Abstract

From Ryder's (1973) “cloudy future” to Lee's (1980) “moving target,” the malleable nature of fertility preferences is widely accepted; however, tools for conceptualizing and measuring preferences have been slow to evolve. In this article, we seek to demonstrate that underpinning the number people give when asked about their ideal family size, and critical to interpreting it, is either a rigidness or a flexibility that is contextually situated and dynamic over the life course. To illustrate the difference between fixed and flexible orientations to fertility, we draw upon the metaphor of a movable feast. While some religious holidays like Christmas and All Souls Day occur annually on fixed days, others like Passover, Easter, and Pentecost change from year to year depending upon the lunar cycle or other ecclesiastical dates. From the perspective of the Gregorian calendar, movable feasts seem irregular, unpredictable, and sometimes merely seasonal. But movable feasts are no less regular than fixed feasts, nor are they of secondary importance; they differ in being governed by a distinct, flexible logic. Like many major religious feasts, some people's fertility preferences are indeed fixed, but it is the movable ones we endeavor to theorize here. Our interest in flexibility is anchored in two recent empirical and theoretical developments in the demographic literature. First, although instability in fertility preferences is higher in developing contexts than in the West (Bankole and Westoff 1995; Kodzi, Johnson, and Casterline 2010), new evidence suggests that this preference instability is not simply random noise but frequently patterned. In Malawi, for example, the setting of our present study, women change their numeric and timing preferences in response to changes in their relationships (divorce, widowhood, new marriage) and reproductive circumstances (pregnancies and child mortality) (Sennott and Yeatman 2012; Yeatman, Sennott, and Culpepper 2013). Furthermore, instability itself has predictive power with respect to short-term fertility outcomes (Kodzi et al. 2010). Second, scholars have developed more theoretically sophisticated ways of thinking about fertility. The Theory of Conjunctural Action (TCA) (Johnson-Hanks et al. 2011) critiques theories of fertility that reduce it to planned action because they rely on assumptions of clarity about a predictable, imagined future. Such an idealized view of reality is difficult or impossible to reconcile with the messiness of real lives, in which young adults lack “definite knowledge about their future family formation, education, and career paths, and so are likely to form their fertility intentions based on general social norms rather than specific desires” (Hayford 2009: 767). Fertility preferences, therefore, should only rarely be treated as a fixed statement of a feasible plan, and researchers should expect fertility behaviors to respond to contingencies, inputs, and shifts that occur at the micro and macro levels. Despite widespread agreement on these points, how can demographers go about integrating notions of flexibility into empirical research on fertility? Our interest in the nature of flexibility is further motivated by the puzzle of fertility transition in sub-Saharan Africa. The shape of the African fertility transition is distinct from the patterns of decline that characterized Latin America and Asia during the second half of the twentieth century (Bongaarts and Casterline 2013; Casterline and El-Zeini 2007). The debate about whether Africa's fertility transition is late, stalled, or simply different is ongoing (Bongaarts 2017; Bongaarts and Casterline 2013; Caldwell and Caldwell 2002; Caldwell, Orubuloye, and Caldwell 1992; Casterline and Agyei-Mensah 2017; Mbacké 1994; Moultrie, Sayi, and Timæus 2012; Shapiro and Gebreselassie 2008; Smith 2004). Despite differing in their views of the nature of the transition, scholars agree that fertility preferences play a central role in the transition and in our capacity to develop a better understanding of it. Fertility preferences are critical because, in the words of Bongaarts and Casterline (2013: 159), they “represent a key link in the chain of causation between fertility and its socioeconomic determinants.” Questions about the nature of fertility preferences raise especially challenging issues for researchers working in sub-Saharan contexts. While some researchers maintain that fertility rates in sub-Saharan Africa remain high precisely because desired fertility has remained high (Bongaarts 2006; Bongaarts and Casterline 2013; Pritchett 1994), others read the evidence differently. Günther and Harttgen (2016), for example, document that across the region realized fertility has exceeded wanted fertility by two children for more than two decades. They interpret this gap as evidence that African women are less capable of translating child preferences into outcomes than are women in other