short-term changes, however, are important for understanding how women create and revise their family size ideals dynamically, and may be particularly relevant for young women at the beginning of their reproductive careers.

Third, it is possible that the life events we examine are correlated with other unobserved time-varying characteristics or events that influence fertility preferences. Although our fixed-effects models control for all time-constant variables that might influence fertility preferences and life events, unobserved time-varying variables remain a potential problem. For example, a change in economic circumstances might conceivably contribute to both the dissolution of a marriage and a decrease in ideal family size. We acknowledge our inability to isolate the single effect leading to a change in ideal family size, but we are reassured in our findings by the salience of reproductive and relationship events in young Malawian women’s lives, their relevance for changes in fertility preferences, and the short time periods over which we measure these events. Additionally, we must acknowledge the possibility—as unlikely as it may be—that the life events we measure occur in response to changes in ideal family size instead of the other way around.

Life course research has not had much purchase in developing nations (Dannifer 2003). Yet, our findings suggest that women in Malawi think about their future childbearing, articulate an ideal family size, and then adjust that family size goal in response to important life events that ostensibly change the way they think about their family. A recent qualitative study of young Malawian women underscores the applicability of life course theory in this context. Frye (2012) demonstrated that young Malawian schoolgirls rely on their (highly unrealistic) imagined educational futures to construct their current selves and shape their actions and behaviors. Although the young women in that study faced innumerable barriers to achieving the largely unrealistic educational goals they articulated, they continued to aspire to great heights because these goals were important in how young women viewed their lives and futures (Frye 2012). Just as those young women clearly articulated goals and plans for futures they knew might not materialize, we find young women in this study engaging in a similar practice: planning for a number of children that is ideal, given their current circumstances. As we would expect, based on what we know about the life course, these plans change as women gain life experience—especially with childbearing and relationships. Thus, as their life circumstances evolve, young Malawian women often revise their desired family size preferences. Many of these changes are patterned, which we interpret as evidence that IFS is a meaningful construct in this context.

We are careful not to argue that women pinpoint a specific ideal family size preference that does not waver even in the absence of altered circumstances. Instead, we adopt Coombs’ (1979) position that ideal family size is a meaningful point on a continuum around which there is some fluctuation. We find evidence of this in the residual variance of our models and our inability to explain all changes in ideal family size. As the results demonstrate, women for whom reproduction is a more distant goal (e.g., younger, lower parity) change their ideal family size in less predictable ways, some of which is likely corrective around a narrow range of desired children. We note two exceptions to this conclusion. First, women who are enrolled in school have more predictable preferences than women who are not. This may reflect the stability of the imagined futures held by young women in school. Second, formerly married women—those who are divorced, widowed, or separated—have less predictable preferences than women who are married. We offer two possible explanations here. It is possible that these women’s preferences are responsive to life events other than the ones included in our models. Alternatively, these women’s preferences are more volatile perhaps because they are living through a period of intense uncertainty about what their future might hold without the benefit of the naïve ideals to which less-experienced women can ascribe. While they undoubtedly have hopes for their future, they are also familiar with the challenges of reality and the barriers that often divert envisioned futures off course.

This study empirically tests the dynamic theory of fertility for young Malawian women using closely spaced panel data on family size preferences. Consistent with the sequential model, our sample of young Malawian women altered their ideal family size reports over time and in response to key life events. In this context of high and transitioning fertility, we find support for this theory in its more complex expression. It is not simply that decisions about births are made one by one, but rather that preferences are flexible and responsive to life circumstances, including relationship changes and undoubtedly other factors—such as economic conditions—that we have not considered here. Are family size preferences fixed? Definitively not. Are they perfect guides for reproductive behavior? Certainly not. And yet, we think that even in a higher-fertility context, they represent a moving target that takes into account internalized societal norms, relationship context, reproductive experiences, and individual hopes. Family size preferences are dependent on a “cloudy” view of the future but are best guesses at the time of interview as to what that future might entail (Ryder 1973:502).

