Women’s Life Trajectories in Rural Timor-Leste: A Life History and Life Course Perspective on Reproduction and Empowerment

Abstract

Women’s reproductive decisions and life trajectories are shaped by an interplay of biological, social, and ecological factors. While Life History Theory (LHT) has traditionally been applied in biological sciences to examine reproductive trade-offs, its integration with Life Course Theory (LCT) and empowerment frameworks offers a novel approach to understanding how structural and environmental conditions shape women’s reproductive behaviours and household roles. This study applies Categorical Principal Component Analysis (CATPCA) to identify key profiles of women’s lives in two ecologically distinct rural communities in Timor-Leste—Ossu and Natarbora—and examines how these patterns relate to early life conditions. Building on a longitudinal study conducted in these communities, our findings reveal four distinct profiles: (1) Tech and Sanitation, linked to household labour-saving technology and higher education; (2) Traditional, reflecting large household size and livestock ownership; (3) Contraception, associated with fertility control, particularly among younger cohorts; and (4) High Fertility, characterised by more births, greater child mortality, and being born in high-altitude regions. By combining LHT, LCT, and the empowerment framework, this study analyses how reproductive strategies and household ecology intersect with structural inequalities. These findings offer key insights for policies aimed at improving women’s autonomy, access to resources, and reproductive health in rural Timor-Leste.

1. Introduction

1.1. Timor-Leste

Timor-Leste is a country located in the Sunda archipelago of Southeast Asia between the Bandar and Timor Seas. The area that is now Timor-Leste was colonised by the Portuguese in the 16th century and declared independence in 1975 (Durand 2011). It was occupied by the Indonesian army from 1975 until 1999, and achieved formal independence in 2002 (Durand 2011). Violence was common during the Indonesian occupation, with an estimated 20–25% of Timorese dying (Durand 2011) and women and children being exposed to family and community conflict and trauma during those decades (Silove et al. 2014; The Asia Foundation 2016).

As a young country, Timor-Leste still faces challenges to development. Two thirds of its 1.4 million population live on less than USD 2 a day (Central Itelligence Agency 2023; Department of Foreign Affairs and Trade 2019). As of 2023, 67.5% of its population reside in rural areas (Central Itelligence Agency 2023), and most of the country’s population have rural livelihoods that depend on seasonal subsistence agriculture (Da Costa et al. 2013; Lopes and Nesbitt 2012). Subsistence agriculture is mostly done either via small garden plots (kintal), larger gardens with a variety of crops (to’os) or rice paddies (natar) (Thu and Judge 2017). Animal protein is scarce since livestock are used mainly as an income source for larger expenses and in ritual exchange practice (AMSAT International 2011).

Women’s livelihoods in rural Timor-Leste are challenging. Women and girls undertake labour-intensive tasks such as carrying water and wood, farming in the garden plot, and preparing meals for household members (Seeds of Life 2007). In addition, women in rural areas (when compared to urban regions) have less access to education and health care, and experience poorer health (Ministry of Health 2018). These challenges overlap with their already vulnerable positions in income earning possibilities, land ownership, and decision-making power (Narciso and Henriques 2020).

The challenges faced by Timorese women stem from a longstanding interaction between customary traditions, colonial rule, and post-conflict experiences, shaping gender norms and reproductive roles. Traditional gender structures position women as primary caregivers and household managers, with decision-making power often mediated by kinship systems and patriarchal norms (Loney 2015; Niner and Loney 2020). The Portuguese colonial administration reinforced male authority through taxation and governance structures, restricting women’s access to formal education and economic autonomy (Loney 2015). This “double exploitation”—first under colonial rule and later within customary systems—persisted into post-independence Timor-Leste, where sociopolitical pressures continue to maintain patriarchal norms (Loney 2015).

Gender roles were further reshaped by conflict. During the occupation, Timorese women actively participated in the nationalist struggle, taking on roles in armed resistance and clandestine networks (Niner and Loney 2020). The Popular Organisation of Timorese Women (Organização Popular da Mulher Timorense—OPMT) framed women’s liberation as intertwined with national independence, expanding their responsibilities beyond traditional domestic labour. However, despite their contributions, post-independence Timor-Leste remained dominated by a militarised male elite, limiting women’s political influence and broader gender equality reforms (Niner and Loney 2020).

In addition to gender-based inequalities, women endured severe violations of their sexual and reproductive rights during the occupation. The Commission for Reception, Truth, and Reconciliation in East Timor (CAVR) documented systematic sexual violence perpetrated by Indonesian security forces, including rape, sexual slavery, and torture (Commission for Reception Truth and Reconciliation Timor-Leste 2005; Wayte et al. 2008). Women were also subjected to forced contraception and sterilisation under Indonesia’s national family planning program, often without informed consent, as part of a broader strategy to control indigenous population growth (Commission for Reception Truth and Reconciliation Timor-Leste 2005). These abuses resulted in physical and psychological trauma, significantly affecting women’s reproductive health and their capacity to care for children (The Asia Foundation 2016). The legacy of this violence persists, contributing to ongoing maternal health challenges. As of 2016, Timor-Leste reported a maternal mortality rate of 195 per 100,000 live births, with only 57% of births attended by skilled health personnel (United Nations Population Fund 2024). Customary beliefs continue to influence women’s reproductive autonomy, as male relatives often hold decision-making power over family planning (Tribess et al. 2015). In response, the Timorese government launched a national family planning campaign in October 2024, promoting a three-year birth spacing strategy to reduce maternal and infant mortality rates (United Nations Population Fund 2024).

Despite these historical and structural barriers, contemporary women’s movements, local NGOs, and activists have successfully advocated for gender-equity policies, including the Law Against Domestic Violence and national gender action plans (Freitas Belo 2015; Niner and Loney 2020). These ongoing efforts highlight the persistent tension between deeply rooted gender norms and emerging opportunities for women’s empowerment, reflecting both progress and continued challenges in achieving gender equality in Timor-Leste.

Women’s life patterns and their levels of empowerment are key to improving family welfare (Bonis-Profumo et al. 2021; Sell and Minot 2018; Sharaunga et al. 2016) and to realising the government campaign “feto forte, nasaun forte” (strong women strong nation). Analysing women’s livelihoods in Timor-Leste is important to understanding the factors that may foster the wellbeing of women and their families.

In this paper, we build upon our ongoing longitudinal study of family ecology and children’s growth (Spencer 2018) in the rural communities of Ossu and Natarbora, in Timor-Leste to examine how women’s reproductive outcomes, empowerment indicators, and household ecology intersect. By identifying patterns within these domains, we aim to provide insights into how socio-ecological and historical factors shape women’s life trajectories and opportunities for empowerment.

Ossu is a mountain subdistrict of the municipality of Viqueque located 600 to 1000 m above sea level. Natarbora is a subdistrict of the municipality of Manatuto located 5 to 50 m above sea level on the south coastal plain of Timor-Leste. The research team visits both sites on a six-month to one-year basis since 2009 in Ossu, and from 2012 onwards in Natarbora. Both subdistricts include a local health clinic, weekly markets, schools, and have subsistence agriculture as their main economic activity (Thu and Judge 2017). However, they differ in their social affiliation practices and in mother-tongue language. In Ossu, social affiliation is patrilineal (male descendants can inherit the family’s assets) and the predominant languages are Makassae and Kai Rui. In Natarbora, social affiliation follows a matrilineal/bilineal line and Tetun-Terik language predominates (Thu and Judge 2017). See Thu and Judge (2017) for a more detailed description of both sites.

We used Categorical Principal Component Analysis to objectively assess how women’s reproductive outcomes, indirect indicators of empowerment, and household ecology co-occur and cluster. We then compared these clusters (profiles) across the two sites to determine whether they are specific to each locale or consistent across both rural communities. We assessed the extent to which women’s profiles were related to their early life conditions as indexed by birth cohort and birthplace. Finally, we analysed the patterns of women’s profiles using Life History Theory (LHT), Life Course Theory (LCT) and the women’s empowerment framework. In the next section, we present a detailed overview of LHT and LCT and their application in studying women’s life trajectories.

1.2. Life History Theory and Life Course Theory

LHT and LCT provide two useful theoretical frameworks to analyse women’s life trajectories. LHT stems from evolutionary biology and focuses on the allocation of resources during an organism’s life history so as to maximise fitness (Hill and Kaplan 1999). This framework facilitates comparing life history strategies across different types of organisms including humans and across humans under various ecologies (Stulp and Sear 2019). In contrast, LCT originated in the social sciences, and seeks to analyse the mechanisms over an individual life course that explain the association between developmental experiences and later life outcomes (Sampson and Laub 1992). Both terms, life history and life course allude to the progression of time and its implications on the changes experienced by an organism (LHT) or a human (LCT) during their lifespan.

