Gifting Relationships and School Dropout in Rural Malawi: Examining Differences by Gender and Poverty Level.

Research from sub-Saharan Africa has shown the heightened likelihood of dropping out of school for students in sexual relationships, particularly girls. However, our knowledge is limited as to whether the risk of school dropout is exacerbated by the exchange of gifts in the relationship as well as students' poverty level. Drawing on longitudinal survey data from rural Malawi, this study explores these questions, examining differences by gender and poverty level in the association between being in a sexual relationship in which gifts are exchanged and school dropout for adolescents in primary school. Our findings show that for both boys and girls, being in a gifting relationship heightens the risk of school dropout and eliminates the protective advantages of being nonpoor on dropout. However, non-gifting sexual relationships also erase the protective advantage of being nonpoor for girls, but not for boys. These results point to the value of examining poverty-gender interactions to gain a more nuanced understanding of the impact of sexual relationships on adolescent trajectories.

Research from across sub‐Saharan Africa has shown that adolescents in sexual relationships have a higher likelihood of dropping out of school, particularly girls (Clark and Mathur 2012; Frye 2017; Sunny et al. 2019). In many settings, gift‐giving from boyfriends to girlfriends is a routine part of adolescent romantic relationships (Kaufman and Stavrou 2004; Mojola 2014; Moore et al. 2007; Poulin 2007; Verheijen 2011), and yet we know little about whether the exchange of gifts in these relationships heightens the risk of dropout. Additionally, the financial dynamics of gift‐giving within relationships raise questions about how the gendered risks of dropout may vary by poverty level. Poverty could magnify the consequences of gift‐giving such that poor girls in these relationships might be additionally disciplined or disinvested in compared to wealthier girls. For poor boys, the pressure to earn money in order to support the provision of gifts may impede school attendance. These hypotheses suggest that gift‐giving might be particularly consequential for poor students, but require further analysis.

In this article, we explore these gender and poverty dynamics around gifting relationships and school dropout, using longitudinal data from rural Malawi. Specifically, we study how being in a sexual relationship in which gifts are exchanged is associated with dropping out of school amongst adolescent girls and boys enrolled in primary school. Our findings show that being in a gifting relationship heightens the risk of dropout for both girls and boys, but that the implications of poverty level differ by gender. We find that girls and boys from nonpoor households are significantly less likely to drop out of school than youth from poor households, but contrary to our expectations, this advantage disappears for nonpoor students who participated in gift‐giving relationships—their probability of dropping out of school is not statistically different from those of poor students in gift‐giving relationships. Non‐gifting sexual relationships also erase the protective advantage of being nonpoor for girls, but not for boys. We then explore three potential mechanisms that might explain this differential association by poverty level for boys: school absence, working for pay, and the reported consequences of carrying out chores or paid work such as being unable to concentrate in school. Of these mechanisms, we find that only school absence was significantly associated with school dropout for boys, but it failed to explain the associations for poverty or gift‐giving. For girls, we find that gifting's association with school dropout may in part reflect the movement toward a more serious relationship as girls in gifting relationships were significantly more likely to marry and have a birth.

THEORETICAL PERSPECTIVES ON SEX AND EXCHANGE

Gifts, often in the form of money, have been found to shift power dynamics in romantic relationships, compromising when young women start having sex as well as whether condoms are used (Blanc 2001; Luke 2003; Stoebenau et al. 2016). Subsequent research has added nuance, showing that sexual relationships in which gifts are exchanged include, but also extend beyond, the archetypal Sugar Daddy‐schoolgirl dynamic (Luke 2005a; Moore et al. 2007; Wyrod et al. 2011). For one, in many instances, gift‐giving is a routine aspect of being in a romantic relationship, including those between age‐mates (Kaufman and Stavrou 2004; Poulin 2007). Survey data from Ghana, Malawi, and Uganda showed that between 70 and 80 percent of adolescent girls (12–19 years) in a sexual relationship with a partner who was similar in age to them had received money or gifts (Moore et al. 2007). In addition, women who receive gifts in relationships often enter into these exchanges willingly and even strategically (Mojola 2014; Verheijen 2011; Wamoyi et al. 2011). Despite this added depth, relatively little remains known about how sexual relationships involving gift‐giving impacts young men and women differently, and to what extent these associations differ by poverty level.

In recent years, scholars have developed frameworks to make sense of the wide variety of relationship types in which money or gifts are exchanged. For instance, Luke (2005b) focuses on the relation between the two people involved as well as the meaning of the gift exchanged and any reciprocal obligations it entails. According to this model, gifting—the exchange of relatively low value but romantically symbolic items between age‐mates at school, for example—would be less likely to impact sexual health outcomes, given the more minimal power differential between the couple. In another framework, to distinguish between types of transactional relationships, Stoebenau and colleagues (2016) propose three continua: the level of deprivation of a particular social context, how much agency women have in these relationships, and lastly, the extent to which women's entry into these relationships is motivated by financial or status gain (instrumentality) (Stoebenau et al. 2016).

Given their emphasis on negotiation and women's reasons for entering certain types of relationships, these frameworks are particularly valuable for hypothesizing sexual health outcomes, but they can also serve to inform hypotheses around other outcomes of interest, such as school dropout in the case of this paper. Although student status has been found to be protective against transactional sex (Gichane et al. 2020), the emphasis on the symbolic value of the gift could lead us to expect that gifting relationships would heighten the risk of school dropout, given that the gifts indicate a certain level of seriousness about the relationship. Alternately, we might expect the association between gifting and dropout to be particularly large for poor girls, who might be more motivated to enter such relationships for instrumental reasons, which in turn could sway the power dynamics of the relationship, making it more disruptive for girls’ schooling.

Focusing on relative rather than absolute levels of socioeconomic status, that gifts tend to flow from men to women in romantic relationships is understood to reflect that the gift‐giver (man) is better off economically than the gift‐receiver (woman) (Blanc 2001; Luke 2003). This focus on power differentials has produced important insights around sexual health behaviors and outcomes; for example, large age asymmetries combined with the exchange of gifts/money were found to be positively associated with the non‐use of condoms in Kisumu, Kenya (Luke 2005a). At the same time, the focus on relative status may have hindered attention to variation in men's economic positions. The framing of gift‐giving as a clear indication of economic power is sound in contexts where gift‐giving is the purview of only wealthier men, but appears overly simplistic in contexts in which the exchange of money and gifts is a fundamental aspect of relationships, something that all men, rich or poor, young and old, engage in.

