AGE MODERATES THE ASSOCIATION BETWEEN MICROFINANCE MEMBERSHIP AND PHYSICAL ABUSE, RELATIONSHIP POWER, AND TRANSACTIONAL SEX IN HAITIAN WOMEN

Abstract

Microfinance interventions may have differential effects on relationship dynamics among subpopulations of women. We estimated the association between microfinance participation duration and physical abuse, relationship power, and transactional sex in a sample of Haitian women (n=304). Further, we tested for moderation by age. In older women, microfinance tended to be associated with reduced risk of violence, low relationship power, and transactional sex. These associations were not observed for younger women. Thus, older Haitian women may benefit from microfinance in ways that younger women do not. Future studies should examine whether additional training and resources could improve outcomes in younger women.

INTRODUCTION

Intimate partner violence (IPV), defined as physical, sexual, or psychological abuse and controlling behaviors within an intimate relationship, is a significant global problem that disproportionately affects women (World Health Organization, 2012). Globally, an estimated 30% of women who’ve ever had an intimate partner have experienced IPV in their lifetime, with notable variation by country and region (Devries et al., 2013; Garcia-Moreno et al., 2006; World Health Organization, London School of Hygiene and Tropical Medicine, & South African Medical Research Council, 2013). Complex relationships between relationship power, IPV, and transactional sex have been documented (Dunkle et al., 2004; Dunkle et al., 2007; Jewkes, Dunkle, Nduna, & Shai, 2010). Notably, an increased risk of IPV has been repeatedly associated with both formal sex work and informal transactional sex, which is defined as an exchange of money or goods for sex (Dunkle et al., 2007; Dunkle et al., 2006; Jewkes, Morrell, Sikweyiya, Dunkle, & Penn-Kekana, 2012). In addition to the direct adverse health effects of engaging in transactional sex and experiencing IPV, both are established risk factors for STIs such as HIV (Fielding-Miller, Dunkle, Cooper, Windle, & Hadley, 2016; Jewkes, Dunkle, Nduna, & Shai, 2012; Wamoyi, Stobeanau, Bobrova, Abramsky, & Watts, 2016).

Women living in poverty have an elevated risk of experiencing IPV (Dalal, Rahman, & Jansson, 2009; Diop-Sidibe, Campbell, & Becker, 2006; Kim et al., 2007; World Health Organization, 2005) and engaging in transactional sex (i.e. survival sex) (Miller et al., 2011). In Haiti, the poorest country in the Americas, approximately 60% of the population lives below the poverty line of $2.41 a day and nearly a quarter live in extreme poverty, or less than $1.23 a day (World Bank & Observatoire National de la Pauvreté et de l’Exclusion Sociale, 2014). The prevalence of IPV is also high in Haiti. Of ever-partnered women, 34% report experiencing IPV in their lifetimes and 22% in the last year, according data collected in the 2016–2017 Demographic and Health Survey in Haiti (Institut Haïtien de l’Enfance & ICF International, 2018).

In Haiti, as well as in other low- and middle-income countries, microfinance-based interventions are commonly deployed, disbursing small loans for income generation enterprises and providing other basic financial services to poor individuals (Khandker, 2005; Swain, Van Sanh, & Van Tuan, 2008). Microfinance has been lauded not only as a key tool for poverty reduction but also, when targeted to women, as an important means of improving their status and power in the household and in society more broadly (Luke & Munshi, 2011; Mayoux, 1998; Weber & Ahmad, 2014). However, there is controversy surrounding the impact of microfinance, particularly on relationship dynamics and IPV.