LHT examines the allocation of resources and energy into three main organismal functions: growth, maintenance, and reproduction (Hill and Kaplan 1999). In human populations, resources can be material—such as food, water, wealth (e.g., income, land, and other assets)—or social, such as cooperation (e.g., altruism, alliances) (Chisholm 1999). Resources are limited; investing more in one function means investing less in another, resulting in trade-offs (Kaplan et al. 2000). The prioritisation of resource allocation across functions is contingent on the characteristics of a specific local environment (ecology) and, as a result, resource allocation strategies have inherent plasticity (Kuzawa and Bragg 2012).

In LHT, resource allocation strategies in human populations are unconscious and prioritise fitness over well-being (Stulp and Sear 2019). Herein, well-being refers to the objective and material indicators of quality of life such as having income, housing, education, health, and social networks amongst other attributes (Diener and Suh 1997; Western and Tomaszewski 2016). In contrast, LCT incorporates the ability of individuals to make continuous conscious decisions (agency) that maximise their well-being over their lifespan (Stulp and Sear 2019).

Bernardi et al. (2019) suggest that in addition to conscious decisions, individuals also adopt “spontaneous behaviours” that emerge in response to historical and environmental factors, and both types of behaviours influence individual life trajectories. Moreover, they describe the life course as a “multifaceted process of individual behaviour” that shapes individuals’ biographical states (Bernardi et al. 2019). They define biographical states as the result of an individual’s attributes across three levels: inner-individual, individual, and supra-individual. They argue that biological, genetic, and psychological characteristics comprise the inner-individual level, while the individual level consists of clear conscious processes that are outcomes of individuals’ conscious behaviour over time and are influenced, for example, by education, social status, and resources. The supra-individual level relates to the socio-cultural environments where behaviour occurs, also influencing individuals’ decisions (Bernardi et al. 2019).

Employing LHT and LCT to analyse the life trajectories of individuals is beneficial for understanding the multiple layers in which an individuals’ life pattern is embedded. For example, the idea proposed by LCT theorists Bernardi et al. (2019) that individuals’ conscious and spontaneous behaviours are affected by the environment experienced by the individual, resembles the notion of human populations acclimating to different ecologies over a multigenerational timeframe as explored through LHT. As displayed in Figure 1, biological, genetical and psychological factors (inner-individual characteristics) may translate into spontaneous behaviours and influence agency (individual level characteristics), which are also affected by the historical and socio-cultural characteristics of the environment (supra-individual level). Unconscious resource allocation strategies that occur at both inner-individual and individual levels, are influenced by the levels of mortality and stability of the environment (LHT, Hill and Kaplan 1999). Moreover, mortality and environmental instability may be a product of the socio-cultural environment and historical context of a particular community.

LHT and LCT are useful for understanding women’s reproduction and empowerment as complex processes that take place within the context of household ecology, influenced by various biological, social, and cultural factors across different stages of life.

Women’s reproductive life stage starts with menarche and ends with menopause (Mishra et al. 2010). Theorists of both frameworks—LHT and LCT—analyse the effects of the environment (as referred to in LHT) or the socio-historical context (as defined by LCT) on women’s reproductive traits throughout their lives and across generations. The mechanisms through which both theories explore the interplay between the environment (or socio-historical context) and women’s reproductive outcomes are similar, but they also possess clear differences. LHT focuses on how evolutionary pressures and resource allocation strategies influence reproductive behaviours, while LCT emphasizes the role of socio-historical contexts and individual agency in shaping reproductive outcomes.

In LHT, reproduction is one of the three primary functions that contribute to fitness (Stearns 1989). To maximize fitness, women must produce offspring that not only survive to reproductive age but who also reproduce (Hill and Kaplan 1999). Because functions require resources and energy and these are finite, women face two reproductive trade-offs: Current vs. future reproduction and Quantity vs. quality of offspring (Hill and Kaplan 1999). Current vs. future arises when a woman either (1) reproduces earlier thus increasing the likelihood of having more offspring, but reduces investment in her own growth and maintenance, or (2) delays reproduction, allowing for greater investment in growth and maintenance at the risk of having fewer children due to the limited reproductive lifespan and potential for pre-reproductive mortality (Hill and Kaplan 1999; Trivers 1972). Quantity vs. quality trade-offs occur when a woman either (1) has more children, raising the odds of leaving at least one surviving offspring but with diminished resource investment per child or (2) has fewer offspring, enabling greater resource investment per child and increasing the likelihood of their offspring’s reproduction (Hill and Kaplan 1999; Trivers 1972). Shorter interbirth intervals (IBI) (period between sequential live births by the same mother) can indicate low parental investment since it limits women’s allocation in each offspring prior to the arrival of a subsequent (and competing) offspring, whereas longer IBIs may suggest higher parental investment per birth (Lawson and Mace 2011).

In LHT, variations in reproductive timing and levels of parental investment can be analysed relative to the local environmental conditions. For example, when local mortality is high, women tend to have earlier age at first birth, short IBIs and high fertility rates, thus maximising the chance of leaving any surviving offspring (Chisholm 1999). Conversely, when more children survive (indicative of low local mortality), competition for resources increases when these children reach adulthood, thus women might adopt longer IBIs facilitating higher parental investment per child, raising the odds of survival of their offspring until adulthood and their competitive ability at adulthood (Chisholm 1993; Chisholm 1999; Sellen and Mace 1997).

The relationship between environmental factors and reproduction is complex, as the same environmental factors can sometimes accelerate reproductive timing and, in other instances, slow it down. For example, some harsh environments in the form of high loads of disease or malnutrition associate with delayed reproduction (Chisholm 1999). This delay may result from an adaptative response by women to postpone reproduction until conditions improve (Chisholm 1999) or due to reduced growth rates caused by limited resources. Sear et al. (2019) argue that most of the studies supporting the hypothesis that early life adversity promotes accelerated reproduction were conducted using samples of people living in western, educated, industrialized, rich and democratic (WEIRD) countries (Henrich et al. 2010). Moreover, they suggest that evidence in non-WEIRD settings may not consistently support this hypothesis (Sear et al. 2019). For example Odongkara Mpora et al. (2014) observe that in post-conflict northern Uganda, conflict-related, early life stressful events did not associate with an early onset of menarche, but higher paternal education (but not maternal) did, as did living in urban areas. In southwestern Nigeria, Sabageh et al. (2015) report that a higher socio-economic status is associated with an earlier onset of puberty in girls as measured by the Modified Sexual Maturity Scale (Sabageh et al. 2015). Kyweluk et al. (2018) also find that early life adversity or familial instability (measured as paternal or maternal absence, or death of a sibling) does not predict early menarche in the Philippines, but better growth and nutrition during childhood do predict early maturity (Kyweluk et al. 2018). Similarly, in Indonesia, having electricity, piped water, a toilet with a septic tank and good health during childhood correlated with earlier menarche (Sohn 2017). These studies demonstrate that improved material resources (i.e., nutrition, indicators of wealth) predict earlier reproductive timings in most cases, whereas the effects of psychosocial factors (i.e., familial instability) on reproduction appear to vary depending on the different socio-cultural characteristics of population studied.

LCT considers women’s reproduction from an epidemiological perspective (Mishra et al. 2010). Reproductive outcomes throughout women’s lives are assessed concerning the independent, cumulative and interactive effect of biological, behavioural, and social factors (herein referred to as exposures) (Mishra et al. 2010). For instance, researchers may examine the effect of birthweight on age at menarche and menopause, analysing these effects in relation to women’s health and well-being while considering mental, social, and cultural factors (Mishra et al. 2010, 2023). In contrast, LHT researchers focus on the effects of women’s reproductive traits on fitness and resource allocation strategies, relating these traits to environmental conditions (Hill and Kaplan 1999; Stearns 1989). Thus, while both LCT and LHT operationalise reproductive traits similarly, LCT focuses on health and well-being (contemporary outcomes), and LHT focuses on resource strategies relative to evolutionary fitness.

