The impact of husbands' involvement in goal‐ setting training on women's empowerment: First evidence from an intervention among female microfinance borrowers in Sri Lanka

Offering women access to microcredit and business training is a prominent approach to stimulate women's empowerment. Whereas men seem to profit from business training, women do not. We adjusted a goal-setting training session on the basis of women's needs in collaboration with a women organization in Sri Lanka. We invited female microfinance borrowers and their husbands to the training as both parties should be involved to change existing gender roles with respect to their income-generating activity. We investigated the impact of the training on goal-setting skills, self-esteem, and the couples' interaction in a subsequent task. In two field experiments, female borrowers and their husbands (nstudy1 = 68; nstudy2 = 76) were randomly assigned to one of three conditions: (a) goal-setting training and setting goals as couple, (b) goal-setting training and setting goals individually, or (c) no training (control condition). Participation in the training increased women's SMART (specific, measurable, achievable, realistic, time bound) goal-setting skills. We coded couples' interactions in a subsequent decision-making task to assess signs of women's empowerment. Descriptively, we found some initial evidence of increased women's empowerment in the interaction (Study 2). We critically discuss results and how gendered power imbalances may need to be addressed to stimulate social change towards gender equity.

initial evidence of increased women's empowerment in the interaction (Study 2). We critically discuss results and how gendered power imbalances may need to be addressed to stimulate social change towards gender equity. KEYWORDS empowerment, goal setting, partner interaction, training, women 1 | INTRODUCTION Today, over 120 million people receive small loans to set up and develop their income-generating activities; 84% are women (Microfinance Barometer, 2017). Besides providing small loans, entrepreneurs can participate in different trainings to mainly improve business skills to help them move out of poverty (Armendáriz & Morduch, 2010). Previous research shows that business training improved business outcomes among male entrepreneurs but not among female entrepreneurs in Tanzania (Berge, Bjorvatn, & Tungodden, 2015). The researchers suggest that women only benefit from training if their special role as female entrepreneurs is also addressed in the training. A large field experiment showed that a combined gender and business training for female entrepreneurs increased business outcomes (Bulte, Lensink, & Vu, 2017) and first signs of women's empowerment such as increased control beliefs and intrahousehold decision-making power in Vietnam (Huis, Lensink, Vu, & Hansen, 2018). Power barriers and prescribed gender roles may strongly inhibit women's abilities in general (e.g., Connell, 1987;Pratto & Walker, 2004).
Inviting women together with their husbands to training may be a first step to stimulate the discussion of business responsibilities and may start changing prescribed traditional gender roles with respect to the income-generating activity (for a similar argument, see Mauchi, Mutengezanwa, & Damiyano, 2014).
In the current research, we describe the adjustment of one training session (selected from the Gender and Entrepreneurship Together Ahead programme, called GET-Ahead, which has been implemented in 17 countries; see Bauer et al., 2004) to the Sri Lankan context, which addressed the needs of female entrepreneurs. We investigate the impact of one short goal-setting session and husbands' involvement in this training on goal-setting ability, women's personal empowerment, and signs of empowerment in videotaped spousal interactions in a subsequent task using self-reports and observational data.

| Culture and gender power relations in Sri Lanka
Sri Lanka is a South Asian nation. Since the end of an armed civil conflict in 2009, Sri Lanka's economy has been developing, and it is now a lower middle-income nation. People in this collectivistic society adhere to a cultural ecology of embedded interdependence (Dissanayake & Semasinghe, 2015). People in such a society are strongly embedded in social networks such as families and communities. Their relationships are characterized by obligations for care and support both within and outside the nuclear household with clear gender roles for women and men (e.g., Kurtiş, Adams, & Estrada-Villalta, 2016). Power barriers and prescribed gender roles often inhibit women's abilities to develop stronger feelings of empowerment (e.g., Connell, 1987;Pratto & Walker, 2004). More precisely, gender relations are structurally determined but are also shaped by individual practices of these relations (Connell, 1987), and gender dynamics take specific forms in colonial and postcolonial contexts (Connell, 2014). Moreover, power is divided along gendered lines such that men commonly hold more power on four different basis: Men hold power † The first two authors contributed equally to this work. over resources (e.g., household income), men have more force (e.g., exert intimate partner violence), women have more obligations (e.g., household responsibilities), and ideologies may enforce this power imbalance (e.g., gender roles: see gendered power model; Pratto & Walker, 2004). In the current context, these power barriers can oftentimes inhibit women. In Sri Lanka, it is important for women to be mothers and wives, and it is expected that they obey their husbands (de Alwis, 2002). In the last decades, women's participation in the paid labour market has increased; many of them work in the textile and garment sectors for the export. Although many workers in this specific sector reported feeling empowered as a result of their work, at a societal level, negative stereotypes shared about women working in this sector corroborated the individual feeling of empowerment (Hancock & Georgiou, 2017). This is one example to illustrate how societal attitudes may still hinder women's empowerment.