This study is grounded in the economic theories of household bargaining, which posit that the bargaining power of actors within partnerships is dependent upon the existence of available (and accessible) alternatives to the marriage (Aizer, 2007, 2010; S. Anderson & Eswaran, 2009; Farmer & Tiefenthaler, 1997; Manser & Brown, 1980). Theoretically, economically empowered women, such as those participating in microfinance programs, may have improved agency to leave abusive or non-serving relationships, thereby increasing their bargaining power within the relationship (Aizer, 2007; Choice & Lamke, 1997; Gelles, 1976; Johnson, 1992; Kaukinen, 2004; Manser & Brown, 1980; Villarreal, 2007). At least one study has documented increased separations and marriage turnover among women involved in an economic empowerment program compared to women not involved (Bobonis, 2011). Several studies have also found consistent connections between economic empowerment and corresponding increases in intra-household bargaining and decision-making power (Hatzimasoura, Premand, & Vakis, 2017; Huis, Lensink, Vu, & Hansen, 2019). Meanwhile, greater relationship power has been repeatedly connected to safer sex and condom use (Teitelman et al., 2016), reduced risk of HIV (Dunkle et al., 2004; Jewkes et al., 2010), and ability to refuse sex altogether (Sano, Sedziafa, Vercillo, Antabe, & Luginaah, 2018). Improved bargaining and relationship power should allow women to refuse transactional sex with their partners. In regard to violence, these theories postulate that economically empowered will have decreased risk of relationship violence since they have the economic means to leave, or even avoid, violent relationships (Brinkerhoff & Lupri, 1988; Gelles, 1976; Kalmuss & Straus, 1982; Straus, 1973, 2017). In contrast, women may be forced to tolerate violence if they lack other options outside of the partnership or the ability to leave (Brinkerhoff & Lupri, 1988; Gelles, 1976; Kalmuss & Straus, 1982; Straus, 1973, 2017). Finally, there are other constraints on women’s agency and ability to avoid and leave relationships that lie outside of the economic realm, particularly in strongly patriarchal and conservative societies where marriage may be arranged or expected and divorce might be prohibited for women (Khan & Klasen, 2018). In these contexts, it is likely that household bargaining theories may not completely or accurately reflect the relationship between female economic empowerment on relationship dynamics.

Much of the empirical research on microfinance indicates that women’s involvement increases their empowerment (e.g., decision-making power, ability to leave abusive relationships, ability to demand better treatment, etc.) (Deininger & Liu, 2013; Kim et al., 2007; Orton et al., 2016; Schuler & Hashemi, 1994; Schuler, Hashemi, & Riley, 1997) and reduces their risk of experiencing IPV (Bates, Schuler, Islam, & Islam, 2004; Kim et al., 2007; Schuler, Hashemi, Riley, & Akhter, 1996; Yoo-MiChin, 2012). However, other studies have found microfinance participation was associated with an increased risk of exposure to IPV (Dalal, Dahlstrom, & Timpka, 2013; Hadi, 2000; Koenig, Ahmed, Hossain, & Khorshed Alam Mozumder, 2003; Murshid, Akincigil, & Zippay, 2016; Naved & Persson, 2005) or increased justification of IPV, especially among certain subgroups of women (Murshid, 2016; Murshid et al., 2016). Still others have shown initial increases in violence in the first couple years of microfinance membership as gender norms and roles were renegotiated, which subsided with time (Ahmed, 2005), or no association between microfinance participation and IPV at all (Bajracharya & Amin, 2013; Kim et al., 2009). Finally, some have indicated reduced exposure to IPV when the microfinance intervention was combined with “gender transformative interventions,” or interventions aimed at addressing gender inequity by reshaping the social norms, attitudes, and power structures that underpin such inequality in society (Pronyk et al., 2006).

Though a few studies have found education and wealth to be modifiers of the effect of microfinance on IPV in Bangladesh (Dalal et al., 2013; Murshid et al., 2016), no studies have examined the impact of microfinance participation on IPV, relationship power, and transactional sex in the Haitian context. In this study, we hypothesize that Haitian women who have been involved in a microfinance program for a longer period of time will have experienced the benefits of greater economic and financial independence and will, therefore, also have greater relationship power, decreased exposure to physical abuse from their partners, and be less likely to have engaged in transactional sex. Further, based upon the existing literature, we hypothesize there may be effect modification by some socio-demographic variables, such as wealth, education, age, or marital status.

METHODS

Study Population

This study was conducted among female microfinance clients of the microfinance institution, Fonkoze, in the southern peninsula of Haiti. Fonkoze is the country’s largest microfinance organization, has 44 branch offices across the country, targets women almost exclusively, and has been serving microfinance clients since 1994 (Tucker & Tellis, 2005). The foremost service of Fonkoze is the solidarity lending program. In this program, women clients form groups of five and organize together to obtain and repay loans. The range for the solidarity group loans is US$100 to US$1300. In general, the loan sizes and benefits increase as clients’ progress through the program. Further, Fonkoze also incorporates training and educational programs into its structure, including those focused on literacy, life skills, business skills, and health education.

