A moveable feast? The flexibility of fertility preferences in a transitioning Malawian community
Jenny Trinitapoli Penn State University
Sara Yeatman
University of Colorado at Denver
Hannah Furnas Penn State University
A moveable feast? The flexibility of fertility preferences in a transitioning Malawian community Abstract: Recent studies suggest a rapid change in fertility preferences among young adults across sub-‐Saharan Africa. In this study, we examine the sensitivity of Malawians’ fertility preferences to a variety of hypothetical (but common) events that may alter fertility preferences and intentions. Using new data from the Tsogolo la Thanzi (TLT) study in southern Malawi, we analyze expected changes in desired number of children (quantum) and the pace of childbearing (tempo) in response to this variety of events. We further employ the Coombs scale, a measure of underlying family size preferences, to predict the direction of both dimensions of fertility preferences. To measure tempo change, the survey questions respondents about their preferred timing to next birth. We find 1) that both the quantum and tempo dimensions of fertility preferences are most responsive to AIDS-‐related conditions and 2) that young adults’ preferences are relatively impervious to changing economic conditions and family issues. Our results indicate that the generalized AIDS epidemic in Malawi is critically important for understanding young people’s fertility preferences and, ultimately, their behaviors. Key Words: Fertility preferences; HIV/AIDS; Malawi; Coombs scale; sub-‐Saharan Africa
A moveable feast? The flexibility of fertility preferences in a transitioning Malawian community
Family size preferences are changing rapidly in the peri-‐urban areas of SSA, a setting characterized by the recent expansion of mass education, heavy out-‐migration to regional centers like Johannesburg and Accra, and the proliferation of new media, including cell phones, DSTV, and wireless internet access. The demographic literature has a long tradition of examining family size preferences as predictors of subsequent fertility and a sensitive measure of changing norms, but most studies are constrained by the questions asked on the Demographic and Health Surveys. Consequently, little has been done to examine the stability or flexibility of these preferences and how this malleability is associated with subsequent behavior. This study represents both the revival of an important but underutilized demographic tool, the Coombs Scale, and the presentation of a true innovation for examining the malleability of fertility preferences cross-‐sectionally by focusing on the factors that young adults believe would trigger tempo changes. We use new data from Southern Malawi designed to assess the conditions under which respondents believe their preference will move, in which directions, and whether or not this move would be temporary (a tempo effect) or permanent (an absolute change.) We address these questions in a setting characterized by high levels of uncertainty surrounding ideal family size.
First, after decades of high and stable fertility rates in rural Africa, new evidence suggests that a fertility transition is underway – particularly in urban and peri-‐urban areas. Numbers from Malawi’s 2004 DHS demonstrate a TFR of 6.4 in rural areas and 4.2 in urban areas. In contrast, ideal family size is considerably lower for men and women in each group: 4.2 rural women; 3.4 urban women; 4.1 rural men; 3.5 urban men (MDHS 2004). Clearly, ideals about and expectations for family life are changing in Malawi, and our study will aid in our understanding of why fertility hasn't fallen in line with fertility preferences. Second, this study considers how Malawi’s generalized HIV epidemic is shaping how individuals think about their futures across multiple dimensions, including fertility (Yeatman 2009, Yeatman Forthcoming, Hoffman et al. 2008). Early in the epidemic, when its societal consequences were not as extensive as now, there was probably little deliberate attempt to limit pregnancies in response to AIDS (Setel 1995): a study using the 1998 and 2001 rounds of the MDICP survey found few differences in actual fertility based on concern about HIV alone (Noel-‐Miller 2003). However, we expect that changes in preferences and behaviors have become increasingly important both for those who suspect or know that they are HIV positive and those who fear infection in the future. In this study, our focus is on the volitional responses to AIDS and the extent to which concern about AIDS engenders greater flexibility about one’s reproductive future. The Coombs scale for measuring fertility preferences was once a well-‐known and widely taught method for assessing preferences about ideal family size. Rather than relying on a one-‐shot question for ascertaining what is often a movable and highly contingent ideal, the scale developed by Lolagene Coombs uses a series of three questions to assign the respondent a scale value that has demonstrably more predictive power for actual
reproductive behavior than single-‐value statements about the number of children wanted (Coombs et al. 1975, Coombs 1974).