Thus, in this paper, we extend these prior frameworks on gifting relationships and gender through focusing on how gender interacts with socioeconomic status to shape the consequences of being in a gifting relationship for both girls and boys.

LOVE AND MONEY IN STUDENT RELATIONSHIPS

Qualitative research on romantic relationships and gift‐giving between young people in sub‐Saharan Africa reflects the complexity of classifying gifting relationships. One key point from this literature is that gifting is rarely transactional in a pure sense: only done to receive sex or the gifts. Rather, gifting often indicates romantic interest before a relationship begins and, once begun, shows that the relationship is ongoing (Poulin 2007). Alongside this intertwining of love and gifts, secondary school girls do not usually enter into relationships with their male peers purely for reasons of need such as for school fees, in part because adolescent boys in low‐income contexts rarely have much money to give (Nyanzi, Pool, and Kinsman 2000). Instead, through these relationships, girls often hope and expect to receive or buy consumer items, including beauty and sanitary products, that have become understood globally as necessities for modern young women (Bhana and Pattman 2011; Kaufman and Stavrou 2004; Mojola 2014; Nyanzi et al. 2000; Ranganathan et al. 2018). Thus, amongst age‐mates, negotiations and exchanges of gifts are largely symbolic—“a ritual performance in which sex, money and prestige are intertwined” (Nyanzi et al. 2000, 96).

Though schoolgirls’ desire for nonessential items may be socially produced, this does not diminish the understanding that they have additional material “needs,” heightening the urge to satisfy them (Mojola 2014). Furthermore, other social factors might limit girls’ ability to cover these needs on their own. For one, girls may be less able to engage in paid work due to demands on their time in the home, negative perceptions of girls engaging in paid work outside the home as well as more limited work opportunities for girls (Kendall and Silver 2014; Nyanzi, Pool, and Kinsman 2000). Considering these dynamics of gendered expectations and economies, scholars have described sexual relationships that involve the exchange of gifts as “somewhere between consensual and coercive” (Hampshire et al. 2011, 227). Also important to note is that in addition to love or satisfying “felt needs,” girls may also enter sexual relationships for physical pleasure, to gain sexual experience, or to assess potential partners to marry or with whom to have a baby (Nyanzi et al. 2000, 88).

These wide‐ranging dynamics characterize gift‐giving in student relationships in Malawi, the site of this study, where premarital romantic relationships begin with a gift and then are sustained by more gift‐giving. As outlined by Poulin (2007), after a boy proposes to a girl to begin a relationship, often through an intermediary, and the girl accepts, he gives her a gift. Though girls are free to reject proposals, agreeing to the proposal signifies an agreement to have sex. Relationships generally require the exchange of gifts to continue, often in the form of money that girls use to buy things they both want and need from soap to lotion, sugar, and doughnuts. The drying up of gifts is taken as a sign that the boy no longer cares as much about the girl and thus is taken as legitimate grounds for a girl to leave the relationship. Given that gifts flow from boys to girls, these exchanges have gendered motivations and consequences. Boys get pride and respect from giving these gifts to a girl, whereas girls risk ridicule if they have sex but do not receive gifts.

Of particular importance to our study is how boys obtain money for gifting. In rural Malawi, where poverty rates are high and casual labor opportunities are scarce, money requires significant effort to accumulate. A quote from a 22‐year‐old man in Poulin's (2007) study reveals the links between labor and gifts: “I worked hard in the field to have money, not less than 30 kwacha at once (to give her). In the first days I gave her no gift but after a month, she asked for an under wear. I tried my best and bought it with almost 100 kwacha.” A study on how adolescents negotiate sexual relationships in south‐western Uganda similarly found that secondary school boys described taking on “odd jobs,” including digging, tending to animals, and gathering firewood and water, to earn additional money to support their relationships (Nyanzi et al. 2000, 87). These findings therefore suggest that young men spend time beyond the actual relationship devoted to obtaining income needed for gifts.

SEX AND SCHOOL DROPOUT: GENDER DIFFERENCES

One of the concerns about adolescents starting sexual relationships while in school is that it may reduce the likelihood of completing their education. Indeed, some studies have found that pregnancy‐related dropouts amongst girls are a major contributing factor to gender disparity in secondary school attainment (Eloundou‐Enyegue 2004). On the other hand, others have found that factors such as lack of school fees play more of a role in why girls drop out of school, but that, still, girls who become pregnant do tend to drop out of school (Mensch et al. 2001).

Studies have found that being in a sexual relationship is associated with dropping out of school for both boys and girls but that these associations tend to be larger for girls (Bengesai et al. 2018; Biddlecom et al. 2008; Clark and Mathur 2012; Frye 2017; Sunny et al. 2019). This differential association appears to be due in part to the unequal consequences of sexual relations and pregnancy for schoolboys and schoolgirls (Lloyd 2005). A study from Kisumu, Kenya, found that for girls, being in a sexual relationship resulted in a “cascade effect”: sexually active girls were less likely to graduate from secondary school, largely explained by pregnancy and marital aspirations (Clark and Mathur 2012, 170). Indicating circular processes, girls who were doing worse in school were also more likely to engage in sexual relationships. In contrast to the girls, sexually active boys were not significantly more likely to drop out of school overall, and boys who were doing worse in school were not more likely to become sexually active than their higher‐performing peers. However, the boys who were more likely to drop out of school were those who were aware that their partner had become pregnant as well as those who had started earning an income.