This study was conducted among a random sample of Fonkoze clients served by the branch office in the city of Okay (n=304). The Okay branch office serves over 6500 clients in the city of Okay and surrounding municipalities. The study population was randomly sampled from the Fonkoze membership database of the Okay branch. Eligibility criteria were (1) female gender, (2) being a microfinance client served by the Fonkoze Okay branch office, and (3) being between the ages of 18 and 49 years old.

Data Collection

We conducted a cross-sectional, tablet-based survey in a random sample of 304 female microfinance clients. Data collection was conducted between December 2017 and February 2018 by trained local fieldworkers in the local language of Haitian Creole. Field workers approached and recruited potential participants at their homes. Those who met the eligibility criteria, agreed to participate, and completed a written consent form were asked to self-report various socio-demographic information, the length of their microfinance involvement, sexual behaviors, relationship power dynamics, and experience of physical abuse. This study was approved by the Human Subjects Office of Indiana University (Protocol #1705661852).

Key Variables

Exposure

This study examined the association between amount of microfinance exposure and physical abuse, relationship power, and transactional sex. We measured amount of microfinance exposure with individual duration of microfinance membership. As Fonkoze provides larger loans and more training opportunities to clients as they progress through the program over time, the duration of client membership is a useful proxy for the amount of microfinance services a participant has received. Participants reported the date they became Fonkoze clients and the exposure variable was created by subtracting this date from the date of their interview and dividing by the average number of days in a month, which created a continuous microfinance duration variable measured in months. We categorized this variable into low (≤12 months) versus high microfinance exposure (>12 months) using a cut point of one year. This cut point was chosen since the literature indicates that the effects of microfinance interventions stabilize around one year of participation (Ahmed, 2005).

Outcomes

The primary outcomes of interest were physical abuse in the last year, relationship power, and transactional sex. Physical abuse in the last year was measured by 12 questions, which utilized behaviorally-specific queries to assess exposure to various types of physical abuse in the lifetime and in the past year. If participants reported yes to any of the questions and then also yes to the follow-up question asking if it had happened in the last year then they were coded as having experienced physical abuse in the last year. These behaviorally-specific intimate partner violence questions were drawn from the physical violence subscale of the domestic violence questionnaire utilized by the WHO’s Multi-Country Study on Women’s Health and Domestic violence (World Health Organization, 2005), which was developed from the Conflict Tactics Scale (Straus, Hamby, BoneyMcCoy, & Sugarman, 1996). The WHO evaluated the psychometric properties of the questionnaire and concluded that the items for each subscale had appropriate internal consistency and, therefore, were reliable and valid measures for each type of intimate partner violence (World Health Organization, 2005). Another outcome of interest was relationship power. This variable was derived from 12 survey questions which queried various aspects of relationship power dynamics and asked participants to rate their agreement or disagreement on a 3 item Likert scale: (1) Agree a lot, (2) Somewhat agree, and (3) Do not agree at all. Example questions are: “If I asked my partner to use a condom, he would get angry.” and “If My partner tells me who I can spend time with.” These questions were asked with reference to the participants’ most recent partners. Those without a partner in the last 24 months were not asked these questions. These relationship power questions were drawn from the Sexual Relationship Power Scale (Pulerwitz, Gortmaker, & DeJong, 2000). Answers were then summed into a continuous index variable ranging from 1 to 36. The median was used as a cut point for the creation of a dichotomous low/high relationship power variable (median=25). The median was used as a cut point in order to ensure a mostly equal distribution between the two new groups. Higher scores indicated that the woman has more power in the relationship. Lower scores indicated the woman has little power in the relationship. This variable intended to estimate the level of abusive and controlling behavior in the relationship outside of direct physical violence since many of the questions ask about these types of behaviors.

The final outcome of interest was transactional sex. Informal transactional sex is defined as an implicit exchange of money or goods for sex where the woman feels obligated to have sex because of financial or material gifts (Abels & Blignaut, 2011; Chatterji, Murray, London, & Anglewicz, 2005; Cluver, Orkin, Boyes, Gardner, & Meinck, 2011; Jewkes, Morrell, et al., 2012; Maganja, Maman, Groves, & Mbwambo, 2007; Ranganathan et al., 2016) and can occur in any type of sexual relationship (one-time encounter, casual, committed relationships, etc.) (Maganja et al., 2007). In this study, we defined transactional sex as the self-reported receipt of money or gifts from the most recent partner and feeling obligated to have sex with this partner in return for these received money or goods. This variable aimed to assess informal transactional sex that sometimes occurs between sex partners but is not as overt an exchange of money for sex as more formal sex work. In the creation of this variable, individuals were only included if they had reported at least one sex partner in the last year. This variable was based on three queries in the survey about transactional sex behavior. The final variable was coded dichotomously (yes/no).