developing contexts. An extensive literature on fertility in sub-Saharan Africa attributes the observed gaps between preferences and completed fertility to “unmet contraceptive need” (Bankole and Ezeh 1999; Bongaarts 1991; Frank and Bongaarts 1991; Sedgh and Hussain 2014). Complicating this interpretation are three factors: i) the important but often neglected caution against inferring individual intentions from population rates (Johnson-Hanks 2007); ii) the weak relationship between family planning programs and fertility reduction (Günther and Harttgen 2016); and iii) new evidence that unrealized fertility is far more prevalent in this region than previously recognized. Recent estimates from individual-level analyses reveal that despite high levels of fertility in sub-Saharan Africa, as many as 46 percent of African women fall short of their ideal at the end of childbearing (Casterline and Han 2017). Other common explanations for the fact that preferences and behaviors tend to be misaligned in sub-Saharan Africa include poor data quality (Dare and Cleland 1994), poor construct validity (Bankole 1995), and the possibility that reproductive decisions remain outside the calculus of conscious choice (Coale 1973; van de Walle 1992). To readers familiar with recent developments in the methods and materials of demography, the limits of these explanations are apparent. While concerns about data quality are valid, data availability (if not quality) has been improving, and statistical methods for treating preferences as dynamic processes have advanced considerably. Non-numeric responses to questions about fertility desires such as “Don't know” and “Up to God” have declined (Frye and Bachan 2017), suggesting that the vast majority of women think numerically about the future with respect to their families. Whereas the presence of some uncertainty about one's future is universal, most high-fertility societies are characterized by rampant uncertainty, and scholars from various disciplines have argued that flexibility is a strategic response to the many uncertainties of life in the African context. The connection between uncertainty and flexibility has been elaborated by scholars of rural livelihoods working among Kenyan pastoralists (Butt 2011), Yoruba cacao farmers (Berry 1993), and Mandara Mountain dwellers (Lev and Campbell 1987). In settings where “no condition is permanent” (Berry 1993), flexibility in everything from the selection of land and crops to the timing of labor for planting and harvest is crucial to survival in both the short and long term. Similarly, flexibility in childbearing is a strategic response to life's uncertainties. Describing the landscape of her research site, Johnson-Hanks (2005: 364) underscored the exceptionally high levels of existential and economic uncertainty. “Life in contemporary Cameroon is extremely uncertain, both in the specific sense that death often comes early and unexpectedly and also more generally: few events in everyday experience are predictable or consistent. From buses to paychecks to roadblocks to prices, common things elude standardization.” Where day-to-day life is riddled with uncertainties, being flexible about all sorts of things—including childbearing (how much and when)—is a necessity (Johnson-Hanks 2005). Posited as an alternative to rational-choice perspectives on action, the notion of “judicious opportunism” captures the ease with which individuals can withdraw from their prior intentions (intentions that were real when articulated) by seizing opportunities to reach desirable ends rather than struggling against tides to manifest a fixed and actionable plan. According to Johnson-Hanks, judicious opportunism is found not just in Cameroon but wherever the usual supports for rational choice are wobbly. And while the flexibility that judicious opportunism requires may look, on the surface, like “just waiting” or like indecision or inaction, flexibility is distinct from these responses in that it is strategic. Examples of uncertainty and its relationship to fertility abound; here we point to two additional examples—one emphasizing the existential and the other the economic. Extending the literature on insurance or replacement effects on fertility (Cain 1981; Caldwell et al. 1992; LeGrand et al. 2003; Randall and LeGrand 2003), Sandberg (2006) used network data from an agrarian community in Nepal to demonstrate that uncertainty about child survival (proxied by high levels of infant mortality within a network of conversational partners) accelerated and increased women's fertility. Using qualitative data from peri-urban Mozambique, Agadjanian (2005) looked not to the graveyard but to the market, arguing that stated fertility desires are conditional on current economic and social circumstances and that reproductive aspirations (especially at lower parities) should be treated as tentative because they are shaped by assessments of an unknowable future. While Agadjanian