There are mainly four theoretical models used in life course studies to analyse women’s reproductive events: the critical period model, the critical period model with later effect modifiers, the accumulation of risk model and the model of chains of risk (Mishra et al. 2010). The critical period model (also known as biological programming) suggests that exposures occurring during an individual’s vulnerable period of life may have irreversible long-lasting impacts on later health (Ben-Shlomo and Kuh 2002). The critical period with later effect modifiers model acknowledges that later-life exposures can modify or amplify the effects of earlier critical period exposures (Ben-Shlomo and Kuh 2002). The accumulation of risk model posits that the accumulation of multiple exposures throughout the life course increases the risk of later adverse health outcomes (Mishra et al. 2010). Ben-Shlomo and Kuh (2002) suggest that the accumulation of risk model is similar to the concept of allostatic load (Seeman et al. 2001) in which chronic exposures to stress, and the body’s efforts to maintain stability (homeostasis) in response to stressors, have long-term health consequences (Seeman et al. 2001). The model of chains of risk proposes that risk factors are linked in a chain-like sequence leading to a health outcome (Ben-Shlomo and Kuh 2002). This model differs from the accumulation of risk in that it emphasises a sequential pathway where each risk leads to the next. These theoretical models are not independent of each other and as such, their impacts can overlap (Mishra et al. 2010).

In summary, the analyses of women’s reproduction through both the LHT and LCT frameworks provides a comprehensive understanding of the multifaceted nature of reproductive experiences. While LHT offers insights into the intrinsic biological processes and evolutionary foundations of women’s reproductive traits, LCT illustrates the complex interplay between biological processes and socio-cultural contexts. By considering both frameworks, we can analyse the influence of evolutionary pressures and contemporary social environments on women’s reproductive traits and conscious choices, and their broader implications for health and well-being.

1.3. Household Ecology and Women’s Empowerment

The household is the most immediate environment to which a woman is exposed; therefore, household characteristics should be considered when analysing women’s reproduction and well-being. Becker (1965) defines a household as a unit of production and consumption where members allocate time to market work (activities that generate income), household production (activities that contribute to the well-being of individuals or family), and leisure. Time is limited, thus household members make decisions about time allocation based on the relative costs and benefits of different activities resulting in allocation trade-offs across market work, household production activities and leisure (Becker 1965). Becker (1981) proposed that families share the same preferences and make joint resource allocation decisions to maximise their collective well-being or ‘collective utility’. Alderman et al. (1995) argued that household ‘unitary models’ like Becker’s often overlook the process by which family members distribute resources, whereas ‘collective models’ recognise that household members may have different preferences and bargaining power. Differences in bargaining power may determine the ability of household members to influence resource allocation, which in some cases can lead to intra-household inequality (Alderman et al. 1995; Haddad and Kanbur 1992).

Differences in bargaining power can manifest in various ways; for example, a higher-earning partner may have more influence on income allocation, or societal norms and gender roles may provide men with more power in the decision-making process of distributing resources (Lundberg and Pollak 1993). In subsistence farming societies, women (in comparison to men) have a higher workload of household production activities because they are mostly responsible for food preparation, water and wood collection, and childcare (Grassi et al. 2015), and this workload remains high despite childbearing (Panter-Brick 1989). Increased workload translates into greater time allocation to supply a particular work demand (Becker 1965); hence rural women often do not have time to invest in market-related work, which reduces their access to income and, in some communities, they are not allowed to own land or inherit property (Food and Agriculture Organization 2011; Grassi et al. 2015). High workloads are associated with poor health, and poor health hinders rural women’s ability to supply the workload demand (Sinclair et al. 2019). This phenomenon fosters the cycle of poverty, increasing the risk of food insecurity for women and their household residents (Sinclair et al. 2019).

Some aspects of women’s empowerment can improve their access to resources (Food and Agriculture Organization 2011). Women’s “empowerment” refers to a process that enables women to exercise control and influence over the challenges that are most relevant for them (Kar et al. 1999). It entails the control of resources and the agency to make the choices needed to reach “functional achievements” which involve tangible outcomes such as women’s increased access to education, healthcare, and economic opportunities (Kabeer 1999). These outcomes may be influenced by contextual factors; for instance, in rural Uganda, Sell and Minot (2018) found that women’s education associated with empowerment, but when husbands had a higher educational level, women’s empowerment decreased. The authors suggested that gender education equality might be more effective in promoting women’s empowerment than solely increasing women’s education. In this study, women’s empowerment is framed as the result of “individual, household and community characteristics” (Sell and Minot 2018), arguing that empowerment can affect both the distribution of resources and the welfare of the rest of household members. This view constitutes a similar approach to the one used in the household collective models described by Alderman et al. (1995), where bargaining power can affect the decision-making processes related to resource allocation in the household. These concepts are related in that agency is central to their meaning as it may manifest as negotiation and bargaining (Kabeer 1999).

Women’s empowerment and bargaining power influence reproduction. In contexts where men have more bargaining power, a couple’s number of children is less likely to reflect a woman’s desired fertility (Doepke and Tertilt 2018). Empowerment also has an effect on the length of IBIs and family planning, but the direction can vary, especially in non-western populations. Reviews of women living in developing countries and rural areas (Prata et al. 2017; Upadhyay et al. 2014) find that in some cases, increased empowerment positively associates with longer IBIs and the use of contraception, but in other cases increased empowerment associates with higher fertility. The disparities in these findings can be attributed to the variations in how empowerment is defined and measured, as well as diverse socio-cultural environments (Upadhyay et al. 2014). For example, in rural India women’s autonomy (measured as access to resources) associates with the use of contraception; however, the association is mediated by cultural factors (assessed as region of residency) (Jejeebhoy 2002). In Chad, women who are empowered are more likely to have higher fertility (Atake and Gnakou Ali 2019), possibly because children can potentially contribute to family income (Atake and Gnakou Ali 2019).

In this paper, we examine the interrelation between women’s reproduction, empowerment, and household ecology in Ossu and Natarbora using CATPCA, interpreted through LHT, LCT, and the women’s empowerment framework.

2. Materials and Methods

At the baseline of our ongoing longitudinal study, our research team conducted interviews with the female head of each household who was usually the mother of the resident children (Reghupathy et al. 2012). The team registered the health status of the household residents and collected their sociodemographic information. In addition, they recorded the household’s resources (livestock, crops grown, sanitation, water usage, appliances, house structure, employment status of adult household members, and governmental transfers). Reproductive histories of resident women, the numbers of fostered and biological children, and children’s heights, weights, and upper arm circumferences were also collected (Reghupathy et al. 2012). In later years, information about each household’s diet was collected and, when possible (refer to Table 1), use of family planning by women of reproductive age. It is not culturally appropriate to inquire about this in public or in front of a man and we do not ask about contraception to unmarried, pregnant, or elderly women. Self-reported health status (over the preceding month) of each household resident, household resources, and children’s measures are collected during each visit. Sociodemographic information (date of birth, age and occupation) is updated when applicable.

This paper uses data from women who were 15 years and older in 2018, had at least one birth during their lifetime, and were living in a household that participated in the data collection interviews during that fieldwork year (Ossu n = 116 women, Natarbora n = 140 women), and uses all available raw data collected in previous years through until 2018. For this research, we originally intended to quantify women’s experiences during early childhood via one-on-one interviews, but the 2-year border closure produced by the COVID-19 pandemic made this difficult. Therefore, we use retrospective data collected between 2010 and 2018, drawn from the original longitudinal study, to indirectly assess women’s early childhood environment, reproductive outcomes, empowerment, and other characteristics of their household ecology. Although husbands’ data are relevant to women’s livelihoods, we were unable to incorporate them due to insufficient information for robust statistical analyses.

We include three variables relevant to reproductive history, three for women’s empowerment, eleven for household ecology, and self-rated health, year of birth, and birth region, bringing our total to twenty aspects of women’s lives (Table 1). We first adopt a quantitative method to objectively reduce the number of study variables to sets of correlated groupings (components), within which the categorisation of the variables’ raw data is embedded. Subsequently, in the measures section (Section 2.2), we provide a detailed account of the data source, operationalisation, and the labelling process for each variable of analysis.

2.1. Categorical Principal Components Analysis

We use Categorical Principal Component Analysis (CATPCA) to reveal the profiles of women living in both communities, using the available data on them collected though our ongoing longitudinal study. This methodology has been applied previously in our longitudinal research (Sumich 2021). Principal Component Analysis (PCA) is a statistical method that involves clustering the linear relationships among variables from a particular data set to create independent summary variables (also known as components), which account for the cumulative largest proportion of the variability in the data set (Jolliffe and Cadima 2016). PCA only uses numerical data, whereas CATPCA permits inclusion of categorical variables as well as quantitative variables and does not assume linear relationships between variables (Linting et al. 2007; Linting and van der Kooij 2012). In addition, it allows imputation of the mode for missing values, and to include variables into the analysis without affecting component loadings (Linting and van der Kooij 2012). This method is appropriate for our analysis since our longitudinal study’s information on women includes numerical, ordinal, and nominal variables (Table 1).