Covariates

Covariates queried in the survey included age (in years), marital status (married/living together, divorced/separated, never married), ever attend school (yes/no), education level (preschool, primary, secondary, more than secondary), literacy (yes/no; defined as ability to read a sentence in Creole), household size (defined as number of people living in participant’s household not including the participant themselves), and assets (defined as the sum of the values, at time of purchase in Haitian gourdes, of 20 possible household items including oven, television, radio, cell phone, etc.). The asset index variable was categorized in quartiles, from lowest (Q1) to highest (Q4). Covariates included in the adjusted models to control for potential confounding were household asset quartile, marital status (further categorized as married/living together vs. divorced/separated/never married), and education level (further categorized as preschool/primary vs. secondary/more than secondary). In the models the effect measure modifier, age, was categorized from a continuous variable into a dichotomous variable with cut point at the median age (median age=36). This resulted in two age groups: younger (aged 20–36 years) and older (aged 37–49 years).

Statistical Analysis

First, frequencies for categorical variables and means for continuous variables were calculated for basic sample demographic characteristics in addition to the distribution of the outcomes of physical abuse, relationship power, and transactional sex in the sample. These frequencies were stratified by the exposure, a dichotomous variable of microfinance membership length (≤12 months vs. >12 months). Differences in frequencies between the two strata were calculated using Pearson’s chi-square. We used Fisher’s exact test for frequencies with small cell sizes (i.e. those containing <5 observations). For continuous variables, Student’s t-test was used to compare the differences in means between the two strata.

We used log-binomial models to assess the relationship between length of microfinance membership and the three categorical outcomes, including experience of physical abuse in the last year, relationship power, and transactional sex. These models were run on the entire sample with and without interaction terms between the microfinance membership duration and each of the four sociodemographic variables, including wealth (measured by assets), education, age, or marital status, for each of the three dichotomous outcomes to confirm whether the effect estimates for the strata were statistically different from each other. Wald test p-values for the interaction terms with an a priori cut point of p<0.20 were used to determine statistical significance. Models with significant interactions are reported in this study while non-significant models are excluded from the results. Models were then adjusted for potential confounders, including marital status, education level, and household asset quartile. For the confounder-adjusted models, we utilized Poisson regression models with robust error variance due to convergence issues in the adjusted log-binomial models (Barros & Hirakata, 2003; Zou, 2004). All statistical analyses were conducted using R version 3.6.2 (“Dark and Stormy Night”) (R Core Team, 2019) and SAS 9.4 (Cary, NC) (SAS Institute Inc., 2014).

RESULTS

Demographically, this sample had a mean age of 36 years (SD=7.7), which did not vary significantly between the two microfinance membership strata (≤12 months vs. >12 months) (Table 1). Most women were married or living with their partner (55%). Most of the participants (83%) had attended some schooling with 44% having completed secondary school or more, though few had obtained education past the secondary level. Additionally, household size, excluding the participant, was about 4 people (mean=3.9, SD=1.9). Finally, 12.4% of interviewed women had experienced physical abuse in their lifetime and 10.9% had experience it in the last year. Meanwhile, over 40% reported transactional sex in the last year with no significant differences by microfinance membership. As for relationship power, the mean score was 24.6 (SD=5.8) from a range of 1 through 36 with higher indicating better relationship power.

Table 1.

Characteristics of the study population, stratified by microfinance membership duration