A study in Malawi found similar contrasting associations: Both boys and girls who were in a sexual relationship were more likely to drop out of school, but this association was much stronger for girls than it was for boys (Frye 2017). For boys, being in a sexual relationship was strongly associated with school absence, which may in part be a result of boys working outside of school to earn money for gifts for girls (Poulin 2007). Unlike the Kenya study, however, this study found limited evidence for the classic causal mechanisms of school absence and performance to explain the stronger association of school dropout for girls. These findings, along with teachers’ negative views about girls engaging in sexual relationships, suggest that cultural understandings about the incompatibility of schooling and sex for girls exert an independent negative effect on girls’ schooling trajectories (Frye 2017).

A more recent study of sexual debut on school dropout in northern Malawi found that sexually active girls were five times more likely and sexually active boys twice as likely to drop out of primary school than their non‐sexually active peers (Sunny et al. 2019). Sexually active girls who dropped out of school were most likely to report having done so because of marriage or pregnancy (70 percent cited these reasons). These responses, however, may reflect the retroactive attribution of cause; marriage, for example, rarely precedes school dropout (Mensch et al. 2014) and usually functions as an impediment to reenrollment rather than a reason for leaving school (Grant 2012). In contrast, about 10 percent of the sexually active boys in the study cited marriage or pregnancy as the reason for dropout whereas around half reported “school‐related reasons” for dropping out, particularly those related to school performance, such as failing exams and repeating a grade, as well as losing interest in school (Sunny et al 2019).

Together, these studies show how being sexually active is associated with school dropout, pointing to gendered pathways. Our study builds on this research in a few key ways. First, these studies do not explicitly include whether the sexual relationships involve gifting in their analyses and yet, gifting may be a mechanism behind the association of these relationships and dropout. Second, though all of these studies on sexual relationships and school dropout control for the socioeconomic status of the students, they do not investigate whether the gendered effects of being in a romantic relationship on school dropout differ by socioeconomic status. In our study, we differentiate between sexual relationships in which gifts were exchanged and those in which they are not, and we explore the interaction between these relationships and poverty level on school dropout for both girls and boys in order to further our understanding of the gendered dynamics between young people's sexual relationships and school dropout.

HYPOTHESES

Based on prior research on sexual relationships and school dropout, we expect to find that being in a gifting sexual relationship will be associated with dropping out of school for both boys and girls, but that this association will be larger for girls. Prior research points to potentially different impacts of being in romantic relationships for boys of different socioeconomic classes. Poorer boys in relationships would likely have to spend more time earning money for gifting than relatively wealthier boys who might be able to have some amount of money without engaging in time‐consuming forms of labor. In turn, boys in relationships might be more likely to be absent from school or to spend less time on schoolwork, both of which could lead to a greater likelihood of school dropout. Poorer girls might be more vulnerable in gifting relationships than nonpoor girls, given their greater need for the gifts due to difficulties in getting these items from elsewhere. As a result, their obligations to their romantic partner might be greater, potentially heightening the risks for school dropout. On the other hand, given that gifting relationships do not have the same demands of time and work for girls, we could expect more minimal differentiation by poverty level than with boys. Another potential reason for a hypothesis in this vein is that schoolgirls may receive similar social sanctions for being in a sexual relationship regardless of socioeconomic status. In terms of the pathways that focus on school absence and paid work, we expect to find them to potentially explain the association with gifting and dropout for boys but not girls. For girls, we also expect gifting to be associated with marriage and potentially also motherhood, given how it reflects a more serious relationship.

DATA AND METHODS

This paper uses data from the Malawi Schooling and Adolescent Study (MSAS), a longitudinal dataset collected in southern Malawi from 2007 to 2013 by the Population Council (Hewett and Mensch 2019). The initial 2007 sample consisted of 1,764 14‐ to 17‐year‐old students (875 girls and 889 boys) who were randomly selected from the enrollment rosters at 59 randomly selected primary schools in Machinga and Balaka districts. The probability of a particular school being included was proportional to its enrollment in 2006. An additional sample of 885 out‐of‐school adolescents (462 girls and 423 boys) were identified through key informants located at the school and within the school catchment villages but are not included in this analysis. Survey weights were constructed to correct for the differences in the probability of being included in the sample by school, gender, and enrollment status at round 1. Follow‐up interviews were conducted annually between 2007 and 2013.1

Analysis

Using discrete time logit regression, we first estimate models, separately for boys and girls, of the factors associated with being in a non‐gifting sexual relationship relative to a gifting relationship. These models include respondent characteristics as well as characteristics of the partner and relationship, including variables such as the partner's education level as well as the age difference between the partner and the respondent. Next, we examine how being in a sexual relationship that involves gifting is associated with school dropout from primary school. In two nested models, we assess the association with students’ relationship status, controlling for sociodemographic characteristics, and then we add variables related to students’ school performance, including students’ literacy and numeracy skills, whether they repeated the last grade, and if they expect to attend secondary school. We conduct these nested models to examine whether school performance may in part explain the association between gifting status and school dropout. Then, we estimate a third regression model that includes interactions between household poverty status and gifting to examine whether there are gender and wealth differences in these associations. Lastly, we estimate a set of models to test various pathways that might explain the differential association of gifting and dropout by poverty level. For both boys and girls, we first examine the pathways of absence from school, engaging in paid work, and the reported consequences of doing paid work or household chores. Then we examine how relationship status is associated with subsequent marriage and having a birth.

The first three survey rounds collected data on relationship type from a detailed module on sexual activity that was administered privately with a hand‐held audio‐computer–assisted self‐interviewing device. An experiment with this same dataset found that this interview mode, compared to face to face, was more effective at collecting information on stigmatized sexual behavior (e.g., girls’ relationships with teachers or relatives) and reducing socially desired responses (e.g., boys’ exaggeration of sexual experience) (Kelly et al. 2013). Respondents were first asked their age at first sex; respondents who had never had sex were instructed to respond with the number zero. Next, all respondents (including those who entered “zero” in response to the first question) were asked whether they had ever had sex with a series of potential types of partners (e.g., boyfriend/girlfriend, a “hit and run,” a relative, a teacher, and any other type of person not mentioned).2 If a respondent indicated any sexual partners, they were then asked how frequently the respondent received gifts from (girls) or gave gifts to (boys) their most recent sexual partner. Girls were asked, “How often did this person give you gifts, such as money, food, or other items?” while boys were asked, “How often did you give gifts, such as money, food or other items to this person?” The lack of explicit reference to exchange of these items for sex allows for capture of gifting relationships that are not purely transactional (Luke 2005b; Wamoyi et al. 2019). The answer choices to these questions were “never,” “rarely,” “sometimes,” and “often”—respondents were categorized as not giving/receiving gifts if they answered “never” and as giving/receiving gifts otherwise. Respondents who indicated that they had never had sex were not asked the question about gifting. Therefore, we code this variable into four categories: never had sex, no gifting, gifting, or missing/inconsistent.