acknowledged the persistent poverty that characterizes much of his study population, he emphasized not the poverty itself but “the unpredictability of the economic situation” (2005: 625), including structural factors like labor market opportunities and intimate conjunctures like spousal migration and relationship strain. Because the data demands for examining preference change are high, most evidence comes from the data-rich West (Hayford 2009; Heiland, Prskawetz, and Sanderson 2008; Iacovou and Tavares 2011; Liefbroer 2009; Udry 1983); however, the evidentiary basis from Africa is growing (e.g., Kodzi et al. 2010; Yeatman et al. 2013). The roots of this literature on preference change can be found in theories about the impact of child mortality, post-hoc rationalization, and household bargaining, but new research shows that other types of events provide change as well. In our context of Malawi, a wide array of conjunctures is known to affect both the number of children a woman wants and will subsequently have and the timing of those pregnancies and births. Confirmed or suspected HIV infection, for example, leads women to accelerate their childbearing plans in order to achieve their ideal family size while still in good health (Trinitapoli and Yeatman 2011), while caregiving responsibilities for non-biological children (fostering) often lead women to reduce their numeric preferences and delay their childbearing (Bachan 2015). Labor market opportunities may lead women to adjust their timing preferences (Sennott and Yeatman 2012), while expectations that any serious relationship would be solidified through offspring mean that partnership changes in the wake of death or divorce tend to increase desired fertility (Verheijen 2013; Yeatman et al. 2013). In sum, fertility preferences are contingent and are unstable over time in ways that are patterned. What remains less clear is whether and to what extent strategic flexibility i) varies with perceptions and experiences of uncertainty, ii) may help account for the high levels of preference instability observed in sub-Saharan Africa, and iii) helps explain a unique pattern of fertility-related behaviors and outcomes. Reflecting on his experience directing the US National Fertility Study (NFS), Ryder likened the task of asking American respondents to identify their optimal reproductive target to “asking the respondent to perform a complex conceptual experiment: ‘If everything else in your life were to remain the same, except for your parity, what would you choose for your parity?’” He continued, “I suspect that respondents, faced with this challenge, can scarcely avoid thinking of other things they would like to change in addition to the number of children, such as their health, or their housing, or perhaps their husband, unless, of course, they reject the game altogether and converge on their actual experience” (Ryder 1973: 504). When researchers ask questions such as “If you could have exactly the number of children you want, what number would that be?,” women almost always answer clearly, providing a single number. However, underpinning these numbers are processes that are messy to model but central to understanding fertility preferences and what they do (and do not) tell us. During the heyday of research on fertility preferences, several scholars sought to supplement best-practice measures of ideal family size (IFS) with new constructs that could tap, prospectively, an underlying structure that would tell us more. For example, Coombs advanced both theory and measurement related to fertility preferences, employing the metaphor of preferences unfolding around a personal ideal (the target). She codified this metaphor in a set of preference scales that were part of a broader endeavor to generate more valid, sensitive, and refined measures of preferences: “[I]f we are to explore in more precise fashion than heretofore the antecedents and correlates of preferences, measures beyond global statements about preferred numbers provide valuable tools” (Coombs 1974: 609). Coombs-authored preference scales force respondents to move a beyond their initial target to reveal underlying preferences. To describe her “unfolding theory” of fertility, Coombs began with an exercise that moved respondents to either end of a constrained spectrum of ideal family size: “‘If you couldn't have ___ (number given) would you rather have __ (lower number) or __ (higher number)?’ and so on until the respondent chose zero or six” (1974: 588–89). Today, the most widespread adaptation of the Coombs scale forces respondents to choose second and third preferences, which enables analysts to identify women's underlying preference for a small or a large family but masks variability in movement up and down the IFS spectrum by constraining the amount of variation within each sample to two shifts per person. Concerned primarily with the instability of fertility preferences over the life course, Morgan (1981) built upon Coombs's insights, offering a simple but elegant alternative. Leveraging changes in preferences observed among American women from the National Fertility Studies conducted in 1965 and 1970, he insisted: “This uncertainty is not ‘noise’ in the data that should be ignored, discarded, or removed by some post hoc coding procedure. Rather, it is a real phenomenon inherently part of fertility decision making” (Morgan 1981: 268). The 1965 NFS asked those respondents who indicated their intention to have more children, “Do you think you might later change your minds and decide not to have another child?” And it asked respondents who indicated the intention to stop, “Do you think you might later decide to have another child?” Learning that 7 percent did not know their intentions to begin with, 13 percent were uncertain of their intention to stop, and 50 percent were uncertain of their stated intention to have more, Morgan commented that these high levels of inconsistency between intentions and behaviors were “not surprising” (p. 280). Morgan's work pointed to flexibility as an inherent part of fertility intentions, worthy of further theoretical and empirical attention. But despite the fact that these two simple questions actually did, to some extent, index individuals’ willingness to revise their preferences, such questions are rarely used by researchers today and are seldom asked outside of the West. Like Coombs, we believe that a structure underlies each person's stated ideal. However, rather than conceptualizing this structure numerically, as a type of statistical uncertainty, we focus on the level of flexibility that characterizes individual preferences. By flexibility we mean the extent to which preferences are designed to shift in the wake of the evolving contingencies. We posit that flexibility is measurable and intrinsically linked to fertility preferences and that measuring the prevalence of and variation in flexibility can increase understanding of fertility processes broadly. When viewing world fertility patterns from a fixed-feasts perspective, an unacceptably large proportion of fertility preferences appear unstable, invalid, unpredictable, and untrustworthy; we argue here that for large portions of the world's population, this instability is not an anomaly to be corrected for, but, like movable feasts, an essential aspect of their nature. The data for our study come from Tsogolo la Thanzi (TLT), a longitudinal study conducted in Balaka, Malawi designed to examine how, in the context of a generalized AIDS epidemic, young adults navigate the sometimes incompatible goals of enjoying sexual relationships, bearing children, and avoiding HIV infection. Balaka is a bustling township located in Malawi's southern region at the crossroads between a major road linking the country's political capital (Lilongwe) with its cultural capital (Zomba) and the rail route that ferries goods between Salima and Blantyre. The common refrain “In Balaka, every day is market day” attests to the vibrancy of this rapidly growing trading town. Several other pieces of information contextualize the setting in which we examine young adults’ fertility goals in the context of broader concerns. First, the economic conditions characterizing Balaka are harsh. Despite the commercial activity, southern Malawi is poorer than the rest of the country. The southern region features lower levels of educational attainment and higher levels of poverty than the northern and central regions (MDHS 2010). Most residents of Balaka are subsistence farmers; there is just one paved road, and in 2009 only 12 percent of households had access to electricity. Second, of epidemiological relevance, the southern region has the country's most severe AIDS epidemic; in 2010, 15 percent of the population aged 15–49 in the southern region was infected with HIV, compared to 8 percent in the central and 7 percent in the northern region (ibid.). In pan-African perspective, Balaka's epidemic might best be described as severe but improving: HIV prevalence in Malawi's southern region, estimated at 17 percent in 2004, has fallen to 12 percent but remains twice as high as prevalence in nearby regions. Recent causes for optimism include a decline in new infections, expanded access to antiretroviral treatment, and falling AIDS-related mortality (ibid.). Despite these improvements, however, the epidemic has engendered widespread uncertainty: a sizable proportion of the population is unsure of their current HIV status, and worry about future infection is omnipresent for the vast majority of young adults (Kaler and Watkins 2010; Trinitapoli and Yeatman 2011; Watkins 2004). Third, the transition to adulthood unfolds quickly: women become sexually active around age 17, marry for the first time about a year later, and give birth to their first child about a year after that (Boileau et al. 2009; Clark, Poulin, and Kohler 2009; Poulin 2007). The first