		Microfinance membership
	Total (n=304)	≤12 months (n=135)	>12 months (n=169)	P-value*
Socio-demographic characteristics
	Mean (SD)	Mean (SD)	Mean (SD)
Age, years	36.0 (7.7)	35.6 (8.3)	36.3 (7.1)	0.44a
Household size	3.9 (1.9)	3.8 (1.8)	4.1 (1.9)	0.18b
Number of children	3.0 (1.8)	3.0 (1.9)	2.9 (1.7)	0.80b
	n (%)	n (%)	n (%)
Marital status				0.63
Married/Living together	164 (55.0)	73 (54.5)	91 (55.5)
Divorced/separated	26 (8.7)	14 (10.5)	12 (7.3)
Never married	108 (36.2)	47 (35.1)	61 (37.2)
Missing	6	1	5
Education level				0.06
None	51 (16.9)	29 (21.5)	22 (13.3)
Preschool/Primary	119 (39.5)	56 (41.5)	63 (38.0)
Secondary or more	131 (43.5)	50 (37.0)	81 (48.8)
Missing	3	0	3
Literacy				0.10
Yes	223 (74.1)	93 (69.4)	130 (77.8)
No	78 (25.9)	41 (30.6)	37 (22.2)
Missing	3	1	2
Asset Quartiles				0.30
Q1	76 (25.2)	40 (29.9)	36 (21.4)
Q2	75 (24.8)	32 (23.9)	43 (25.6)
Q3	76 (25.2)	34 (25.4)	42 (25.0)
Q4	75 (24.8)	28 (20.9)	47 (28.0)
Missing	2	1	1
Relational and violence outcomes
	n (%)	n (%)	n (%)
Physical abuse in last year				0.96
Yes	30 (10.9)	14 (11.0)	16 (10.8)
No	245 (89.1)	113 (89.0)	132 (89.2)
Missing	29	8	21
Physical abuse ever				0.80
Yes	34 (12.4)	15 (11.8)	19 (12.8)
No	241 (87.6)	112 (88.2)	129 (87.2)
Missing	29	8	21
Transactional sex				0.50
Yes	112 (41.2)	50 (39.1)	62 (43.1)
No	160 (58.8)	78 (60.9)	82 (56.9)
Missing	32	7	25
	Mean (SD)	Mean (SD)	Mean (SD)
Relationship power	24.6 (5.8)	24.5 (5.9)	24.8 (5.7)	0.70b

^*P-values were derived from Chi-Square tests for categorical variables and Student’s t-tests for continuous variables.

^aSatterthwaite method used because variances were unequal.

^bPooled method used because variances were equal.

In the full population, we found little to no effect of microfinance participation length (dichotomized as ≤12 months vs. >12 months) on our three outcomes of interest [physical abuse prevalence ratio (PR): 0.98, 95% confidence interval (CI): 0.50–1.93; low relationship power PR: 0.90, 95% CI: 0.69–1.17; and transactional sex PR: 1.10, 95% CI: 0.83–1.47] (Table 2). However, due to previous research indicating effect modification on various sociodemographic variables, we tested for effect modification in our models. There was significant multiplicative effect measure modification by age on the relationship between microfinance duration and all three outcomes: experiencing physical abuse in the last year, having low relationship power, and engaging in transactional sex. This moderating effect was identified using log-binomial regression models, which produced significant Wald test p-values for the interaction terms (i.e. Wald test p-values<0.20). Significant Wald test p-values for the interaction terms in the models indicated that the protective associations we observed in older women were significantly different from the non-protective associations we observed in younger women (Table 2).

Table 2.

The association between length of microfinance membership (≤12 months vs. >12 months) and physical abuse in the last year, relationship power, and transactional sex among Haitian women, stratified by age (n=304)

	Crude models			Adjusted models*
Outcomes	PR (95% CI)	p-value	Wald p-value for interaction term	PR (95% CI)	p-value	Wald p-value for interaction term
Physical abuse in last year
Total	0.98 (0.50–1.93)	0.96	-	1.45 (0.69–3.14)	0.33	-
Age strata			0.04			0.17
Young (18–36)	1.65 (0.74–3.69)	0.22		2.03 (0.89–4.64)	0.09
Old (37–49)	0.26 (0.05–1.24)	0.09		0.53 (0.09–3.43)	0.48
High relationship power
Total	1.11 (0.86–1.42)	0.44	-	1.16 (0.89–1.52)	0.27	-
Age strata			0.05			0.07
Young (18–36)	0.84 (0.57–1.23)	0.37		0.88 (0.60–1.28)	0.50
Old (37–49)	1.39 (0.99–1.96)	0.06		1.47 (0.98–2.21)	0.06
Transactional sex
Total	1.10 (0.83–1.47)	0.51	-	1.05 (0.77–1.43)	0.75	-
Age strata			0.17			0.03
Young (18–36)	1.34 (0.90–1.99)	0.15		1.40 (0.94–2.98)	0.09
Old (37–49)	0.89 (0.59–1.35)	0.59		0.70 (0.43–1.13)	0.14

^*adjusted for assets (proxy for SES), marital status, and education level.