The final category of “missing/inconsistent” consists of respondents who provided an age at first sex but indicated that they had never had any sexual partners and subsequently were not asked any additional questions about their sexual behavior, including gift exchange. We are unable to discern whether this inconsistent reporting within a survey round is due to a respondent who ever had sex failing to report a sexual partner or a respondent who never had sex erroneously provided an age at first sex. Almost all missing/inconsistent values for gifting were in the first survey round; in subsequent rounds, fewer than 1% of respondents have missing/inconsistent information about gift exchange. Prior analyses of these data found that students were more likely than out‐of‐school youth to provide inconsistent responses (Soler‐Hampejsek et al. 2013). In our analytic sample of students, inconsistent reporting is not associated with any of the key variables and appears to be random.

Respondents were considered to have dropped out of school in a given survey round (t) if they were enrolled in school in year t − 1 but were no longer enrolled in school in year t. All students who left school without passing the primary school leaving certificate exam (PSLCE) taken near the end of standard (grade) 8 are coded as dropping out of school. Respondents were censored if they passed the PSLCE, enrolled in secondary school, or were still enrolled in primary school at the time of their last interview. All time‐varying independent variables are lagged one survey round (t − 1) behind the measure of school dropout. Furthermore, all analyses are weighted and adjusted to control for the stratified cluster sampling design.

Household poverty is derived from an index of household asset ownership (Filmer and Pritchett 1999). A principal components analysis was run for 14 household assets: mattress, sofa, table, chair, lamp, television, radio, cell phone, mosquito net, motorcycle, car, tin roof, electricity, and boat. The first principal component estimated for the entire sample at round 1 was used to calculate the asset index for all subsequent survey rounds, such that values on the asset index are standardized relative to the average round 1 asset ownership. Preliminary analyses indicated that there are significant differences in school dropout between the highest asset quartile and the three lower quartiles, but no significant differences between the three lower quartiles. Therefore, the asset quartiles were converted into a binary variable, such that respondents in the highest wealth quartile were coded as nonpoor and all remaining respondents were coded as poor. By standardizing asset ownership, this variable is flexible as to how household wealth changes over time.

In the analyses of school dropout, our second model adds a set of school performance variables measured at time t − 1 to the first model: whether the respondent had repeated their current grade, whether the respondent expected to attend secondary school, and a measure of a respondent's literacy and numeracy skills. Literacy in Chichewa, the national language, is based on the ability to read two sentences aloud without difficulty. Numeracy was based on the ability to correctly answer 12 questions that tested basic numeracy and knowledge of money. Given that basic literacy in Chichewa was necessary to answer the math questions, these two skills were combined into the following categorical variables: not literate, literate but not numerate, and literate and numerate. Missing skill levels in subsequent rounds were inferred from the respondent's performance in the previous survey round.

The final sets of models examine pathways through which gifting relationships may lead to school dropout. First, we explore a set of three variables related to school attendance and work. The first variable from this set is the number of days the student reported being absent from school in the past two weeks, which was calculated from the answers to two questions, one on the number of school days attended in the previous week and another on the number of school days attended in the week before that. The second is a variable on whether the respondent was currently working for pay, based on the survey question: “Are you currently working for kwacha or some other form of payment?” Third, we include a binary variable to indicate whether the respondent reported any negative consequences of engaging in paid work or household chores on their schooling. This variable is based on the following four yes/no questions: “In the last 7 days, because of your chores/work have you… A. not been able to go to school at all? B. not been able to go to school for part of the day? C. been tired or unable to concentrate in school? D. not been able to perform well in class or on exams?” If the respondent answered “yes” to any of these four questions, they were categorized as impacted by non‐school work, and not impacted if they answered “no” to all the questions.

For the second pathway, we examined the association between gifting relationships and whether the respondent had married or given birth by the next survey round through a set of discrete time logistic regressions. These analyses are restricted to girls, as too few boys had either married or had a birth by the next survey round to conduct the same analyses. We examine marriage and birth as outcomes rather than independent variables because almost no girls in the sample continue schooling after experiencing these events.

All analyses also include a set of sociodemographic control variables and measures of school experience known to be associated with school dropout. Sociodemographic control variables included age, ethnic group (Yao, Chewa, Lomwe, or Other), current grade of enrollment, and binary indicators of whether the respondent's mother or father had died.

RESULTS

Table 1 presents descriptive statistics on the sample at the survey baseline. Slightly less than half of boys and slightly more than two‐thirds of girls reported that they had not had sex at the time of the first survey round. Around a fifth of boys and 15 percent of girls reported that they had given gifts to or received gifts from their most recent sexual partner. In line with qualitative research, the data show these gifts to be relatively small in value: At round 1, on average, girls estimated the value of the last gift received to be about 460 kwacha (3.3 USD) and boys estimated the value of the last gift given as 360 kwacha (2.6 USD). An additional fifth of boys and almost 5 percent of girls reported that they had ever had sex but did not report giving or receiving gifts with their most recent sexual partner. Finally, 14 percent of boys and 12 percent of girls indicated elsewhere in the baseline survey that they had ever had sex but are missing information about gifting. For all person‐years of observation (not shown), missing values decrease to less than 8 percent of boys and girls, with complementary increases in the percentage of students reporting gift exchanges. Girls were significantly more likely to be enrolled in a higher grade and to be literate than boys. Additionally, boys were significantly more likely than girls to be working for pay and to report that work negatively impacted their school performance. There were no other substantive differences in the sociodemographic characteristics of boys and girls at the first survey round.