However, despite its notoriety, the Coombs scale has only seldom been implemented in the collection of large-‐scale survey data. Our analysis of Coombs’s most heavily cited articles reveal surprising limited use of this important tool. Most notably, the Chitwan Valley Family Study (CVFS) conducted in 1996 in Nepal utilized this method to aid our understanding of changing fertility preferences in a context transitioning from high to low fertility (Biddlecom et al. 2005, Pearce 2002) . However aside from the CVFS, the Coombs scale has been implemented almost exclusively in East Asia, generally with a narrow focus on sex-‐preferences . Indeed, the only study in sub-‐Saharan Africa (SSA) to have used the Coombs scale considered the sex preferences of men in Tanzania (Mwageni et al. 2001).
DATA The data for the study come from Tsogolo la Thanzi (TLT), a longitudinal study in Balaka, Malawi designed to study how young people navigate reproduction in an AIDS epidemic. The first wave of data collection took place between May and August 2009. Fifteen hundred female respondents were randomly selected from a sampling frame of 15 to 24 year olds living in census enumeration areas within 7 kilometers of Balaka, Malawi. The catchment area includes a mix of rural and peri-‐urban communities around Balaka, a growing town one and a half hours from the southern city of Blantyre.
One particularly unique feature of TLT is the use of a centrally located research center for conducting interviews. Respondents were first contacted in their homes and asked to set up a time for an interview. On their assigned day (or more accurately close to it), respondents came to the research center and were interviewed in a private room where their responses could not be overheard. The survey took approximately one and a half hours to complete. Refusal at the time of making an appointment and passive refusal by not showing up at the research center were relatively rare (our current estimate is <5%). Quantum: Coombs Scale TLT measured total desired fertility using the Coombs scale. Rather than using a measure of the number of children a respondent reports as his or her ideal, the Coombs scale value represents each respondent’s position on a continuum, distinguishing between respondents who report the same ideal family size but whose actual desired preferences have subtle distinctions. Respondents’ position on the scale is determined based on their second and third choices in addition to the first articulated ideal family size.
Figure 1: Coombs Scale
Tempo: Timing of Next Child The timing of childbearing is an important for understanding how fertility preferences are manifest. We measure timing to next birth with the following question asked of all respondents: “How long would you like to wait before having your first/next child?” Responses ranged from: “As soon as possible” (0) to “Five or more years,” (5). Flexibility of Quantum and Tempo Preferences After the administration of the Coombs scale, each respondent was asked a set of 18 items to tap the malleability of their fertility preferences. In other words, faced with significant events that commonly occur in Malawi (e.g., food shortage, death of a parent, relationship instability), would their preference for number of children increase, decrease or stay constant? 1In order to differentiate tempo from absolute changes, we also asked about whether or not such events would alter the timing of their childbearing. Socio-‐Demographic Factors & Controls We control for key socio-‐demographic factors that the previous literature has established as associated with ideal family size: gender, age, sibship size, marital status, and parity (number of living children). We measure education in years of schooling completed. To measure socio-‐economic status, we created an index of household goods that ranges from 0 to 11.2 PRELIMINARY RESULTS On average, respondents report 3.3 children as their ideal family size, a number that is slightly higher for men than for women. The distribution by gender (see Figure 3) also shows that more men than women (3 percentage points) report wanting 4 children, while an ideal of 2 children is more common among women. The average Coombs scale score is 8.77. A position of 8 on the Coombs scale is assigned to a respondent who indicates 3 children as the ideal number, moves down to 2 when forced to choose between 2 and 4, but then moves back up to 4 when forced to choose between a 1 or a 4 child family. A 9 is assigned to respondents who first indicate 3, move up to 4 and then back down to 2 when choosing between 2 and 5 children (see Figure 1). At the bivariate level, ideal family size is