CATPCA requires positive integer values for both numerical and categorical data. To avoid recoding raw data manually, the method provides an in-built option to automatically transform the raw values of the variables of interest before the clustering analysis is performed, a process known as discretisation (Linting and van der Kooij 2012). Plotting the quantifications (that is, the transformed values after discretisation) is useful to visualise and amend data to preserve the stability of the CATPCA solution (Linting and van der Kooij 2012). Stability is compromised if sample size within a variable’s categories is below eight cases (Linting and van der Kooij 2012); therefore, we reviewed the distributions of raw categorical data and merged the categories of some variables on theoretical grounds, before running the first iteration of the CATPCA analysis. Next, we combined some categories of other variables in subsequent CATPCA iterations based on the transformation plots after discretisation. Linting and van der Kooij (2012) advise doing this to account for any outstanding quantification generated by categories with low marginal frequencies. Explanations on which categories and variables were modified are included in the measures section.

2.2. Measures

To profile women living in each community, we use their (a) health status, (b) reproductive outcomes such as age at first birth, numbers of births and child deaths, (c) indicators of empowerment like family planning, education and income independent of a partner, (d) household ecology such as subsidies (Bolsa da mae—vulnerable mothers’ pension—and age pension), sanitation, water usage (type of supply in Ossu and source in Natarbora), livestock (cows and pigs), crops, electric appliances, and number of residents, and (e) year of birth and birth region. Health status, age at first birth, family planning, education, income, water source, year of birth and birth region are treated as trait variables, whereas number of births, number of child deaths and household ecology variables are assessed as state variables with data as of 2018 (Table 1).

2.2.1. Health Status

Women were asked whether they had experienced good health or not (‘saude diak ka lae’) in the preceding month. Women’s overall health status ratio is assessed as the ratio of reported good health to total data collection field trips in which a response was recorded.

2.2.2. Reproductive Outcomes: Age at First Birth, Births, and Child Deaths

Age at first birth (AFB), number of births and number of child deaths are used to assess women’s fertility pattern as of 2018. AFB was calculated by subtracting the year of a woman’s own birth from the year of birth of her first child. For some cases, we could only calculate the year of their first birth based on the current age of their oldest child. Cases where the first birth was stillborn, and/or the year of birth is unknown are designated missing. Number of births is a count of a woman’s reported births. Child deaths are operationalised as the sum of reported child deaths and stillbirths as of 2018. Miscarriages are not included due to the lack of updated data on this matter, largely due to the difficulty in accessing the information. Births and child deaths are minima since older women sometimes may not report the child deaths that occurred decades ago.

2.2.3. Indicators of Women’s Empowerment: Family Planning, Education, and Income

Women’s use of family planning, education level and independent income from a partner, are used as indirect indicators of their empowerment. Use of family planning refers to a woman reporting that she has used at least once any form of contraception. This variable measures a woman’s conscious decision to attempt to control her births, regardless of the effectiveness of her contraceptive method of choice. This does not imply that the lack of contraception indicates less agency; rather, its use demonstrates a woman’s clear exercise of agency over her fertility. Women’s use of contraception was first grouped into a variable with five categories (one per each fieldwork year where contraception methods were recorded): (1) never used, (2) used a medical method (at least once), (3) currently using a medical method, (4) using a non-medical method, and (5) previously used a non-medical method. Medical contraception in this context includes injection, intrauterine device, contraceptive implant and surgery. Non-medical contraception refers to rhythm and other types of natural methods. Yearly family planning data was summarised into a dichotomous variable to be included in the CATPCA. The variable comprises report of use of any type of family planning at least once in a woman’s lifetime as of 2018, or no evidence of use.

Education is the highest level of school attended. Women are more eager to report when they have completed some years of schooling, and less likely to report when they have none since it is culturally valued to be formally educated in these communities. In addition, older women had historically less access to formal education (Ministry of Health 2018). Therefore, women’s answers were initially grouped as: no school (includes cases where women did not report their education level at all), some or complete primary, some or complete SMP (lower secondary education), and some SMA/SPP until tertiary (upper secondary education until tertiary). Subsequently, based on the transformation plot (product of discretisation process) of this variable generated during the first iteration of the CATPCA analysis, we merged its categories and created a binary education variable (one per each community): “Under SMP” and “Some SMP to Tertiary” in Ossu, and in Natarbora, “Until complete SMP” and “SMA to Tertiary”.

Income independent of a partner is defined as a woman’s access to cash (excluding government subsidies) via selling goods, providing services, or formal employment. Women’s income, as reported at interview, is coded using ordinal categories defined by women’s level of access to cash: having no income, selling non-animal/agricultural goods, selling animal/agricultural goods, day labour, having a business, and wage earnings. Day labour includes casual services such as sewing or washing clothes. Having a business includes buying and reselling or producing non-animal/agricultural goods or providing a regular service to community members such as being a seamstress or weaving Tais (a traditional cloth made for largely ceremonial purposes). Yearly income variables were summarised into one variable with two categories: evidence of having income at least once over the years or no evidence of having individual income at any point. Having had income at some point is being used as an indicator of women’s intentions of generating income independent of their husbands’ earnings because, in the Timorese context, husbands are usually the primary source of family income (Ministry of Health 2018).

2.2.4. Household Ecology Variables

Household ecology refers to the household’s economy, sanitation, water usage and its amenities. Particularly, these group of variables include subsidies (Bolsa da mae and age pension), sanitation, water supply method (Ossu), type of water source (Natarbora), livestock (pigs and/or cows), having a garden plot (to’os), having labour-reducing and/or non-labour reducing electronics and number of household residents.

• Subsidies. Introduced in 2008, the ‘Bolsa da mae’ is a governmental cash transfer for women who are mothers of school age children living in vulnerable households (Fernandes 2015). Women need to register at local health clinics to receive 5 USD per each child per month for up to three children (Fernandes 2015). Similarly to Bolsa da mae, Timor-Leste’s age pension was legislated in 2008 with the aim of providing 30 USD per month to any Timorese 60 years of age and older (Bongestabs 2016) who did not receive a veterans or employment pension. Both subsidies are included in the analysis as of 2018. Bolsa da mae is operationalised as receiving at least one by participating women (individual level) and age pension as the total number of pensions received by a particular household (household level). Few women receive pensions which are of a higher value (e.g., veteran pensions); therefore, they are not included in the analyses;
• Sanitation. The household’s type of sanitation was recorded in 2018 in 3 levels: having no facility, a traditional (pit latrine), or a developed facility (with an adjacent water tank or mandi). Based on the variable’s transformation plot after discretisation, we recategorized it into two categories for subsequent CATPCA iterations: having no facility and having a traditional or developed facility (Table 1);
• Water usage. Because water usage varies across communities, we use two different assessments: water supply in Ossu and water source in Natarbora. In Ossu, access to water is seasonal; therefore, water supply was initially categorised as spring, tap or pipe (bamboo canals) per each fieldwork year, and then the most common type of water supply across years was used for creating this variable. The variable was later recategorized into a binary variable (spring or tap, and pipe) based on the transformation plot after discretisation. In Natarbora, water source is measured in two categories: superficial (hand drawn from well, hand pump, or river) or deep source (sourced from a tank, an electric pump or piped water). We use water source data as of 2018;
• Livestock. Cattle ownership is assessed as the number of pigs and cows in the household reported by the respondent in 2018. Having pigs is measured as the total number of pigs and having cows as an ordinal variable with 3 levels: zero cows, between 0 to 10 and more than 10. Cows are recorded using these three categories during fieldwork as an estimate due to their extensive distribution in numbers. Respondents usually report “many” when they have high numbers of cows instead of providing an exact number;
• Garden plot (to’os). The existence of a to’os is assessed as a dichotomous variable. If the interviewee indicated that at least one crop was harvested in 2018 from a patch within their place of residence, the household was coded as having a garden plot;
• Electronics. We use two variables based on the effectiveness of electronics in reducing women’s household labour in the context of these rural communities. Labour-reducing electronics include rice cookers, sewing and washing machines, refrigerators, kettles, and mixers. Non-labour reducing electronics include light bulbs, telephone chargers, radios, music devices, televisions, electric saws, computers, irons, fans, photocopiers, motors, telephones, and internet cables. Both variables contribute to the total number of different appliances enumerated during the 2018 household interview. Only one item was counted when there were multiple of a particular category (i.e., light bulbs).
• Household residents. The number of people living in the household represents the total number of adults and children who typically sleep and eat in the household as recorded at interview.