TABLE 1

Descriptive statistics, 14‐ to 17‐year‐old students, 2007, MSAS

	Boys (n = 889)	Girls (n = 875)
Gifting		***
Never had sex	46.01	68.57
No gifting	17.66	4.69
Gifting	22.16	14.63
Missing	14.17	12.11
Poor	70.64	67.77
Days absent in two weeks (SD)	1.25 (2.01)	1.39 (2.22)
Working for pay	38.02	10.29***
Consequences of work on school	25.65	19.09**
Mother dead	12.82	12.80
Father dead	22.38	21.60
Age (SD)	15.24 (0.90)	15.11 (0.88)**
Ethnic group
Yao	40.83	39.54
Chewa	20.36	20.57
Lomwe	24.52	25.26
Other	16.31	16.11
Expects to attend secondary	93.48	91.54
Current grade		***
Std. 4	12.15	5.49
Std. 5	18.45	15.77
Std. 6	20.47	22.97
Std. 7	23.96	26.74
Std. 8	24.97	29.03
Academic skills		**
Not literate nor numerate	10.24	6.06
Literate, not numerate	69.29	75.89
Literate and numerate	20.47	18.06
Repeated last grade	14.96	16.00

p < 0.05. **p < 0.01. ***p < 0.001.

We first examined the factors associated with girls and boys being in gifting or non‐gifting relationships, illustrated in Table 2. For both girls and boys, we found that gifting is more a function of the partner's characteristics, which in turn reflect the level of seriousness of a relationship. Both girls and boys were significantly more likely to be in gifting relationships relative to non‐gifting relationships if they had had sex more than once with their partner or their partner was also enrolled in school, although the latter association was only marginally significant. They were less likely to be in a gifting relationship if their partner was more than 3 years older or was classified by the respondent as not being a boyfriend or girlfriend. Altogether, these results suggest that gifts are exchanged in more serious and involved relationships.

TABLE 2

Average marginal effects, discrete time logit regression, likelihood of being in a non‐gifting relationship relative to a gifting relationship, 2007–2010, MSAS

	Boys		Girls
	dy/dx	SE	dy/dx	SE
Nonpoor	0.038	0.035	−0.068	0.044
Mother dead	0.053	0.056	0.017	0.057
Father dead	0.022	0.050	0.045	0.040
Age	−0.009	0.032	0.042	0.026
Ethnic group (ref. Yao)
Chewa	0.071	0.060	0.029	0.068
Lomwe	0.086	0.043	0.066	0.071
Other	−0.015	0.073	−0.006	0.065
Expects to attend secondary	0.007	0.089	0.110	0.051*
Grade (ref. Std. 8)
Std. 4	0.009	0.107	0.175	0.141
Std. 5	−0.020	0.073	−0.039	0.066
Std. 6	−0.058	0.050	0.003	0.055
Std. 7	−0.064	0.046	−0.091	0.053
Skills (not literate or numerate)
Literate, not numerate	0.001	0.075	−0.035	0.093
Literate and numerate	0.063	0.108	0.005	0.104
Repeated last grade	−0.033	0.068	0.015	0.053
Partner type (ref. girl/boyfriend)
Acquaintance	0.206	0.068**	0.390	0.109**
Hit and run	0.239	0.092*	0.127	0.065
Relative	0.407	0.081***	0.213	0.101*
Someone else	0.074	0.103	0.211	0.056***
Partner age difference (ref. same)
Younger	0.005	0.069	0.171	0.139
1–2 years older	0.075	0.062	0.012	0.053
3+ years older	0.139	0.057*	0.185	0.063**
Don't know	0.241	0.083**	0.099	0.104
Partner enrolled in school	−0.070	0.040	−0.111	0.066
Don't know	−0.107	0.073	−0.112	0.067
Partner's education level (ref. none)
Some primary	−0.092	0.067	−0.061	0.085
Completed primary	−0.122	0.090	0.016	0.101
Secondary or higher	−0.067	0.097	−0.087	0.090
Don't know	−0.047	0.095	0.086	0.102
Number of times R had sex with partner (ref. once)
2–4 times	−0.148	0.050**	−0.149	0.053**
5 times or more	−0.134	0.065*	−0.234	0.045***
Missing	–	–	–	–
Round (ref. 1)
2	0.021	0.047	0.024	0.033
3	0.057	0.070	−0.066	0.126
Number of observations	721		295
F	2.84**		1.24

p < 0.05. **p < 0.01. ***p < 0.001.

Table 3 presents the average marginal effects from the discrete time logit regressions, which examine the association between relationship type and school dropout. These models clearly indicate that both boys who gave gifts and girls who received gifts were significantly more likely to drop out of school relative to their peers who had never had sex. Gifting increased the probability of school dropout for boys by 9.3 percent, both with and without the school performance variables. For girls, the marginal effect of gifting on dropping out of school was almost twice as high as that for boys, at 16.4 percent. This figure declined to 15 percent when the school performance variables were included, suggesting that for girls, poor school performance only explains a small part of the association between gifting and drop out. In contrast, the likelihood of dropout for boys and girls who did not participate in gifting with their last sexual partner did not significantly differ from those who had never had sex. Youth with missing/inconsistent values for gifting also did not have significantly different probabilities of school dropout, suggesting that inconsistent reporting of sexual experience was not underreporting participation in gifting relationships.