In crude models, the effect of microfinance participation on all three outcomes fell on opposite sides of the null for younger women and older women, when comparing longer microfinance participation (>12 months) to shorter (≤12 months). Among older women, those with longer microfinance participation were significantly less likely to have low relationship power than those with shorter participation [PR: 0.62, 95% CI: 0.40–0.98]. Whereas among younger women, those with longer microfinance participation tended to be more likely to have low relationship power [PR: 1.15, 95% CI: 0.84–1.58]. Even though not significant, longer microfinance participation also seemed to have a protective effect on exposure to physical abuse in the last year among older women [PR: 0.26, 95% CI: 0.05–1.24] and the opposite effect among younger women [PR: 1.65, 95% CI: 0.74–3.69]. Additionally, similar trends were seen for transactional sex with older women seeming to experience benefits from longer microfinance participation while younger women seemed to experience more adverse outcomes with longer microfinance participation compared to shorter participation. Among older women, those with longer microfinance participation tended to have less experience with transactional sex [PR: 0.89, 95% CI: 0.59–1.35] while among younger women, those with longer microfinance participation tended to have greater experience of transactional sex [PR: 1.34, 95% CI: 0.90–1.99] (Table 2). Though some of these effects are not statistically significant, they do reveal significant differential effects (i.e. effect modification) by age on the relationship between microfinance duration and relationship and violence outcomes among this sample of Haitian women. While we also tested for effect modification by wealth (measured by assets), education, and marital status, none of these other variables demonstrated significant modifying effects.

Adjusted models showed similar results to the crude models with effect sizes changing only marginally with the exception of the estimate for physical abuse, which changed more dramatically, but remained non-significant. After adjusting for confounding by household asset quartile, marital status, and education level, longer microfinance participation (>12 months) still tended to have a protective effect on exposure to physical abuse in the last year among older women [PR: 0.86, 95% CI: 0.22–3.40] and the opposite effect among younger women [PR: 2.15, 95% CI: 0.94–4.90] (Table 2). Additionally, similar trends were seen for relationship power and transactional sex with older women and young women still falling on opposite sides of the null. For relationship power, among older women, those with longer microfinance participation tended to be less likely to have low relationship power [PR: 0.57, 95% CI: 0.32–1.00] compared to those with shorter participation whereas among younger women, those with longer microfinance participation tended to be more likely to have low relationship power [PR: 1.10, 95% CI: 0.81–1.51] compared to those with shorter participation when adjusting for household asset quartile, marital status, and education level. Further, among older women, those with longer microfinance participation tended to have less experience with transactional sex [PR: 0.70, 95% CI: 0.42–1.15] while among younger women, those with longer microfinance participation tended to have greater experience of transactional sex [PR: 1.40, 95% CI: 0.93–2.10] when adjusting for household asset quartile, marital status, and education level. Again, though none of these effects are not statistically significant, there is still evidence of significant effect modification by age on the relationship between microfinance duration and our three outcomes.

DISCUSSION

In this study, the first of its kind to analyze the relationship between microfinance and relationship dynamics in Haiti, we found that older women tended to be more likely than younger women to experience protective effects from longer microfinance membership on physical abuse, relationship power, and transactional sex. Specifically, age modified the effect of microfinance membership duration on these three outcomes among Haitian women. However, our results were measured imprecisely due to low sample sizes in the age strata.

Our exploratory findings add to a growing body of literature that indicates that the impact of microfinance, and more broadly the empowerment of women, on intimate partner dynamics may vary in different socio-demographic subgroups of women (Dalal et al., 2013; Murshid et al., 2016). This literature seems to indicate that the relationship between microfinance participation, empowerment, and IPV exposure among women may be dependent upon a variety of contextual factors, which may include culturally prescribed gender norms and roles, family support, existing partner dynamics, and other factors (Devries et al., 2013; Garcia-Moreno et al., 2006; Koenig et al., 2003; Ngo & Wahhaj, 2012; Sabarwal, Santhya, & Jejeebhoy, 2014).