2.2.5. Year of Birth (YOB) and Birth Region

In Timor-Leste, women’s year and place of birth can inform the likelihood of exposure to events related to civil unrest during their childhood years, such as war-related violence, famines, and forced displacement (Silove et al. 2014; The Asia Foundation 2016). We use YOB in both research sites and birth region for women living in Natarbora as indirect indicators of the type of environment to which women were exposed during their development. YOB was self-reported by women verbally and some also showed their identification card. For women who were not able to report their specific YOB, we calculated it using their reported age in the year that they entered the study. YOB was used for the CATPCA as a supplementary variable. For a later stage of the analysis (Section 2.3), we created a birth cohort variable by dichotomising YOB into two categories: born before 1969 or in 1969 and after. This cut-point identified women who experienced any of their first six years of life during the Indonesian occupation (from 1975 to 1999) because of the relationship between the social context in which this developmental stage occurs and later life outcomes (Likhar et al. 2022).

Table 1. Operationalisation of variables used for CATPCA.

Variable	Collection Period	Measurement at Interview	Pre-CATPCA Operationalisation	Recategorization During CATPCA
Health status a	Every visit	Healthy or unhealthy report	Ratio of reported good health to total years with valid data.
Age at first birth b	Baseline	Year of first birth	Subtraction of first birth year minus woman’s year of birth.
Births b	Every visit	No. of births
Child deaths b	Every visit	No. of child deaths and no. of still births.	Sum of child deaths and still births.
Family planning c	Ossu: 2010, 2013, 2014, 2015 and 2018. Natarbora: 2012, 2013, 2014, 2015 and 2018.	Never used. Used medical method at least once. Using medical method Using non-medical method Used non-medical method	Evidence of use of family planning at least once across years or no evidence of use.
Education d	Baseline	None Some or complete primary Some or complete SMP Some SMA/SPP until tertiary		Ossu: Under SMP and Some SMP to Tertiary Natarbora: Until complete SMP and SMA to Tertiary.
Income c	Every visit	Having no income Selling non-animal/agricultural goods Selling animal/agricultural goods Day labour Business Wage earnings	Evidence of having income at least once across years or no evidence of income.
Bolsa da Mae c	Every visit	No. of subsidies received per household	Receiving at least one (per woman) or not receiving at all.
Sanitation e	Every visit	None, traditional (pit latrine) or developed facility		No toilet Traditional or developed.
Water supply c (Ossu)	Every visit	Spring, tap or pipe	Most common type of water supply across fieldwork years grouped as: spring, tap or pipe.	Spring or tap. Pipe.
Water source c (Natarbora)	Every visit	Hand drawn from well. Hand pump River Tank Electric pump Piped water	Superficial (hand drawn from well, hand pump and river), or deep source (tank, electric pump and piped water)
Cows d	Every visit	0 cows, between 0 to 10, or more than 10
Pigs b	Every visit	No. of pigs
Garden plot c	Every visit	List of crops harvested	At least 1 crop harvested or none
Non-labour-reducing Electronics b	Every visit	List of appliances	No. of different non-labour reducing electronics: light bulbs, telephone chargers, radios, music devices, televisions, electric saws, computers, irons, fans, photocopiers, motor, telephones, and internet cables
Labour-reducing Electronics b	Every visit	List of appliances	No. of different labour-reducing electronics: rice cooker, sewing and washing machine, refrigerator, kettle, or mixer
Age pension c	Every visit	No. of age pensions received by household
Residents b	Every visit	No. of adults and children sleeping and eating in household consistently.
Year of birth b	Baseline	Reported verbally or using id card.
Birth region d	Baseline	Reported town of birth	Birth region by altitude level: 0 to 250 m, between 251 m to 500 m and from 501 m onwards

Variables’ measurement levels: ^a Continuous, ^b Discrete, ^c Binary, ^d Ordinal, ^e Nominal.

Table 1. Operationalisation of variables used for CATPCA.

Variable	Collection Period	Measurement at Interview	Pre-CATPCA Operationalisation	Recategorization During CATPCA
Health status a	Every visit	Healthy or unhealthy report	Ratio of reported good health to total years with valid data.
Age at first birth b	Baseline	Year of first birth	Subtraction of first birth year minus woman’s year of birth.
Births b	Every visit	No. of births
Child deaths b	Every visit	No. of child deaths and no. of still births.	Sum of child deaths and still births.
Family planning c	Ossu: 2010, 2013, 2014, 2015 and 2018. Natarbora: 2012, 2013, 2014, 2015 and 2018.	Never used. Used medical method at least once. Using medical method Using non-medical method Used non-medical method	Evidence of use of family planning at least once across years or no evidence of use.
Education d	Baseline	None Some or complete primary Some or complete SMP Some SMA/SPP until tertiary		Ossu: Under SMP and Some SMP to Tertiary Natarbora: Until complete SMP and SMA to Tertiary.
Income c	Every visit	Having no income Selling non-animal/agricultural goods Selling animal/agricultural goods Day labour Business Wage earnings	Evidence of having income at least once across years or no evidence of income.
Bolsa da Mae c	Every visit	No. of subsidies received per household	Receiving at least one (per woman) or not receiving at all.
Sanitation e	Every visit	None, traditional (pit latrine) or developed facility		No toilet Traditional or developed.
Water supply c (Ossu)	Every visit	Spring, tap or pipe	Most common type of water supply across fieldwork years grouped as: spring, tap or pipe.	Spring or tap. Pipe.
Water source c (Natarbora)	Every visit	Hand drawn from well. Hand pump River Tank Electric pump Piped water	Superficial (hand drawn from well, hand pump and river), or deep source (tank, electric pump and piped water)
Cows d	Every visit	0 cows, between 0 to 10, or more than 10
Pigs b	Every visit	No. of pigs
Garden plot c	Every visit	List of crops harvested	At least 1 crop harvested or none
Non-labour-reducing Electronics b	Every visit	List of appliances	No. of different non-labour reducing electronics: light bulbs, telephone chargers, radios, music devices, televisions, electric saws, computers, irons, fans, photocopiers, motor, telephones, and internet cables
Labour-reducing Electronics b	Every visit	List of appliances	No. of different labour-reducing electronics: rice cooker, sewing and washing machine, refrigerator, kettle, or mixer
Age pension c	Every visit	No. of age pensions received by household
Residents b	Every visit	No. of adults and children sleeping and eating in household consistently.
Year of birth b	Baseline	Reported verbally or using id card.
Birth region d	Baseline	Reported town of birth	Birth region by altitude level: 0 to 250 m, between 251 m to 500 m and from 501 m onwards

Variables’ measurement levels: ^a Continuous, ^b Discrete, ^c Binary, ^d Ordinal, ^e Nominal.

Birthplaces of women living in Natarbora were grouped in three birth regions based on altitude: from 0 to 250 m, between 251 m and 500 m, and from 501 m onwards. Birth region was not included in our analysis of women living in Ossu given that there was not enough information on this matter (60% of missing values) for proper analyses. In addition, there was little variation among those women who had birthplace data; almost all (90%) were born in mountainous areas from the Ossu region.

2.3. Analyses

The following process was implemented separately for data collected from each community. First, we imputed missing information manually based on our experience in the field for the variables that had a high number of missing values (

Table 2. t-Tests and Chi-Squares with pre-CATPCA variables across Ossu and Natarbora 2018.

Variable Code, Label	Ossu				Natarbora				p **
	Valid	Missing	Mode	$\overline{\mathbf{x}}$ (SD)/%	Valid	Missing	Mode	$\overline{\mathbf{x}}$ (SD)/%
Health ratio	116	0	1	0.78 (0.29)	140	0	1	0.61 (0.33)	p < 0.001
Age at first birth	105	11	22	22.29 (4.13)	140	0	22	22.11 (4.66)	p = 0.765
Births	116	0	4	5.37 (2.45)	140	0	1	4.17 (2.46)	p < 0.001
Child deaths	116	0	0	0.75 (1.19)	140	0	0	0.26 (0.55)	p < 0.001
Family planning  0 No evidence  1 Used at least once	116	0	0	70.70% 29.03%	140	0	0	67.9% 32.1%	p = 0.625
Education  0 None  1 Some or complete primary  2 Some or complete SMP  3 Some SMA/SPP until tertiary	82	34	0	37.8% 20.7% 23.2% 18.3%	140	0	0	35.5% 19.4% 22.6% 22.6%	p = 0.919
Income  0 No  1 At least once	116	0	0	52.6% 47.4%	140	0	0	75% 25%	p < 0.001
Bolsa da mae  0 No  1 At least one	116	0	0	94% 6%	140	0	0	68.6% 31.4%	p < 0.001
Toilet  0 No toilet  1 Traditional  2 Developed	116	0	2	20.7% 3.2% 67%	140	0	2	11.4% 6.4% 82.1%	p = 0.013
Water supply  0 Spring  1 Tap  2 Pipe	116	0	2	21.6% 24.1% 45.7%
Water source  0 Superficial  1 Deep source					140	0	1	21.4% 78.4%
Cows  0 No Cows  1 Between 1 to 10  2 >10 cows	113	3	0	61.1% 26.5% 12.4%	139	1	1	43.2% 46.4% 10%	p = 0.004
Pigs	104	12	0	1.66 (2.18)	132	8	0	2.17 (0.65)	p = 0.093
Garden plot  0 non-active  1 at least one crop	109	7	1	26.6% 73.4%	136	4	1	23.5% 76.5%	p = 0.580
Non-labour-reducing electronics	110	6	4	4.39 (1.41)	136	4	4	4.67 (0.43)	p = 0.184
Labour-reducing electronics	110	6	0	1.12 (1.06)	136	4	0	1.29 (1.86)	p = 0.230
Age pension  0 No pension  1 At least one	116	0	0	90.5% 9.5%	140	0	0	68.6% 31.4%	p < 0.001
Residents	112	4	5	7.37 (2.79)	139	1	6	7.48 (0.47)	p = 0.742
Year of birth	114	2	1984	1975.61 (12.44)	138	2	1990	1976.14 (15.07)	p = 0.759
Birth Region  0, 0 to 250 m  1, 251–500 m  2, 501 m and more					105	35	0	69.5% 16.2% 14.3%