TABLE 3

Average marginal effects, discrete time logit regression, school dropout, 2007–2010, MSAS

	Boys				Girls
	Model 1		Model 2		Model 3		Model 4
Gifting (ref. Never had sex)
No gifting	0.011	0.028	0.013	0.026	−0.043	0.054	−0.050	0.050
Gifting	0.093	0.027**	0.093	0.027**	0.164	0.034***	0.147	0.033***
Missing	−0.015	0.035	−0.016	0.037	0.010	0.055	0.011	0.052
Nonpoor	−0.093	0.025**	−0.089	0.025**	−0.072	0.022**	−0.063	0.023*
Mother dead	0.001	0.034	−0.002	0.032	−0.014	0.040	−0.001	0.037
Father dead	0.065	0.019**	0.064	0.019**	−0.010	0.033	−0.033	0.032
Age	0.030	0.013*	0.029	0.013*	0.049	0.017**	0.041	0.017*
Ethnic group (ref. Yao)
Chewa	0.015	0.031	0.016	0.031	−0.022	0.040	−0.025	0.040
Lomwe	−0.011	0.025	−0.007	0.024	0.014	0.040	0.012	0.039
Other	−0.041	0.045	−0.034	0.044	0.051	0.040	0.045	0.040
Expects to attend secondary			−0.214	0.059**			−0.198	0.069**
Grade (ref. Std. 8)
Std. 4	0.100	0.044*	0.062	0.049	0.062	0.077	−0.009	0.079
Std. 5	0.070	0.039	0.032	0.037	0.065	0.055	0.010	0.054
Std. 6	0.039	0.036	0.008	0.038	−0.012	0.039	−0.026	0.043
Std. 7	−0.043	0.027	−0.055	0.028	−0.071	0.035	−0.083	0.038*
Skills (Not Literate or Numerate)
Literate, not numerate			−0.035	0.033			−0.167	0.070*
Literate and numerate			−0.104	0.038**			−0.173	0.086*
Repeated last grade			−0.006	0.026			0.006	0.037
Round (ref. 1)
2	0.092	0.024***	0.097	0.026***	0.077	0.034*	0.084	0.034*
3	0.092	0.032**	0.101	0.032**	0.062	0.075	0.064	0.069
Number of observations	1,641		1,641		1,342		1,342
F	7.90***		9.45***		4.02***		5.33***

p < 0.05. **p < 0.01. ***p < 0.001.

Our analyses also show that poverty had similar effects on dropout for boys and girls. When controlling for school performance, the probability that a poor boy would drop out of school was around 9 percent higher than that of a nonpoor boy, and the probability that a poor girl would drop out was 6.3 percent higher compared to a nonpoor girl. Boys and girls were both significantly less likely to drop out of school if they expected to attend secondary school and if they were literate and numerate. Boys, but not girls, were also significantly more likely to drop out if their father had died.

In a third regression model, we estimated the interaction between household poverty status and gifting. The key results from this model are presented in Table 4, which shows the average marginal effects of being nonpoor relative to poor on the likelihood of school dropout by relationship status controlling for the other covariates presented in Table 3. These results clearly show that the impact of poverty on dropout varies by relationship status and gender. Among both girls and boys who never had sex, nonpoor students were significantly less likely to drop out of school than poor students. In contrast, among both girls and boys who had ever had sex and exchanged gifts with their last partner, the probability of drop out did not significantly differ by poverty status. The experiences of girls and boys diverge, however, when we examine the average marginal effect of being nonpoor on students who ever had sex but did not exchange gifts with their last partner. For girls, the probability of dropout did not differ significantly by poverty status, whereas nonpoor boys had a significantly lower probability of dropout than poor boys who had sex but did not exchange gifts.

TABLE 4

Average marginal effect of being nonpoor relative to poor by gifting with partner

Girls	dy/dx	SE
No sex	−0.073	0.029*
No gifts	−0.001	0.126
Received gifts	−0.003	0.051
Gifts NA	−0.153	0.121
Boys
No sex	−0.085	0.029**
No gifts	−0.147	0.032***
Gave gifts	−0.060	0.066
Gifts NA	−0.064	0.066

Figure 1 shows the predicted probabilities of dropping out of school, predicted from this model and based on a respondent who is 15 years old, enrolled in standard 7, is literate and numerate, did not repeat the last grade, is a member of the Yao ethnic group, and has surviving parents. The figure shows that the predicted probabilities of dropout across all categories were lower for boys than for girls. More than a fifth of poor girls who never had sex were predicted to drop out of school in any year, in contrast to 16 percent of nonpoor girls who had never had sex, a statistically significant difference. There was also a relatively large difference by poverty status in the likelihood of dropout for girls who reported inconsistent information about their sexual relationships and were therefore missing the gifting data, although, as shown in Table 4, this difference was not statistically significant. In contrast, for girls who had sex—both those who reported receiving gifts and those who did not—there was no difference in the probability of school dropout between poor and nonpoor girls. However, the probability of dropout for girls who received gifts was about twice as high as the predicted probability for girls who had sex but did not receive gifts: Approximately 34 percent of girls were predicted to drop out of school if they had received gifts from their most recent sexual partner compared to 17 percent of girls who did not. When compared to the predicted probabilities of dropout for girls who had never had sex, these results suggest that the protective benefits of not being poor were effectively erased.

FIGURE 1

Predicted probability of school dropout by poverty status and gifting with partner

As with girls, the predicted probability of dropout differed significantly by poverty level for boys who had never had sex, but the difference by poverty status was larger for boys who had sex but did not give gifts to their last sexual partner; the probability of dropout for poor boys in this group was just above 15 percent compared to around 4 percent for nonpoor boys. Poor boys who reported giving gifts to their most recent sexual partner were predicted to have a higher likelihood of dropping out of school than nonpoor boys (20 percent compared to 15 percent), although this difference was not statistically significant. These results suggest that gifting, but not being sexually active generally, erases the protective effect of being nonpoor for boys.

Finally, we explored some potential mechanisms for the association between gifting and school dropout, including school absence, working for pay, and negative consequences of engaging in non‐school work. Results from these analyses are presented in Table 5 for boys and Table 6 for girls. For boys, school absence increased the marginal probability of school dropout. Each additional day absent from school was associated with a 1 percent higher marginal probability of school dropout. In contrast, there was no association between days absent from school and dropout for girls. Additionally, working for pay and experiencing negative consequences of engaging in paid work—school attendance, concentration, and performance—also increased the probability of dropout for boys, but these associations were only marginally significant. For girls, working for pay itself was not associated with dropout, though girls who reported negative consequences of chores/paid work were also marginally significantly more likely to drop out. Including these three pathway variables, however, did not affect the estimated association between gifting and dropout for either girls or boys.