Our results have implications for demographic survey research more generally. Family size preferences reflect the conditions an individual is experiencing at one particular point in time: namely, the time of the survey. Researchers should be careful not to label such preferences as meaningless when they fail to predict completed family size at the end of one’s reproductive life. Nor do mismatches between ideal family size at the beginning of one’s reproductive career and at the end necessarily reflect rationalizations of a reality that did not turn out as planned. As we have shown, even young women early in their reproductive careers regularly alter their preferences, often in predictable ways.

Acknowledgements

This study was supported by two grants from the National Institute of Child Health and Human Development: R01 HD058366 and R03 HD067099. We are indebted to Jenny Trinitapoli, Stefanie Mollborn, Margaret Frye, Rob Warren, colloquium participants at the University of Colorado’s Institute of Behavioral Science, and Demography’s reviewers for helpful comments on earlier versions of this article; any errors are our own. This research was made possible by the efforts of the Tsogolo la Thanzi fieldwork team, headed by Abdallah Chilungo, Sydney Lungu, and Hazel Namadingo.

Notes

1

Fig. 1 in the  appendix shows ideal family size reports by five-year birth cohorts using Malawi DHS data from 1992, 2000, 2004, and 2010. Overall, there is little change within birth cohorts over this period, although there is a slight decrease between 1992 and 2000 and a slight increase between 2000 and 2010.

2

Tsogolo la Thanzi is a research project designed by Jenny Trinitapoli and Sara Yeatman, and funded by Grant R01 HD058366 from the National Institute of Child Health and Human Development. Details are available online (http://projects.pop.psu.edu/tlt).

3

The evidence for panel conditioning is most clear in studies with surveys spaced less than one month apart. The evidence for surveys spaced between one month and one year is more mixed (Warren and Halpern-Manners 2012).