Chi squares and t tests were used to compare variables across sites. p ** Bold p values are significant with a 95%. Gray rows are unused variables for that field site. Blank cells represent non-comparable variables across sites. Italics correspond to supplementary variables.

). For variables with fewer missing data, we imputed the modes within the CATPCA analysis. All modes were reviewed and discussed to assess their accuracy and the implications of using modes before imputation. Next, we iterated CATPCAs, removing at each iteration the variable that fell under the recommended component loading threshold of ±0.400 (Linting and van der Kooij 2012). Only one variable was removed at a time. The iterations stopped once all the variables’ component loadings were above the threshold across all components. As explained in the previous section, some variables were recoded based on their transformation plots after discretisation, and these recategorizations were reviewed and discussed in each CATPCA iteration. Discretisation methods were selected based on the variables’ measurement level (Table 1).

The number of components (profiles) was determined by the variance accounted for (VAF) of the CATPCA models. We determined that the best solution was to use 4 components due to its consistent >45% VAF (for both communities) throughout iterations. Final CATPCA models include birth region and YOB as supplementary variables to understand their relationship with women’s profiles. The object scores associated with the final CATPCA model for each case (woman) indicated the extent to which each woman was described by each particular component of the CATPCA. Subsequently, t-tests and ANOVAs were performed with women’s object scores to understand the relationship between women’s profiles and their birth cohort (for women living in both communities) and birth region (for women living in Natarbora). All analyses were performed using IBM SPSS Version 27.

3. Results

3.1. Comparison of Pre-CATPCA Variables Across Field Sites

Women differed significantly by community in their health ratio, births, child deaths, income, government transfers, and in having cows (Table 2). Compared to Natarbora, women in Ossu had a higher mean number of births ($\overline{\mathrm{x}}$ = 5.4 > 4.2; t = −3.89, p < 0.001) and child deaths ($\overline{\mathrm{x}}$ = 0.7 > 0.3; t = 4.09, p < 0.001) and reported being healthier ($\overline{\mathrm{x}}$ = 0.8 > 0.6; t = 4.3, p < 0.001). More Ossu women had access to an independent source of income (X²₍₁₎ = 13.98, p < 0.001), but were less likely to receive Bolsa da Mae (X²₍₁₎ = 25.64, p < 0.001). More Ossu women were living in households that did not receive age pensions (X²₍₁₎ = 18.11, p < 0.001), had no sanitation (X²₍₂₎ = 8.54, p = 0.013), and no cows (X²₍₂₎ = 10.95, p = 0.004). Seventy percent of women in Natarbora were born on the coastal plains, 16% migrated to the region from areas between 251 and 500 metres above sea level, and 14% migrated from mountainous regions (501 m and higher).

3.2. CATPCA

Each of the CATPCA models explains more than 45% of the variance (

Table 3. Variable component loadings, eigenvalues for final CATPCAs of both communities.

	Ossu Components				Natarbora Components
	1 Tech and Sanitation	2 Contraception	3 Traditional	4 High Fertility	1 Tech and Sanitation	2 Traditional	3 Contraception	4 High Fertility
Health Ratio					0.104	−0.207	0.652	−0.059
Age at first birth					−0.171	−0.266	−0.444	0.289
Births	−0.068	0.041	0.132	0.854	−0.086	−0.044	0.117	0.807
Child deaths	−0.169	−0.153	0.042	0.702	0.095	0.033	−0.134	0.815
Family planning	0.250	0.685	−0.185	0.150	−0.068	0.06	0.741	0.059
Education	0.351	0.194	−0.464	−0.085	0.437	−0.334	−0.062	0.217
Income	−0.016	0.154	−0.053	0.538	0.101	−0.067	0.415	0.268
Bolsa da mae	−0.080	0.709	−0.169	0.059	−0.276	−0.17	0.508	−0.11
Sanitation	0.625	−0.368	−0.059	−0.133	0.510	−0.134	−0.164	−0.074
Water source					0.518	−0.107	−0.005	−0.08
Water supply	−0.197	0.524	0.285	−0.213
Cows	0.378	0.100	0.650	0.166	0.033	0.783	0.11	0.06
Pigs	0.119	0.599	0.445	0.024
Garden plot	−0.296	−0.084	0.477	−0.048
Non-labour reducing electronics	0.726	0.052	0.086	−0.043	0.722	0.132	0.092	0.075
Labour-reducing electronics	0.710	0.090	0.031	−0.129	0.733	0.087	0.168	0.049
Age pension					−0.177	0.625	−0.246	0.006
Residents	0.165	0.046	0.771	−0.002	−0.002	0.709	−0.134	−0.086
Year of birth	0.165	0.108	0.138	−0.686	0.071	−0.117	0.179	−0.556
Birth region					−0.154	−0.057	0.08	0.337
% Variance	13.98	13.26	13.11	11.83	13.06	12.18	11.95	10.45
(Rotated eigenvalue)	1.957	1.857	1.836	1.657	1.958	1.827	1.792	1.568
% Total variance	52.19				47.64

Loadings >|±0.400| are highlighted in bold. Supplementary variables are in italics. Blank spaces represent variables not included in analysis. Gray cells indicate variables that dropped out of the CATPCA model.

) and four profiles emerged per community: Tech and Sanitation (Ossu 13.98%, Natarbora 13.06%), Traditional (Ossu 13.11%, Natarbora 12.18%), Contraception (Ossu 13.26%, Natarbora 11.95%), and High Fertility (Ossu 11.83%, Natarbora 10.43%). Self-reported health, age at first birth, and age pension dropped out of the CATPCA in Ossu and having pigs and a garden plot dropped from the model in Natarbora. In both communities, women scoring high on Tech and Sanitation “Tech and Sanitation women” (Component 1 Ossu, Component 1 Natarbora) resided in households with non-labour-reducing (0.726 OS, 0.722 Natarbora) and labour-reducing (0.710 Ossu, 0.733 Natarbora) electronics and better sanitation (0.625 Ossu, 0.510 Natarbora). Women scoring high on Traditional “Traditional women” (Component 3 Ossu, Component 2 Natarbora) lived in households with more residents (0.771 Ossu, 0.709 Natarbora) and cows (0.650 Ossu, 0.783 Natarbora). Women who scored high on Contraception “Contraception women” (Component 2 Ossu, Component 3 Natarbora) used contraception (0.685 Ossu, 0.741 Natarbora) and received at least one bolsa da mae (0.709 Ossu, 0.508 Natarbora). Women who had high scores on High fertility “High fertility women” (Component 4 Ossu, Component 4 Natarbora) displayed more child births (0.854 Ossu, 0.807 Natarbora) and child deaths (0.702 Ossu, 0.815 Natarbora), and older age (−0.686 Ossu, −0.556 Natarbora).

Site-specific variations within component loadings exist. In Natarbora, the Tech and Sanitation profile associated with having at least some years of education post high school (0.437) and a deep source of water (0.518), in contrast to Ossu, where Tech and Sanitation tended to correlate with having cows (0.378) and at least some years of high school (0.351). In Ossu, the Traditional profile comprised less education (−0.464), pigs (0.445) and a garden plot (0.477), whereas in Natarbora this profile included having at least one age pension in the household (0.625). In Ossu, the Contraception profile (Component 2) is composed of strong positive loadings of use of family planning, living in a household that received water through a pipe (0.524), had pigs (0.599), and tended to associate with not having sanitation (−0.368). In Natarbora, family planning (0.741) associated with better self-reported health (0.652), lower age at first birth (−0.444), and having experienced an independent income at least once (0.415). High fertility in Ossu women associated with having an independent income (0.538). In Natarbora, the profile tended to correlate with being born in the mountains (0.337).