wave of data collection for TLT took place between May and August 2009. A simple random sample of 1,505 female respondents was drawn from a sampling frame of 15 to 25 year olds living in villages within a seven-kilometer radius of Balaka's main market. The catchment area includes a mix of rural and peri-urban communities around the trading center. TLT interviewers first contacted respondents in their homes and arranged a time for an interview. Respondents came to the research center, adjacent to the town's main market, and were interviewed in private rooms where their responses could not be overheard by family members and neighbors. Each survey took approximately 90 minutes. At Wave 1, refusal at the time of making an appointment and passive refusal by not showing up were rare (97 percent of sampled and eligible respondents completed a baseline interview). Eight waves of data were collected from this cohort of women through 2011, with survey rounds scheduled at four-month intervals; the response rate at Wave 8 was 81 percent. In 2015, a ninth wave of data (TLT-2) was collected from the original sample of respondents; TLT-2 had an 80 percent response rate. The first panel of Table 1 describes all measures and summarizes the characteristics of the sample at Wave 1 (N = 1,505). At baseline, respondents ranged in age from 15 to 25, with a mean age of 19.5 years. Variability in rural/urban residence within the catchment area is captured by a standardized measure of distance from the town center. Mean education was 7.7 years—just shy of primary school completion. Half of respondents had ever been or were currently married. Thirteen percent of the sample was pregnant at baseline (not shown), and 10 percent of women reported having experienced a miscarriage or child death. Average parity was 0.79; about half the sample had no children, while others (N = 11) had four or five children at baseline. In contrast to van de Walle's (1992) respondents from Bamako nearly three decades ago, young women in Balaka had no trouble giving numeric responses to questions about ideal family size. Only two failed to respond to our question about IFS by giving a number. Ideal family size ranged from 0 to 8 children, with a mean of 3.22 children. By 2015 (six years later), 85.5 percent had at least one child (not shown), and mean parity was 1.92. Our two key measures of fertility preferences are ideal family size (a numeric preference) and ideal time to next birth (a timing preference). To measure IFS, we asked: “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?” Response categories for ideal time to next birth range from one (“as soon as possible”) to six (“five or more years”). Immediately following these questions, we endeavored to measure the flexibility of preferences. Interviewers asked each respondent how she would respond to each of 18 events and circumstances. Faced with scenarios that commonly occur in Malawi (food shortage, death of a parent, relationship instability), would her preference for the number of children she stated earlier increase, decrease, or stay the same? Would the event alter her desired timing (sooner, later, no change)? Questions in the flexibility module (see Appendix) were asked during the 2009 baseline survey and again in 2015.1 life-course markers (age, marital status, parity); socio-demographic factors including urban/rural residence, household wealth (following closely from the DHS household goods index—Rutstein and Johnson 2004), and educational attainment; the calculus of conscious choice, proxied in three ways: i) a direct measure of numeracy (wherein respondents were asked to solve simple math problems), ii) a single-item measure of planning for the future (“How often, if at all, do you think about or plan for your future?” with Likert-style responses), and iii) a binary indicator of agreement with the statement “You don't plan on having children, they just happen”; and measures designed to operationalize existential uncertainty. The first distinguishes women who report in their pregnancy and childbearing histories ever having experienced a miscarriage or the death of a child. The second gauges experience of death in the respondent's own network by indexing the number of funerals she attended in the past month. The third gauges each respondent's own sense of mortality using an interactive solicitation method in which respondents are given a pile of 10 beans and asked to shift from one plate to another the number of beans representing the likelihood a given event will occur within a specified time frame. Ten beans indicate absolute certainty the event will occur, zero beans absolute certainty it will not, and five beans a 50-50 chance. We measure the perceived likelihood of imminent HIV infection using the prompt: “Pick the number of beans that reflects how likely it is that you will die within a 1-year period beginning today.” This technique has been used successfully in a variety