| Women's empowerment
Women's empowerment is a multifaceted process involving individual and collective awareness, beliefs, and behaviour embedded in the social structure of a cultural context (Huis, Hansen, Otten, & Lensink, 2017). Empowerment is seen as the process through which women acquire and use resources agentically to reach achievements that were previously denied to them (Kabeer, 1999). In the field of microfinance services, women's empowerment has been assessed with diverse indicators ranging from self-esteem to financial household decision making, and women's social network size (for an overview, see Huis et al., 2017). In Sri Lanka, the social world is perceived as interdependent, and women are not individual agents but embedded in their social relations (Kurtiş et al., 2016).
In this research, we therefore invited female microfinance borrowers to join one training session modified for the cultural context with their husbands. We examined training effects on signs of women's personal empowerment (self-esteem) and relational empowerment (couple interaction in a subsequent decision-making task). We videotaped the decision-making task to examine how women may influence decisions, aiming to capture signs of relational empowerment during actual decision making (e.g., Farrell, Simpson, & Rothman, 2015).

| Microfinance services and women's empowerment
Microcredits were first offered in the 1980s to help people set up small income-generating activities and thereby reduce poverty and empower the disadvantaged. Mohammed Yunus first introduced this concept in Bangladesh.
Small loans are provided to people who often do not qualify to receive loans from commercial banks. Today, people often receive loans together with additional services such as business training (Armendáriz & Morduch, 2010). The provision of microfinance services is one prominent approach to strengthen the position of women. Women's economic participation is expected to stimulate their empowerment (e.g., Armendáriz & Morduch, 2010). However, this approach is also criticized. First, the prevalent focus on women's individual growth may be less compatible in cultural ecologies of interdependence where interpersonal networks are powerful (e.g., Kurtiş et al., 2016). Furthermore, female microfinance borrowers may have to repay loans that they took out for male relatives. As they have little control over these loans, they often end up indebted and unempowered (Mayoux, 2001).

| Training in the context of microfinance services
The majority of training programmes in the context of microfinance services focus on business skills such as bookkeeping (e.g., Armendáriz & Morduch, 2010). Research conducted in Tanzania showed that business training improved business outcomes for male but not for female entrepreneurs. The researchers suggest that women's special role as female entrepreneurs should be addressed (Berge et al., 2015). Research conducted in Sri Lanka also suggests that programmes need to address both business skills and women's position in society to stimulate women's business success (Herath, Guneratne, & Sanderatne, 2016).
One training that addresses the specific role of female entrepreneurs is the GET-Ahead training. It was developed by the International Labour Organization and has been implemented in over 17 countries. This gender and business training aims to stimulate women's entrepreneurial skills including basic business and human management skills by focussing on the specific role of women (i.e., gender roles: nine training sessions, 32 exercises; Bauer et al., 2004).
A recent longitudinal field experiment showed that participation in the GET-Ahead training increased women's personal (e.g., personal control beliefs) and relational (e.g., financial household decision making) empowerment among female loan borrowers in Vietnam than in a control group of borrowers who did not receive training (Huis et al., 2018). This evaluation provided first evidence that especially gender knowledge and not only business knowledge increased women's empowerment.
In the current research, we selected one training session from the GET-Ahead training based on a literature review, advice from our local collaboration partners, and qualitative data (focus group discussions with female borrowers). Previous research shows that successful entrepreneurship requires entrepreneurs to be proactive and self-starting (for meta-analyses, see Frese & Gielnik, 2014). The same authors have developed two types of business training in close collaboration with local partners in Africa showing that training focussing on personal initiative stimulated business success among male and female entrepreneurs (e.g., Campos et al., 2017). Personal initiative involves long-term-oriented self-starting behaviour and persistence when problems and opportunities appear. Both personal initiative and an agentic approach in setting goals, forming and executing action plans, and monitoring progress are needed for successful entrepreneurship (e.g., Frese & Gielnik, 2014). Thus, training on setting business goals is a first important step towards business success. Additionally, we conducted short focus group discussions with female entrepreneurs, which confirmed the relevance of learning to set business goals (see Section 2.2). Therefore, we adapted one exercise of the GET-Ahead training programme on goal setting for entrepreneurs' income-generating activity. We expected that participating in a short goal-setting training would increase women's goal-setting skills (Hypothesis 1).