TABLE 5

Average marginal effects, discrete time logit regression, school dropout with mediating variables, boys, 2007–2010, MSAS

	Model 1		Model 2
	dy/dx	SE	dy/dx	SE
Gifting (ref. Never had sex)
No gifting	0.006	0.024	0.006	0.024
Gifting	0.088	0.028**	0.087	0.028**
Missing	−0.013	0.038	−0.015	0.038
Nonpoor	−0.088	0.025**	−0.087	0.025**
Days absent in past two weeks	0.013	0.005**	0.012	0.005*
Working for pay	0.004	0.002
Consequences of work on school			0.040	0.022
Mother dead	−0.001	0.030	−0.002	0.029
Father dead	0.060	0.019**	0.060	0.019**
Age	0.029	0.013*	0.027	0.013*
Ethnic group (ref. Yao)
Chewa	0.008	0.028	0.007	0.029
Lomwe	−0.007	0.024	−0.010	0.024
Other	−0.038	0.044	−0.041	0.041
Expects to attend secondary	−0.202	0.058**	−0.195	0.058**
Grade (ref. Std. 8)
Std. 4	0.053	0.049	0.050	0.047
Std. 5	0.022	0.036	0.024	0.036
Std. 6	0.001	0.038	−0.001	0.038
Std. 7	−0.059	0.029*	−0.060	0.029*
Skills (not literate or numerate)
Literate, not numerate	−0.034	0.032	−0.033	0.034
Literate and numerate	−0.105	0.038**	−0.102	0.039*
Repeated last grade	−0.005	0.027	−0.006	0.027
Round (ref. 1)
2	0.101	0.026***	0.099	0.025**
3	0.102	0.033**	0.104	0.032**
Number of observations	1641		1641
F	8.59***		8.61***

p < 0.05. **p < 0.01. ***p < 0.001.

TABLE 6

Average marginal effects, discrete time logit regression, school dropout with mediating variables, girls, 2007−2010, MSAS

	Model 1		Model 2
	dy/dx	SE	dy/dx	SE
Gifting (ref. Never had sex)
No gifting	−0.050	0.051	−0.047	0.051
Gifting	0.146	0.033***	0.144	0.033***
Missing	0.012	0.052	0.006	0.050
Nonpoor	−0.064	0.023**	−0.063	0.023**
Days absent in past two weeks	−0.003	0.005	−0.005	0.005
Working for pay	−0.011	0.040
Consequences of work on school			0.058	0.033
Mother dead	0.000	0.037	0.001	0.037
Father dead	−0.033	0.033	−0.036	0.033
Age	0.041	0.017*	0.041	0.016*
Ethnic group (ref. Yao)
Chewa	−0.026	0.040	−0.022	0.040
Lomwe	0.012	0.039	0.011	0.039
Other	0.046	0.041	0.047	0.040
Expects to attend secondary	−0.202	0.067**	−0.193	0.066**
Grade (ref. Std. 8)
Std. 4	−0.009	0.080	−0.015	0.078
Std. 5	0.010	0.055	−0.000	0.054
Std. 6	−0.026	0.043	−0.033	0.044
Std. 7	−0.082	0.038*	−0.086	0.039*
Skills (not literate or numerate)
Literate, not numerate	−0.169	0.070*	−0.167	0.070*
Literate and numerate	−0.176	0.085*	−0.177	0.085*
Repeated last grade	0.006	0.037	0.003	0.037
Round (ref. 1)
2	0.082	0.034*	0.078	0.033*
3	0.068	0.063	0.052	0.066
Number of observations	1342		1342
F	4.92***		5.26***

p < 0.05. **p < 0.01. ***p < 0.001.

Another potential explanation is that gifting indicates movement toward parenthood and marriage, which in turn could increase the likelihood of dropout. Fewer than 1 percent of boys had married or had a child by the next survey round, so we focus these analyses on girls. We examined these pathways with the average marginal effects illustrated in Table 7. None of the sociodemographic variables were associated with being married at the next survey round. Girls in gifting relationships were significantly more likely to be married in the subsequent survey round than girls who had never had sex. In contrast, girls in non‐gifting relationships were less likely to be married, although the association was only marginally significant. Older girls had significantly higher probabilities of having a birth, but none of the other sociodemographic variables were associated with births. Girls in gifting relationships had marginally significantly higher probabilities of having a birth relative to girls who had never had sex, but there was no significant association between being in a non‐gifting relationship and birth.

TABLE 7

Average marginal effects, discrete time logit regression, marriage and birth, girls, 2007–2010, MSAS

	Marriage		Birth
	dy/dx	SE	dy/dx	SE
Gifting (ref. Never had sex)
No gifting	−0.054	0.030	−0.005	0.023
Gifting	0.123	0.029***	0.052	0.028
Missing	−0.018	0.036	0.014	0.033
Nonpoor	−0.040	0.025	0.013	0.015
Mother dead	0.018	0.029	0.019	0.018
Father dead	−0.022	0.024	0.001	0.019
Age	0.002	0.012	0.018	0.007*
Ethnic group (ref. Yao)
Chewa	−0.053	0.035	−0.018	0.019
Lomwe	−0.048	0.031	−0.036	0.021
Other	−0.041	0.045	0.009	0.021
Expects to attend secondary	−0.074	0.054	−0.027	0.038
Grade (ref. Std. 8)
Std. 4	−0.067	0.056	–	–
Std. 5	0.016	0.043	0.000	0.029
Std. 6	−0.034	0.034	0.016	0.025
Std. 7	−0.029	0.037	−0.003	0.016
Skills (not literate or numerate)
Literate, not numerate	0.009	0.047	−0.010	0.039
Literate and numerate	−0.016	0.058	−0.016	0.041
Repeated last grade	0.014	0.027	0.018	0.015
Round (ref. 1)
2	0.007	0.024	0.005	0.015
3	−0.011	0.056	−0.001	0.036
Number of observations	1344		1288
F	4.19***		2.62**

p < 0.05. **p < 0.01. ***p < 0.001.