Appendix

References

Agadjanian, V. (
2005
).
Fraught with ambivalence: Reproductive intentions and contraceptive choices in a sub-Saharan fertility transition
.
Population Research and Policy Review
,
24
,
617
645
. 10.1007/s11113-005-5096-8
Bankole, A. (
1995
).
Desired fertility and fertility behaviour among the Yoruba of Nigeria: A study of couple preferences and subsequent fertility
.
Population Studies
,
49
,
317
328
. 10.1080/0032472031000148536
Bankole, A., & Singh, S. (
1998
).
Couples’ fertility and contraceptive decision-making in developing countries: Hearing the man’s voice
.
International Family Planning Perspectives
,
24
,
15
24
. 10.2307/2991915
Bankole, A., & Westoff, C. F. (
1998
).
The consistency and validity of reproductive attitudes: Evidence from Morocco
.
Journal of Biosocial Science
,
30
,
439
455
. 10.1017/S0021932098004398
Barbieri, M., Hertrich, V., & Grieve, M. (
2005
).
Age difference between spouses and contraceptive practice in sub-Saharan Africa
.
Population
,
60
,
617
654
.
Becker, G. S., & Lewis, H. G. (
1973
).
On the interaction between the quantity and quality of children
.
Journal of Political Economy
,
81
,
S279
S288
. 10.1086/260166
Bledsoe, C. (
1990
).
Transformations in sub-Saharan African marriage and family
.
The Annals of the American Academy of Political and Social Science
,
510
,
115
125
. 10.1177/0002716290510001009
Bledsoe, C., Fatoumatta, B., & Hill, A. G. (
1998
).
Reproductive mishaps and Western contraception: An African challenge to fertility theory
.
Population and Development Review
,
24
,
15
57
. 10.2307/2808121
Bulatao, R. A. (
1981
).
Values and disvalues of children in successive childbearing decisions
.
Demography
,
18
,
1
25
. 10.2307/2061046
Caldwell, J. C., & Caldwell, P. (
1987
).
The cultural context of high fertility in sub-Saharan Africa
.
Population and Development Review
,
13
,
409
437
. 10.2307/1973133
Casterline, J. B., & el-Zeini, L. O. (
2007
).
The estimation of unwanted fertility
.
Demography
,
44
,
729
745
. 10.1353/dem.2007.0043
Coombs, L. C. (
1979
).
Underlying family-size preferences and reproductive behavior
.
Studies in Family Planning
,
10
,
25
36
. 10.2307/1966175
Coppola, L., & Di Cesare, M. (
2008
).
How fertility and union stability interact in shaping new family patterns in Italy and Spain
.
Demographic Research
,
18
(
article 4
),
117
144
. 10.4054/DemRes.2008.18.4
Culpepper, S. A. (
2010
).
Studying individual differences in predictability with gamma regression and nonlinear multilevel models
.
Multivariate Behavioral Research
,
45
,
153
185
. 10.1080/00273170903504885
Dannifer, D. (
2003
).
Whose life course is it, anyway? Diversity and “linked lives” in global perspective
. In R. A.Settersten Jr. (Ed.),
Invitation to the life course: Toward new understandings of later life
(pp.
259
268
).
Amityville, NY
:
Baywood Publishing Company
.
Debpuur, C., & Bawah, A. A. (
2002
).
Are reproductive preferences stable? Evidence from rural northern Ghana
.
Genus
,
LVIII
,
63
89
.
DeRose, L. F., & Ezeh, A. C. (
2005
).
Men’s influence on the onset and progress of fertility decline in Ghana, 1988–98
.
Population Studies
,
59
,
197
210
. 10.1080/00324720500099496
Ezeh, A. C., Seroussi, M., & & Raggers, H. (
1996
).
Men’s fertility, contraceptive use, and reproductive preferences
(Demographic and Health Survey Comparative Studies No. 18).
Calverton, MD
:
Macro International Inc
.
Freedman, R., Coombs, L. C., & Bumpass, L. (
1965
).
Stability and change in expectations about family size: A longitudinal study
.
Demography
,
2
,
250
275
. 10.2307/2060117
Freedman, R., Freedman, D. S., & Thornton, A. D. (
1980
).
Changes in fertility expectations and preferences between 1962 and 1977: Their relation to final parity
.
Demography
,
17
,
365
378
. 10.2307/2061151
Frye, M. (
2012
).
Bright futures in Malawi’s new dawn: Educational aspirations as assertions of identity
.
The American Journal of Sociology
,
117
,
1565
1624
. 10.1086/664542
Griffith, J. D., Koo, H. P., & Suchindran, C. M. (