There are several possible explanations that might explain why the effect of microfinance membership may be different among older women compared to younger women. Several of these explanations fall under theoretical models established in the scientific literature. First, household bargaining theories (Aizer, 2007, 2010; S. Anderson & Eswaran, 2009; Farmer & Tiefenthaler, 1997; Manser & Brown, 1980) extend that participation in microfinance activities may decrease a woman’s exposure to adverse relationship and violence outcomes because their financial independence and access to economic opportunities help to balance the power dynamic between partners, recalibrating it to a more equal distribution of power between the husband and the wife (Aizer, 2010; Farmer & Tiefenthaler, 1997; Guarnieri & Rainer, 2018). Expanding upon this theory, women who are financially empowered may be leaving abusive and violent relationships altogether (Gelles, 1976). In this study, older women seem to be benefiting from microfinance participation and are thus able to reduce their exposure to violence from their partners, increase their power within the household, and better refuse transactional sex. These positive associations may be due to the fact that they are less dependent on their partners and are better equipped to fulfill their own basic needs.

Second, older women may be positioned to benefit more from microfinance participation for a variety of reasons (Brody et al., 2015; Vyas & Watts, 2009). Relative resource theory (Bowlus & Seitz, 2006; Goode, 1971; Macmillan & Gartner, 1999; McCloskey, 1996; Melzer, 2002; Whaley & Messner, 2002) argues that female economic empowerment could have a negative effect on heterosexual partnerships with the male partner feeling threatened by the redefinition of gender roles and power relations and, thus, resorting to violence to reestablish and maintain his dominance and control (Finnoff, 2012; Guarnieri & Rainer, 2018; Macmillan & Gartner, 1999). As such, it may be that older women begin with lower relative status to their partners and, therefore, their financial empowerment might establish a more equal distribution of power within the relationship but without presenting as a perceived threat to male authority. On the other hand, younger women may already have more equal status relative to their partners. Thus, their financial empowerment may throw off the balance of power and be perceived as more threatening to their male partners. Previous research has shown that female-dominance in decision-making increases exposure to intimate partner violence among women in Haiti (Gage & Hutchinson, 2006). Additionally, studies elsewhere have shown that both educated and wealthier women, who already have more power in familial decision-making and are more equal to their partners in the relationship, experience increased risk of exposure to intimate partner violence when participating in microfinance interventions compared to those not participating in microfinance services (Dalal et al., 2013). Building upon the previous idea, it may not simply be the renegotiation of relationship power dynamics that increase violence risk but rather, power asymmetries that favor the female partner that lead to increased violence risk for women. Preliminary research has shown that power asymmetries that favor the female partner (i.e. when women are earning more than their partners, are employed while their partners are not, or are more educated than their partners) may be associated with increased IPV risk (K. L. Anderson, 1997; Kaukinen, Meyer, & Akers, 2013; Macmillan & Gartner, 1999; Yount & Carrera, 2006).

Finally, it may be that older women are in more established, even legally-recognized (i.e. wedded), partnerships and their economic empowerment and earnings resulting from their microfinance participation are perceived as an important and helpful contributions to the household. Meanwhile, younger women are likely to be in less established and less official partnerships compared to older women, and thus, their partners may be afraid that they will leave the relationship or feel threatened by their economic independence. Clearly, the existing relationship dynamics of older women and younger women may significantly differ preceding microfinance participation and these and other contextual factors may influence the effects of economic empowerment on physical abuse, relationship power, and transactional sex.

The heterogeneous effects of microfinance suggested here, and demonstrated in the literature, emphasize the necessity of further research to elucidate the causal relationships between microfinance and relationship dynamics and violence and to identify all moderators of these relationships. Specifically, this is important in order to develop and target at-risk subgroups involved in microfinance programs with extra trainings and violence prevention strategies to minimize possible adverse outcomes. One group of researchers has theorized that the variation in the effects of microfinance are due to interactions between household dynamics, individual characteristics of the microfinance participant, and design of microfinance programs (Ngo & Wahhaj, 2012). They propose that involving the male partner, instead of simply targeting women microfinance programs, may have a much more positive impact on household welfare and the client’s wellbeing, particularly in poor settings where traditional and inequitable gender norms are more pervasive (Ngo & Wahhaj, 2012). Other researchers have also emphasized the need for the addition of gender-transformative approaches aimed at both the microfinance client and her male partner (Dworkin, Fleming, & Colvin, 2015; Gibbs, Willan, Misselhorn, & Mangoma, 2012; Green, Blattman, Jamison, & Annan, 2015; Gupta et al., 2013).