DISCUSSION

In this study, we demonstrate the importance of considering whether gifts were exchanged as well as the interaction between poverty level and gender when studying adolescent sexual relationships and their consequences. Prior studies on sexual relationships and school dropout in sub‐Saharan Africa have examined differences by gender (Clark and Mathur 2012; Frye 2017; Sunny et al. 2019), but not how the consequences of sexual relationships may vary by gender and poverty level. We found that both boys who gave gifts and girls who received gifts were significantly more likely to drop out of school relative to their peers who had never had sex. In line with prior research, the association was larger for girls. We also found that poverty had a similar association with school dropout for boys and girls with the association slightly higher amongst boys. Gifting relationships, however, erased the significant protective effects of being nonpoor for girls and boys. Additionally, sexual partnerships that did not involve gift exchange—locally a signal of less serious or potentially more exploitative relationships—also erased the difference by poverty status for girls but not for boys.

In terms of school dropout, how do gifting relationships erase the benefits of being nonpoor for both girls and boys? And why do the risks of being in a non‐gifting relationship differ by poverty level for boys, but not for girls? Based on prior research, we expected that there would be differences by poverty level for boys, driven in part by the instrumental cost of gift‐giving: Poor boys would need to spend time earning money to fund gifting and that this time, in turn, negatively impacts their schooling performance. Poulin's (2007) study features young men talking about taking on additional time‐consuming labor in order to pay for gifts, and Frye's (2017) study finds that sexual relationships are associated with absenteeism from school for boys, but not for girls. In our testing of mechanisms, we found that while school absence and the consequences of work (including absence but also performance and concentration) for boys were positively associated with school dropout, though the latter only marginally, they did not explain the association between gifting relationships and dropout. One potential explanation is that the work variables in the dataset may be obscuring variation in the intensity of work carried out by respondents. Another is that even nonpoor boys may face difficulties finding money to cover gifting and engage in similar strategies to poor boys. Non‐gifting sexual relationships, on the other hand, may affect boys relatively minimally, leaving the negative impact of coming from a poor background intact.

That there was no significant difference in the likelihood of dropout by poverty level for girls in sexual relationships, both gifting and non‐gifting, suggests that sexual relationships generally are a risk factor for girls’ school trajectories. This may be explained by the fact that, though parents and school administrations are highly disapproving of school girls being in any type of sexual or romantic relationship, existing research does not suggest that the penalties are especially harsh for poor girls (Frye 2017; Grant 2012). Further, poor girls are not significantly more likely to be in a gifting relationship than nonpoor girls, suggesting that these relationships are not necessarily motivated by economic desperation. However, the likelihood of dropping out of school is higher for girls in gifting relationships compared to girls in non‐gifting sexual relationships, which may be related to how gifting itself makes relationships more visible, alerting others—including school authorities and parents—to girls’ engagement in such relationships. Teachers keep a close eye out for relationships in school and are quick to punish any suspects, particularly girls (Frye 2017). Additionally, parents may not be as keen to pay the costs of a daughter's education, no small investment in this high‐poverty context, or may be less adamant about encouraging girls to stay in school (Grant, 2012).

Additionally, for girls, gifting's association with dropout does appear in part to reflect an intensification of relationships. Being in a gifting relationship was associated with a higher likelihood of being married and having a birth at the subsequent round. In contrast, non‐gifting relationships were negatively associated with marriage, but also associated with school dropout. If sexual activity prompts dropout but does not lead to marriage, it raises questions about girls’ economic survival strategies if they are not in school and have limited work opportunities (Kendall and Silver 2014). However, one limitation to note in these marriage and birth analyses is that we are unable to identify whether a girl married or had a child with the partner who was giving her gifts. Thus, it is difficult to determine the causal role of gifting relationships in these trajectories, particularly given that courtships leading to marriage in rural Malawi can be extremely quick (Bertrand‐Dansereau and Clark 2016). A gifting relationship may signal an imminent marriage or, equally, a gifting relationship may cause others to discourage or disinvest in the schooling of a girl who then drops out and marries as a result.

This study has some other limitations. First, the nature of the questions means that we are only capturing gifting that took place with the respondent's most recent sexual partner and not instances of gifting in relationships that did not involve sex or that occurred with previous or concurrent sexual partners. As such, we are likely underestimating the prevalence of gifting in the sample. However, focusing on gifting within sexual relationships also limited our study to more involved instances of gifting excluding, for example, gifting to show romantic interest but that might not have led to a relationship. Second, again due to the nature of the data, we have limited details on what type of gifts were exchanged as well as students reported reasons for leaving school. Finally, we focus on relationships reported the survey round prior to when school dropout is measured. Although this time lag clarifies the causal ordering of relationships and school dropout, it does not capture relationships during the intersurvey interval that may have more acute consequences for school dropout.

Nonetheless, this study provides some novel findings for research on gifting and sexual relationships amongst adolescents in sub‐Saharan Africa. First, it shows that gifting, independent from being in a sexual relationship, raises the likelihood of dropout for both boys and girls. Indeed, so much so that it effectively erases the protective advantage of being nonpoor on school dropout. We also find some differential impact of poverty by gender with poor boys in sexual relationships without gift exchange more likely to drop out of school than their nonpoor peers. There is no poverty differential for girls in such relationships. This study thus points to the value of differentiating between types of adolescent sexual relationships as well as considering how the gendered risks to drop out may differ for subgroups of girls and boys, poor and nonpoor but also along other axes of difference. These results have potential implications for policy trying to reduce school dropout, showing the importance of supporting poorer students financially in their studies, but also encouraging students regardless of their known or suspected relationship status to continue to pursue their education. While most policy work focuses on the negative consequences of girls’ receiving gifts, this research also shows the impact on boys, suggesting that frank and open, but not penalizing, co‐ed discussions about these dynamics in schools might be fruitful.

In sum, qualitative research has shown that the dynamic in sexual relationships involving exchange is often more complex than the standard dichotomy of boys and men as powerful gift‐givers and girls and women as powerless gift‐receivers. The gender–poverty interactive approach presented in this paper is one additional way that studies using survey data can take this nuance into account. As the socioeconomic gap in schooling attainment outpaces the gender gap in the Global South (Jones and Ramchand 2016), future research will need to include greater attention to the nuanced interactions between poverty and adolescent romantic relationships.