1985
).
Childbearing and family in remarriage
.
Demography
,
22
,
73
88
. 10.2307/2060987
Hauser, P. M. (
1967
).
Review: “Family planning and population programs”: A book review article
.
Demography
,
4
,
397
414
. 10.2307/2060379
Hausman, J. A. (
1978
).
Specification texts in econometrics
.
Econometrica
,
46
,
1251
1271
. 10.2307/1913827
Hayford, S. R. (
2009
).
The evolution of fertility expectations over the life course
.
Demography
,
46
,
765
783
. 10.1353/dem.0.0073
Hayford, S. R., & Morgan, S. P. (
2008
).
The quality of retrospective data on cohabitation
.
Demography
,
45
,
129
141
. 10.1353/dem.2008.0005
Heiland, F., Prskawetz, A., & Sanderson, W. C. (
2008
).
Are individuals’ desired family sizes stable? Evidence from West German panel data
.
European Journal of Population
,
24
,
129
156
. 10.1007/s10680-008-9162-x
Hout, M. (
1978
).
The determinants of marital fertility in the United States, 1968–1970: Inferences from a dynamic model
.
Demography
,
15
,
139
160
. 10.2307/2060519
Iacovou, M., & Tavares, L. P. (
2011
).
Yearning, learning, and conceding: Reasons men and women change their childbearing intentions
.
Population and Development Review
,
37
,
89
123
. 10.1111/j.1728-4457.2011.00391.x
Johnson-Hanks, J. (
2007
).
Natural intentions: Fertility decline in the African Demographic and Health Surveys
.
The American Journal of Sociology
,
112
,
1008
1043
. 10.1086/508791
Joyce, T., Kaestner, R., & Korenman, S. (
2000
).
The stability of pregnancy intentions and pregnancy-related maternal behaviors
.
Maternal and Child Health Journal
,
4
,
171
178
. 10.1023/A:1009571313297
Joyce, T., Kaestner, R., & Korenman, S. (
2002
).
On the validity of retrospective assessments of pregnancy intention
.
Demography
,
39
,
199
213
. 10.1353/dem.2002.0006
Kirk, D., & Pillet, B. (
1998
).
Fertility levels, trends, and differentials in sub-Saharan Africa in the 1980s and 1990s
.
Studies in Family Planning
,
29
,
1
22
. 10.2307/172178
Knodel, J., & Prachuabmoh, V. (
1973
).
Desired family size in Thailand: Are the responses meaningful?
.
Demography
,
10
,
619
637
. 10.2307/2060887
Kodzi, I. A., Casterline, J. B., & Aglobitse, P. (
2010
).
The time dynamics of individual fertility preferences among rural Ghanaian women
.
Studies in Family Planning
,
41
,
45
54
. 10.1111/j.1728-4465.2010.00223.x
Lee, R. D. (
1980
).
Aiming at a moving target: Period fertility and changing reproductive goals
.
Population Studies
,
34
,
205
226
.
Liefbroer, A. C. (
2009
).
Changes in family size intentions across young adulthood: A life-course perspective
.
European Journal of Population
,
25
,
363
386
. 10.1007/s10680-008-9173-7
Lillard, L. A., & Waite, L. J. (
1993
).
A joint model of marital childbearing and marital disruption
.
Demography
,
30
,
653
681
. 10.2307/2061812
Lindstrom, D. P., & Kiros, G-E (
2007
).
The impact of infant and child death on subsequent fertility in Ethiopia
.
Population Research and Policy Review
,
26
,
31
49
. 10.1007/s11113-006-9018-1
McCarthy, J., & Oni, G. A. (
1987
).
Desired family size and its determinants among urban Nigerian women: A two-stage analysis
.
Demography
,
24
,
279
290
. 10.2307/2061635
Malawi Demographic and Health Survey 1992
. (
1994
).
Zomba, Malawi, and Calverton, MD
:
National Statistical Office
.
Malawi Demographic and Health Survey 2010
. (
2011
).
Zomba, Malawi, and Calverton, MD
:
National Statistical Office
.
Miller, W. B., & Pasta, D. J. (
1995
).
Behavioral intentions: Which ones predict fertility behavior in married couples?
.
Journal of Applied Social Psychology
,
25
,
530
555
. 10.1111/j.1559-1816.1995.tb01766.x
Namboodiri, N. K. (
1972
).
Some observations on the economic framework for fertility analysis
.
Population Studies
,
26
,
185
206
.
Malawi Demographic and Health Survey 2000
. (
2001
).
Zomba, Malawi, and Calverton, MD
:
National Statistical Office and ORC Macro
.
Petersen, T. (
2004
).
Analyzing panel data: Fixed- and random-effects models
. In Hardy, M., & Bryman, A. (Eds.),