of cultural contexts to generate assessments of child mortality, HIV prevalence, food shortages, and adult mortality (Delavande, Giné, and McKenzie 2011; Delavande and Kohler 2009; Delavande and Rohwedder 2011; Trinitapoli and Yeatman 2011). Third, we assess the extent to which relevant preferences, behaviors, and outcomes vary by level of flexibility. We start by estimating the relationship between flexibility at baseline and observed instability in IFS over the subsequent seven waves. We further test whether flexibility affects fertility-related behaviors (use of modern contraception) and outcomes (pregnancy and surprise pregnancy). Fourth, we re-examine flexibility in this same cohort of women six years later. Combining a dynamic view of flexibility with a dynamic view of fertility preferences allows us to adjudicate between trait-based explanations and perspectives on flexibility that view it as part of a developmental trajectory, responsive to context, and, perhaps, a more temporary state. Table 2 presents detailed information on the diversity of conditions under which respondents indicated in 2009 whether and how they would adjust their fertility preferences. We categorized our conditions into three domains: AIDS-related factors (the first group in the table), economic factors, and family factors.2 On average, respondents indicated movement in fertility preferences on six of the 18 conditions for both their desired number of children and the desired timing of pregnancies. While 10.5 percent of young women in Balaka reported no movement in their numeric preferences for any of the conditions, 3.5 percent anticipated a change for every one of the conditions presented to them (see Figure 1). With respect to timing preferences, 14.5 percent reported no movement, and 4.7 percent anticipated movement in response to every condition we inquired about. Level of flexibility (range 0–36) among women in Balaka, 2009 NOTE: Figure shows quintiles from most fixed to most flexible. N=1,505 women. SOURCE: Tsogolo la Thanzi, Wave 1 For 13 conditions in Table 2, including all of the economic conditions (e.g., winning the lottery, new policies to make the education of children more affordable, a steep rise in the price of food) and most conditions related to family crises such as the illness or death of a parent, less than 50 percent of the sample reported that they would respond by changing their fertility preferences—either in number or timing. Our data provide no evidence of strong sex preferences of offspring, although more than one-fifth of women expressed a clear desire for a mixed-sex household by stating they would continue having more children if they had only boys or only girls.3 A few conditions do, however, appear to incite change for a majority of the TLT sample. Of the five conditions that elicit numeric and timing changes for the majority, four are AIDS-related — two explicitly (suspecting AIDS for yourself or your partner due to weight loss) and two for which AIDS is strongly implied (hearing rumors of a partner's unfaithfulness and fostering children following a sibling's death). The other most consequential condition is one's partner wanting fewer children.4 With respect to timing preferences, AIDS-related conditions are the only conditions that lead a sizable minority of respondents to say they would accelerate their childbearing; all other conditions tend to elicit delays. While 55 percent of women report that they would want fewer children if they heard rumors of an unfaithful partner, 25 percent say that they would have their children sooner—presumably as a strategy for maintaining the relationship or having children before becoming infected with HIV themselves (Hayford, Agadjanian, and Luz 2012; Trinitapoli and Yeatman 2011).5 To operationalize flexibility, we distinguished respondents who list “No change” (coded 0) from those who indicate a likely change (more/fewer or sooner/later, coded 1) for each item, without consideration to direction of change. For each respondent, we then summed responses to all 36 conditions to capture her level of flexibility in a simple additive scale. The analyses that follow center primarily on this scale, which ranges from 0 to 36; occasionally we employ alternate specifications, such as a measure examining numeric flexibility only (0–18) or a distilled measure in which the total flexibility score is converted into quintiles (as in Figure 1). In 2009, over 90 percent of the TLT sample indicated some flexibility, with almost 20 percent expressing more than 20 likely changes in response to the 36 hypothetical conditions. This finding raises questions about the ways in which flexibility is patterned. Is flexibility level tied to life-course steps—chronological and/or social age? Is it patterned socio-demographically, along the same lines of disadvantage we observe for other fertility-related outcomes?

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call