3.3. t-Tests: Birth Cohort and Birth Region

t-tests were used to examine the effect of birth cohort and birth region on the means of women’s individual scores (object scores) derived from the CATPCA, and on women’s age at first birth. In Ossu, there was a statistically significant difference in the means of object scores of High fertility between the two birth cohorts. Women born prior to 1969 had higher means ($\overline{\mathrm{x}}$ = 0.525, t = 2.95, p = 0.004) when compared to the means of women in the younger cohort ($\overline{\mathrm{x}}$ = −0.120) (Figure 2).

In Ossu, being born before 1969 also correlated with having an older age at first birth by 3.6 years when compared to the younger cohort ($\overline{\mathrm{x}}$ = 25.1 > 21.6, t = 2.63, p = 0.016). There was a tendency of younger women in this site to associate with the Contraception profile (t = 1.88, p = 0.061).

In Natarbora, the mean object scores of the High Fertility component differed significantly across birth cohorts ($\overline{\mathrm{x}}$ = 0.757 > −0.284, t = 5.02, p < 0.001); older women were more likely to exhibit the higher fertility profile. In addition, the older cohort had higher Traditional mean object scores compared to the mean of younger women ($\overline{\mathrm{x}}$ = 0.292 > −0.130, t = 2.26, p = 0.025). Birth cohort also influenced the Contraception scores; Natarboran women born in or after 1969 had higher means on this profile when compared to the older cohort ($\overline{\mathrm{x}}$ = 0.182 > −0.400, t = 13.43, p < 0.001).

Regarding birth site, there were significant differences in the mean High fertility scores across birth regions in Natarbora (F 2, 102 = 6.02, p = 0.003). A Games-Howell test for multiple comparisons found that High Fertility scores were higher in women who were born at altitude exceeding 500 m ASL compared to those born at lower altitudes (0 to 250 m) regions ($\overline{\mathrm{x}}$ = 0.888 > −0.071, p = 0.014, 95% C.I. = 0.174, 1.74).

4. Discussion

The aims of this study were to characterise the life experiences of women using their health, reproductive outcomes, indirect indicators of empowerment and household ecology in the communities of Ossu and of Natarbora, to compare women’s profiles across sites, and to relate profiles to women’s birth cohort and region, interpreting findings relative to LHT, LCT and the framework of women’s empowerment.

In this section we first discuss the differences between sites in their frequencies of the variables used for the CATPCA (Table 2), then we discuss the effects of birth cohort and region on women’s profiles and age at first birth derived from the CATPCA and ANOVA analyses (Figure 2 and Table 3).

4.1. Differences Between Communities: Health, Fertility, and Income

Women in Ossu generally reported better health in comparison to those in Natarbora (Table 2). This result appears counterintuitive, particularly considering comparatively lower availability of food and water in Ossu (Thu and Judge 2017) which are conditions necessary for maintaining good health. This finding may be in part due to varying attitudes towards health between the two regions. Rural communities often define their health in terms of their independence and self-sufficiency and may display stoic attitudes towards disease (Gessert et al. 2015). A previous study in Timor-Leste demonstrated that Ossu households are more “independent” and less engaged in reciprocity in comparison to households in Natarbora (Sumich 2021).

In other rural contexts, self-rated good health has been linked to wealth (Jaeggi et al. 2021) and, among rural women, to the quality of marital relationships (Agadjanian 2023). To explore whether these correlations exist in Ossu, we conducted a supplementary analysis comparing the means of women’s health ratios across their education level, independent income, and marital history (married at least once in life or never been married) within this community. However, our analysis did not reveal significant differences in self-rated health across these variables thus, more research is needed to better understand this matter as it was not the focus of this study.

Another significant difference between sites is the high level of child deaths and births observed amongst women in Ossu, which aligns with previous research on fertility patterns conducted in this area (Sumich 2015). The high number of child deaths and births may be attributed to cultural practices and greater geographic isolation. Unlike Natarbora, Ossu has patrilineal family affiliations, and the tradition of bride price (barlake) is more prevalent in this region (Spencer and Judge 2021). Studies in Timor-Leste have found evidence of a correlation between these practices and higher fertility (Samad et al. 2022; Wallace et al. 2018). With higher numbers of births, there is a greater probability of experiencing child deaths. However, the mountainous geography of Ossu may also contribute to a higher child mortality by limiting access to health care services.

The difference in access to an income independent of a partner across field sites might be partially explained by the variation in infrastructure between regions. Women in Ossu reported higher access to income, and a previous study in this community found that households had more access to cash when compared to Natarbora due to an all-year accessible and well-connected road in the region that facilitated engaging in off-farm paid activities (Thu and Judge 2017). However, in the Timorese context, women’s independent income does not necessarily represent a better economic status for women, as most of their income-earning activities are lower paid in comparison to men, and earnings are usually spent on their family’s welfare (Samad et al. 2022).

The higher engagement of women in low-paid, income-earning activities in Ossu aligns with additional indicators suggesting poorer living conditions in this community in comparison to Natarbora. These include more limited access to sanitation and subsidies such as Bolsa da mae and age pension, along with living in households that are less likely to own cows. In rural Timor-Leste, cows are primarily used as a long-term resource investment (Spencer et al. 2018), thus, it is expected that women in Ossu, with fewer resources, are less likely to live in households that raise cows. These findings are consistent with the results of previous studies (Spencer et al. 2018; Thu and Judge 2017) that illustrate that living conditions in Ossu are worse than in Natarbora.

4.2. Women’s Profiles and Reproductive Timing

In both communities, birth cohort and region influenced women’s High Fertility, Traditional, and Contraception profiles, and age at first birth (Table 3). In this section, we address first the profiles (High Fertility and Traditional) associated with the older birth cohort (born before 1969) and their delayed reproduction. We then discuss the Contraception profile and the earlier age at first birth of the younger cohort of women. Following that, we explore the Tech and Sanitation profile which was not linked to year or place of birth but rather to having cash. These findings, including the differences within the elements of profiles across sites (Table 3), are discussed using LHT, LCT and the women’s empowerment framework.

In both communities, women born before 1969 significantly associated with the High Fertility profile characterised by a greater number of child births and deaths (Table 3). In Ossu, the older cohort of women had their first child at a significantly older age, while in Natarbora this was a trend (Figure 2).

From a LCT perspective, these results can be partly framed as a product of the context of these women’s reproductive years, since in LCT decisions are understood as shaped, to a degree, by the socio-historical contexts in which individuals exist (Sampson and Laub 1992). For the older cohort of women, the increased birth rates, child deaths, and postponed reproduction might partly stem from a lack of contraceptive access during their youth and the societal stigma linked to fertility control in the Timorese context (Wallace et al. 2018). Moreover, many of these women reached reproductive age around the onset of the 1975 occupation. Mortality was high during conflict (Durand 2011), and it is possible that some women encountered challenges to earlier childbearing due to a lack of potential marital partners.

Postponing the age at first birth can also be understood using the LHT framework. In LHT, reproductive timing is analysed in relation to environmental conditions. Earlier or delayed reproduction may constitute an adaptive response to environmental characteristics and thus maximize fitness (Chisholm 1999). In some instances, delayed reproduction evolves in response to environments characterised by resource scarcity, allowing adult organisms to direct resources predominantly to maintenance until conditions become favourable for reproduction (Chisholm 1999; Sear et al. 2019). In Natarbora, women who were born before 1969 tended to have an older age at first birth (Figure 2) and had higher scores on the High fertility component. This component tends to correlate with high altitude birth regions (Table 3). Such birthplaces might have exacerbated the resource scarcities they encountered in their developmental years, as mountainous areas in Timor-Leste struggle with poorer access to water, food, and healthcare (Ministry of Health 2018). We suggest that the older age at first birth of these women in both communities may be in part due to the characteristics of the environment during their development which were marked by limited resources (Durand 2011).

Being born before 1969 also aligned with the Traditional profile in Natarbora, characterised by larger household sizes, the presence of at least one person receiving an age pension, cow ownership, and a tendency toward fewer years of formal education (Table 3). We infer that Traditional women might reside in three-generation households where grandmothers commonly assist in child-rearing. A study in Timor-Leste revealed that, compared to Ossu, most households in Natarbora have family structures encompassing grandparents and a greater number of fostered children, and that children in these types of households have better growth (as measured by body mass index) than their peers in two-generation or single generation households in the same community (Spencer and Judge 2021). Future research should examine the correlation between women’s profiles and child growth; we examine this in following work.