Handbook of data analysis
(pp.
331
346
).
Los Angeles, CA
:
Sage
.
Rahman, M. (
1998
).
The effect of child mortality on fertility regulation in rural Bangladesh
.
Studies in Family Planning
,
29
,
268
281
. 10.2307/172273
Reniers, G. (
2003
).
Divorce and remarriage in rural Malawi
.
Demographic Research, Special Collection
1
,
175
206
. doi:10.4054/DemRes.0.S6.1
Rocca, C. H., Doherty, I., Padian, N. S., Hubbard, A. E., & Minnis, A. M. (
2010
).
Pregnancy intentions and teenage pregnancy among Latinas: A mediation analysis
.
Perspectives on Sexual and Reproductive Health
,
42
,
186
196
. 10.1363/4218610
Ryder, N. B. (
1973
).
A critique of the National Fertility Study
.
Demography
,
10
,
495
506
. 10.2307/2060877
Ryder, N. B. (
2010
).
Norman Ryder on the sociology of fertility reduction
.
Population and Development Review
,
36
,
607
620
. 10.1111/j.1728-4457.2010.00348.x
Ryder, N. B., & Westoff, C. F. (
1967
).
The trend of expected parity in the United States: 1955, 1960, 1965
.
Population Index
,
33
,
153
168
. 10.2307/2733064
Sandberg, J. (
2006
).
Infant mortality, social networks, and subsequent fertility
.
American Sociological Review
,
71
,
288
309
. 10.1177/000312240607100206
Schoen, R., Astone, N. M., Kim, Y. J., Nathanson, C. A., & Fields, J. M. (
1999
).
Do fertility intentions affect fertility behavior?
.
Journal of Marriage and the Family
,
61
,
790
799
. 10.2307/353578
Sennott, C., & Yeatman, S. (
2012
).
Stability and change in fertility preferences among young women in Malawi
.
International Perspectives on Sexual and Reproductive Health
,
38
,
34
42
. 10.1363/3803412
Smyth, G. K. (
2002
).
An efficient algorithm for REML in heteroscedastic regression
.
Journal of Graphical and Computational Statistics
,
11
,
836
847
. 10.1198/106186002871
Thomson, E. (
1997
).
Couple childbearing desires, intentions, and births
.
Demography
,
34
,
343
354
. 10.2307/3038288
Thomson, E., Hoem, J. M., Vikat, A., Prskawetz, A., Buber, I., Toulemon, L., & . . . Kantorova, V. (
2002
).
Childbearing in stepfamilies: How parity matters
. In Corijn, M. and Klijzing, E. (Eds.),
Dynamics of fertility and partnership in Europe: Insights and lessons from comparative research
(pp.
87
99
).
New York, and Geneva, Switzerland
:
United Nations
.
Thomson, E., Winkler-Dworak, M., Spielauer, M., & Prskawetz, A. (
2012
).
Union instability as an engine of fertility? A microsimulation model for France
.
Demography
,
49
,
175
195
. 10.1007/s13524-011-0085-5
Toepoel, V., Das, M., & van Soest, A. (
2009
).
Relating question type to panel conditioning: Comparing trained and fresh respondents
.
Survey Research Methods
,
3
,
73
80
.
Udry, J. R. (
1983
).
Do couples make fertility plans one birth at a time?
.
Demography
,
20
,
117
128
. 10.2307/2061230
van de Walle, E. (
1992
).
Fertility transition, conscious choice, and numeracy
.
Demography
,
29
,
487
502
. 10.2307/2061848
Van Lith, L. M., Yahner, M., & & Bakamjian, L. (
Forthcoming
).
Women’s growing desire to limit births in sub-Saharan Africa: Meeting the challenge
.
Global Health: Science and Practice
.
Vikat, A., Thomson, E., & Hoem, J. M. (
1999
).
Stepfamily fertility in contemporary Sweden: The impact of childbearing before the current union
.
Population Studies
,
53
,
211
225
. 10.1080/00324720308082
Voas, D. (
2003
).
Conflicting preferences: A reason fertility tends to be too high or too low
.
Population and Development Review
,
29
,
627
646
. 10.1111/j.1728-4457.2003.00627.x
Warren, J. R., & Halpern-Manners, A. (
2012
).
Panel conditioning in longitudinal social science surveys
.
Sociological Methods and Research
,
41
,
491
534
. 10.1177/0049124112460374
Westoff, C. F., & Ryder, N. B. (
1977
).
The predictive validity of reproductive intentions
.
Demography
,
14
,
431
453
. 10.2307/2060589
Yamaguchi, K., & Ferguson, L. R. (
1995
).
The stopping and spacing of childbirths and their birth-history predictors: Rational-choice theory and event-history analysis
.
American Sociological Review
,
60
,
272
298
. 10.2307/2096387