Several aspects of the design of the study warrant caution in interpreting the results. First, due to the cross-sectional nature of this data, we were unable to fully disentangle the temporal ordering of the key variables of interest. Some of the variables were collected with reference to specific time points (i.e. physical abuse in the last year) which was useful in placing the outcome in relation to the duration of microfinance exposure. However, studies with longitudinal designs should be used to further understand the directionality of the relationships observed here. Additionally, we cannot compare those exposed to microfinance to those not exposed to these services (i.e. a control group) since we purposefully enrolled only microfinance clients. Accordingly, the variation in the exposure comes from the length of participation in microfinance services. However, a control group may not be appropriate under these circumstances since women who self-select to participate in microfinance services will likely systematically differ from a random sample of the general female population. Our design minimizes the potential for this kind of confounding. The survey was also conducted only among women in one region of Haiti and, thus, this sample may not be representative the overall female population living in Haiti. However, we used a rigorous random sampling method with very high participation rates, maximizing the likelihood that the study population was representative of the target population of female Fonkoze clients served by the Okay branch office. The survey responses in this study were self-reported in an interview setting by a trained fieldworker. Self-report, though a valid method of data collection, can affect results if under- or over-reporting occur due to social desirability and recall bias. This bias may be reduced by use more self-controlled data collection procedures such as Audio Computer-Assisted Self-Interviewing (ACASI) in future studies. Finally, there are several additional considerations that we were unable to address in this study due to its design. In particular, we acknowledge that severity of violence might be of interest in the relationships studied here but, unfortunately, we did not collect and therefore did not have access to this information. Additionally, some analyses that we conducted indicated that there were differences in progression in the microfinance program (as measured by loan size, which generally increases as a member progresses) between those women who had experienced physical abuse in the past year compared to those who had not. This effect was only seen at microfinance membership durations longer than one year. Therefore, it seems that experiencing physical abuse does not affect progression for women with shorter participation in microfinance services while, for those with longer progression, experience of physical abuse may be associated with progression. Future research should further investigate these relationships.

This study is important because it suggests that microfinance programs may produce different effects on relationship dynamics and violence depending on the age of the client. Our findings indicate that longer microfinance membership may have a protective effect among older women, but these protective associations were not observed in younger women. Thus, older Haitian women may be benefiting from microfinance in a way that younger women are not. Further understanding the effects of economic empowerment programs, including how participation in them interacts with other individual-level and contextual factors, will allow for the better design, development, and implementation of said programs. Future studies should examine the moderating effects of age on the relationship between microfinance membership and relationship dynamics and violence in larger samples. Additionally, future research should assess whether additional training and resources could improve outcomes in younger women as well. Gender transformative interventions, interventions aimed at reshaping community gender norms and beliefs, or increasing involvement of the male partners in microfinance activities may be effective strategies.

AUTHOR BIOGRAPHIES

Maya Luetke, MSPH, is doctoral student in the Department of Epidemiology and Biostatistics in the Indiana University School of Public Health Bloomington. Her research focuses on the economic and social dynamics of relationships, gender-based violence, and sexual health. She is interested in leveraging research to reduce the negative health impacts of poverty and to minimize interpersonal violence.

Reginal Jules, is the Social Impact Monitoring and Market Research Director at Fondation Kole Zepòl (Fonkoze Foundation), Haiti’s largest microfinance institution. He has worked at Fonkoze for more than 6 years and has extensive experience in data collection.

Sina Kianersi, DVM, is doctoral student in the Department of Epidemiology and Biostatistics in the Indiana University School of Public Health Bloomington. In his research, he aims to use epidemiological and network science methods to study the modifiable risk factors of sexually transmitted infections (STIs). He is interested in exploring the causal pathways that result in STI epidemics.

Florence Jean Louis, MD, is a physician and the Director of the Human Development Program at Fondation Kole Zepòl (Fonkoze Foundation), Haiti’s largest microfinance institution. She has wide-ranging public health knowledge and has substantial experience in the design and implementation of gender trainings.

Molly Rosenberg, PhD, is an Assistant Professor in the Department of Epidemiology and Biostatistics in the Indiana University School of Public Health Bloomington. Her research is focused on identifying how social, structural, and economic factors influence sexual health outcomes. Her work has largely focused on adolescent sexual health in South Africa and other low- and middle-income countries.