1 Table 3 presents the questions in detail.
comparable for these three groups, but they differ in their reported time to next birth, with those certain they are positive reporting a desire to wait the shortest amount of time for their next child (2.6 years), those who are certain they are negative wanting to wait longer (3.6), and the uncertain almost exactly in-‐between. Table 1 presents the diversity of conditions under which respondents indicated whether or not they would adjust their fertility preferences. The conditions are listed in the order of least to most susceptible to change. While we list all of the responses here, we focus our discussion on the conditions to which more than half report some change. More than 70 percent of our sample reports no quantum or tempo change for the first 11 conditions, which include all of the economic conditions (e.g., winning the lottery, new policies to make the education of children more affordable) and most conditions related to family crises such as the illness or death of a parent and the illness of a young child. Our data provide no evidence of strong sex preferences for, though 20 percent of respondents clearly express their desires for a mixed sex household by saying they would have more children if they had only boy or only girl children (these two measures are highly correlated at .86). On average, respondents indicated movement in fertility preferences on 6 of the 18 conditions presented to them. While 11 percent of young people in Balaka report no movement in their preferences in any of the conditions, 3 percent reported some change for every one of the conditions presented to them (not shown). Of the conditions that elicit quantum changes in preferences for the majority of our sample, four of the five are AIDS-‐related. Fostering nieces and nephews, hearing rumors of a partner’s unfaithfulness, and self or partner losing weight (suspicion of AIDS) are associated with a shift towards fewer children for the majority of young people in Balaka. The only non-‐AIDS-‐related condition that strongly influences preferences is having a partner who wants fewer children. Young people in Balaka are more likely to believe their partners’ desires could decrease their family size than increase it, revealing what we consider to be a consensus around the “lowest common denominator” as a solution to discordant fertility preferences within couples.3
In examining how conditions would alter the tempo of childbearing, we see very similar patterns as we do for quantum preferences but with some notable exceptions. Most respondents report no change in the timing of their families; only the AIDS-‐related conditions elicit changes in the timing of children from a majority of respondents. Nearly 40 percent report that they would accelerate their childbearing if they suspected that either they or their partner was infected, and 31 percent report that they would have children sooner if they suspected their spouse of infidelity. The latter result may suggest that childbearing is seen as a strategy for maintaining a family in this context. 55 percent of those who report wanting fewer for this reason (unfaithful partner) say would have them sooner: assuming their health is currently intact, they perceive themselves at high risk of infection and anticipate declines in health that will compromise childbearing. Figures 2 and 3 illustrate the quantum and tempo flexibility, respectively, of preferences by age across three types of conditions: economic changes, family changes, and AIDS-‐related issues. The
3 Across conditions, the preferences of women are less fixed than those of men; however for all the conditions we asked about the general pattern was the same, and we find high levels of consistency in the ordering of these conditions by men and women.
figures both clearly show that our respondents anticipate their preferences being far more influenced by AIDS-‐related concerns than by other common experiences and worries.
Subsequent analyses in advance of the IUSSP conference will include more rigorous analyses that include both respondents’ characteristics and particular conditions to predict quantum and tempo flexibility in preferences, using the Coombs scale scores, which help tap underlying preferences, to validate the kinds of movements respondents anticipate in their childbearing in response to both adverse and positive events that may occur in their lives.
TABLE 1: Flexibility of Fertility Preferences and Timing of Childbearing NUMBER TIMING
More Fewer No Change Sooner Later No Change Ill Mother(in-‐law) 1.31 10.13 88.57 4.84 16.52 78.63 Free Secondary School 6.35 6.59 87.06 5.04 16.67 78.29 Primary Uniforms Materials 5.38 7.56 87.06 4.31 17.54 78.15 Mother(in-‐law) Passes Away 2.08 13.09 84.83 7.56 14.49 77.96 Win Lottery 6.69 8.39 84.92 5.67 17.75 76.58 (Woman) Steady Job 8.77 8.14 83.09 5.14 19.24 75.62 (Man) Steady Job 9.64 7.95 82.41 4.94 20.88 74.18 Ill Youngest Child 4.51 15.47 80.02 6.69 21.33 71.98 Anticipating Maize Shortage 0.19 26.19 73.62 14.70 18.53 66.76 Only Girl Children 23.16 7.22 69.62 6.59 24.67 68.73 Only Boy Children 22.98 7.61 69.41 6.11 25.59 68.30 Partner Wants More 29.16 6.55 64.29 7.43 28.79 63.79 (Male) Partner to RSA 2.13 37.50 60.37 13.91 33.19 52.91 3 Sister(in-‐law) 's Children 3.05 50.53 46.41 20.30 29.46 50.24 Rumors Partner Unfaithful 0.58 55.03 44.39 30.56 26.72 42.71 Partner Wants Fewer 1.60 59.01 39.39 18.56 28.59 52.86 Losing Weight AIDS 0.05 64.97 34.98 38.30 24.03 37.67 Partner Losing Weight 0.05 66.12 33.83 39.65 23.17 37.18 N=2064
Figure 1.
Figure 2.
0.2
.4.6
.81
16 18 20 22 24age
ECONOMIC REASONS
0.2
.4.6
.81
16 18 20 22 24age
FAMILY REASONS
0.2
.4.6
.81
16 18 20 22 24age
AIDS-RELATED REASONS
SOURCE: TLT, W1, 2009
Quantum Flexibility of Fertility Preferences By Condition and Age0
.2.4
.6.8
1
16 18 20 22 24age
ECONOMIC REASONS
0.2
.4.6
.81
16 18 20 22 24age
FAMILY REASONS
0.2
.4.6
.81
16 18 20 22 24age
AIDS-RELATED REASONS
SOURCE: TLT, W1, 2009
Tempo Flexibility of Fertility Preferences By Condition and Age