Women born after 1969 exhibited higher scores for the Contraception profile in Natarbora, a trend that was also observed in Ossu (Figure 2). This profile includes the use of family planning at least once and receiving at least one Bolsa da mae subsidy (Table 3). Accessing these benefits requires time and resources to travel to health clinics and government facilities. This indicates that Contraception women possess the agency and knowledge to invest time in both fertility management and obtaining financial support for caregiving. Drawing from Kabeer (1999) definition of women’s empowerment, which views empowerment as a transition from suboptimal to favourable conditions through conscious choices, it’s clear that Contraception women are empowered. They make decisions to better their and their children’s lives, despite the geography and infrastructure limitations of their rural location.

There are variations in the elements within this profile between Ossu and Natarbora. In Ossu, Contraception women raise pigs and have access to piped water (Table 3). Given the seasonal water scarcity in mountainous Ossu (Thu and Judge 2017), the time commitment to secure water is substantial. The installation of pipes reduces the time women need to invest in obtaining clean water, as it is directly channeled into their households. In addition, livestock is primarily used as a ‘bank account’ rather than for consumption (Spencer and Judge 2021), and in this site, pigs are not allowed to roam free, which implies that owning a pig requires greater time and resource investment to care for it. However, in Natarbora, Contraception women have income independent of their partners (Table 3), suggesting that they use paid work as a means of generating additional resources.

Regarding reproduction, this younger cohort of women had their first child at a younger age than did the older cohort of women in both communities. This may be partially explained by the socio-cultural environments in which these women were raised. In LCT research, this is commonly referred to as a ‘cohort effect’, where the interaction of sociological and ecological factors during a particular historical context have an impact on later life outcomes including reproductive traits (Bernardi et al. 2019; Gilligan et al. 2018). For example, Neal et al. (2016) reviewed how armed conflicts influence adolescent reproductive behaviours, noting both increases and decreases in early marriage and childbirth rates. They noted that the factors influencing early marriages include securing financial stability through bride prices, protecting young girls from sexual violence, disruptions to education systems, and communities reverting to traditional social norms to maintain social cohesion post-conflict. The earlier age at first birth of this cohort of younger women might be influenced by similar factors as they experienced their first six years of life in an environment characterised by instability due to the Indonesian Occupation (1975–1999).

LHT may be useful to understand the underlying mechanisms by which this occurs. Famine, violence against women and children, forced displacement and death were common during the occupation (Silove et al. 2014; The Asia Foundation 2016). Thus, unlike the older cohort of women who might have delayed their reproduction due to resource scarcity, this younger cohort may have an accelerated reproductive timing due to their unstable developmental environment but may be investing more resources into their children than the older cohort since they are limiting their number of births.

Our findings align with Demographic Transition Theory (DTT), which describes how societies transition from high birth and death rates to lower fertility and mortality as economic and social development progresses (Kirk 1996). In our study, older cohorts in mountainous regions exhibited higher fertility and child mortality rates, partly due to historical constraints such as conflict and limited access to healthcare, while younger cohorts in coastal areas displayed lower fertility rates, consistent with DTT.

Additionally, elements of the Second Demographic Transition (SDT)—which emphasizes cultural and ideational shifts leading to delayed reproduction and fertility control (Lesthaeghe 2010; van de Kaa 2004)—may help explain the increased use of contraception among younger cohorts in Natarbora. While SDT has primarily been used to analyse fertility patterns in industrialised nations, recent evidence suggests that rural populations experiencing modernisation and improved access to education may also be undergoing SDT-like transitions (Zhang et al. 2023). For example, in rural China, rising educational attainment has contributed to shifts in gender role attitudes and family dynamics, both of which are key drivers of reproductive decision-making (Zhang et al. 2023).

While our study did not directly assess women’s beliefs about reproduction, our findings suggest that women in Natarbora—who are more educated than their mountain-dwelling counterparts—may be experiencing fertility transitions in response to changing structural conditions, particularly greater access to education. Future research should explore how women’s beliefs and attitudes toward reproduction influence their reproductive behaviours in these rural populations.

4.3. Tech and Sanitation: Wealth and Time Affluence

Tech and Sanitation is the only profile that was not influenced by birth cohort or region, but rather associated with measures of wealth. Tech and Sanitation women are characterised by having appliances and sanitation in their households in both communities (Table 3). In Ossu, Tech and Sanitation women tended to raise cows and to have some high school education. In contrast, in Natarbora, these women have some post-high school education and access to a deep-water source (Table 3). These factors indicate that Tech and Sanitation women possess wealth, as cows represent a long-term investment in this context, and pursuing education beyond high school, usually in Dili, requires resources for relocation and study. Interestingly, Tech and Sanitation women do not have an ‘income independent of a partner’, suggesting that they may have alternative cash sources to fund their education and the purchase of appliances, most of which also require travel to Dili.

Both labour-reducing and non-labour-reducing appliances had the highest loadings within Tech and Sanitation and were positively correlated with each other in both Ossu (r(108) = 0.36, p < 0.001) and Natarbora (r(134) = 0.46, p < 0.001). The moderate strength of these correlations suggests that women accessing Tech and Sanitation are not making trade-offs between the two categories of electronics; rather, they are purchasing both types of appliances with comparable frequency. However, the two categories are different in their impact on women’s time and energy budgets.

In our study, labour-reducing appliances include rice cookers, sewing and washing machines, refrigerators, kettles, and mixers. Traditionally, women in Ossu and Natarbora cook rice and boil water over open fires or stoves, shop for or gather fresh food daily due to being unable to preserve perishables, prepare food items (e.g., dough for sweets to sell) and repair or weave clothes by hand. Labour-reducing appliances significantly mitigate the time burden associated with these activities allowing women to focus on other tasks.

From a LHT perspective, time is a valuable resource that can be allocated to three main biological functions: growth, reproduction, and maintenance (Hill and Kaplan 1999). Tech and Sanitation women are gaining more time by acquiring labour-saving appliances. Time could then be invested in other activities. In LCT (Bernardi et al. 2019), buying appliances to reduce household labour can be understood as a conscious decision that Tech and Sanitation women make to improve their overall well-being. To enact their agency, women in this group are using alternative sources of cash (resources) to achieve a desired outcome, a behaviour that may be an indicator of empowerment (Kabeer 1999). Increased available time has been associated with more empowerment of women in rural contexts (Calvi et al. 2022; Sharaunga et al. 2016). Specifically, in Timor-Leste, less leisure time due to heavy workloads, and gender-biased time allocation for household and caregiving work, as well as poverty, correlate with women’s disempowerment (Bonis-Profumo et al. 2021). Thus, it appears that Tech and Sanitation women in both communities may be making informed decisions to reduce housework time.

5. Conclusions

Using LCT and LHT is useful for analysing the interconnected aspects of women’s histories, environments, and behaviours. LCT allows for understanding the socio-historical context and individual agency in shaping women’s life courses. LHT offers a perspective on how evolutionary pressures and resource allocation strategies influence reproductive behaviours. Our quantitative approach facilitated the identification of significant correlations between indicators of women’s historical context, agency, and reproduction in the Timorese context. We found that older women were more likely to associate with the High Fertility and Traditional profiles and had delayed reproduction, while younger women were more prone to correlate with the Contraception profile and early reproduction. Additionally, irrespective of birth cohort, women in the Tech and Sanitation profile in Natarbora, are highly educated, and are investing in appliances that reduce housework, potentially allowing them to invest more time in other activities.

6. Policy Implications

Our results highlight key areas where targeted policies could enhance women’s health, autonomy, and household well-being in rural Timor-Leste. The association between household infrastructure (Tech Sanitation) and indicators of empowerment suggests that investments in access to sanitation, water, and labour-saving technologies may alleviate time constraints, enabling women to pursue education and economic opportunities. Additionally, reproductive timing variations linked to early life conditions reinforce the need for context-specific reproductive health initiatives that acknowledge both socio-historical influences and structural constraints. Given that fertility control was more common among younger cohorts in Natarbora, strengthening access to family planning resources could support informed reproductive choices across different rural contexts. Moreover, our findings suggest that environmental and structural factors, such as geographic isolation in high-altitude regions, influence fertility and child mortality rates, pointing to the importance of improving healthcare accessibility in remote areas. Addressing these interconnected issues requires collaborative efforts between policymakers, healthcare providers, and community organisations to create interventions that align with the lived realities of rural women while promoting long-term improvements in health and empowerment.