| Involvement of female borrowers' husbands
Many microfinance institutions offer their services to women only. Their assumption is that women are more trustworthy clients; that women invest more in household health, education, and nutrition; and that this should lead to gender equality (Armendáriz & Morduch, 2010).
However, some men felt excluded from microfinance programmes because most borrowers are women, and some women even faced intimate partner violence (e.g., Rahman, 1999). Some men may exert domestic violence or controlling behaviour over their wives (e.g., the loan or their behaviour) to regain greater control that stems from the breaking of traditional gender roles (for a discussion, see Dutt, Grabe, & Castro, 2016). To overcome friction among couples, it may be helpful to involve husbands of female borrowers (Rahman, Hoque, & Makinoda, 2011). This way, women and men may be able to discuss responsibilities, which may change existing gender roles (e.g., Howson & Flood, 2015). Importantly, to achieve social change, members of both advantaged and disadvantaged groups, such as men and women, need to be involved (e.g., Dixon, Levine, Reicher, & Durrheim, 2012). Thus, women and men need to renegotiate their responsibilities and roles before existing patterns may change.
Offering joint training for couples may stimulate changes in gender roles and strengthen women's empowerment with respect to the income-generating activity (e.g., Mbweza, Norr, & McElmurry, 2008). We expected that participating in goal-setting training would result in enhanced personal empowerment for female borrowers receiving goal-setting training and setting goals independently (independent goal-setting condition) or collaboratively with their husband (collaborative goal-setting condition) than for female borrowers who only received training after having completed all empowerment measures (control condition; Hypothesis 2).
Next, we examined women's role in spousal decision making in a subsequent decision-making task to examine relational empowerment. 1 Spouses who reported having joint goals showed more collaboration in problem solving (Hoppmann & Gerstorf, 2013). Thus, we expected more signs of relational empowerment in couples' interaction for female participants in the collaborative goal-setting condition, compared with female participants in the independent goal-setting condition, and even more so compared with female participants in the control condition (Hypothesis 3).

| The current research
We conducted our field work in collaboration with the largest grassroots women organization in Sri Lanka, which aims to reduce poverty through the empowerment of women. Women's Development Federation (WDF) is situated in the south of Sri Lanka (Hambantota District) and was founded in 1989 by women. Women organized themselves in groups and initiated different activities to improve the well-being of their families. Today, WDF provides microfinance services and addresses health and nutrition problems to fight poverty to over 70,000 women (WDF, 2015).
We used a mixed-methods approach employing qualitative and quantitative research methods and examined women's empowerment by using self-reports and coded couple interactions. First, on the basis of a preliminary study, we selected one training session on goal setting from the GET-Ahead training (Bauer et al., 2004) and adapted it to the current context by discussing with our research partners how to adjust the training materials to the local context.
Next, in two field experiments, we examined the impact of a goal-setting training where participants learned how to set goals for their income-generating activity, such as how to increase harvest or sell more products (e.g., candy, papayas, and dresses) on goal-setting ability, women's personal empowerment, and relational empowerment in couples' interaction in a subsequent decision-making task.

| Research relationships
Data for the present research were collected in 2015 in the Hambantota District. The research design developed out of a partnership between the second author and two collaborating organizations, Strømme Micro Credit Sri Lanka, a supporting organization that was founded by Strømme Foundation (a leading Development Organization in Norway with the vision of a world free of poverty), and WDF. The partnership was based on a critical and egalitarian dialogue between researchers and community partners to conduct meaningful research. Researchers were responsible for the theoretically appropriate and scientifically sound methodology, and the community partners were crucial in identifying the needs of their community and helping to develop culturally appropriate study materials. Prior to the data collection and during the field work, regular meetings with all involved parties were held to ensure the collaborative efforts.
Ethical approval for all studies was given by the Ethical Committee Psychology of the University of Groningen, The Netherlands. All women and men participated voluntarily; received informed consent and could withdraw at any time of the study; received participation certificates and refreshments as compensation; and were carefully debriefed. We asked all participants whether we could videotape their interaction in a subsequent task (Studies 1 and 2). Couples' decision to participate or to not participate did not impact current or future loan applications. For all studies, we randomly selected participants from WDF's membership lists. When participants were not able to attend, WDF's field officers approached additional couples to join by inviting couples from their own or participants' personal networks. All participants had received a loan. 1 We focussed on the decision-making process. The task did not allow us to evaluate the quality of the outcome.

| Preliminary study
We conducted four focus group discussions with female microfinance borrowers from WDF (N = 21) to understand which type of training would help them to improve their income-generating activity. We invited women from different saving groups-in which women share a group fund and support each other's repayments-to encourage free expressions of opinion.
The discussions were led by a trained native female research assistant and lasted 75 to 95 minutes (M = 85.03, SD = 8.25). First, we asked women which type of training they would like to attend to learn to improve their income-generating activity. We suggested potential training topics on the basis of interviews with local microfinance experts and asked all women to raise their hands if they would like to join this training; 48% of the women wanted to learn how to set goals and plan to manage income generation, 32% wanted vocational training (e.g., making snacks and dresses), and 20% wanted business training (e.g., money management and approaching the market). Second, we asked women whether they would prefer to participate in training alone or together with their spouse; 93.3% 85% of the couples worked together in their business (e.g., agriculture and selling product).

| Procedure
We offered one training session in which participants learned to set goals for their income-generating activity (i.e., dressmaking, grocery shop) according to SMART characteristics (specific, measurable, attainable, realistic, and time bound) in a 40-min training session (see Figure 1 for the training content). The same native research assistant conducted 29 training sessions. For each session, we invited three couples (M #couples = 2.68, SD = 0.74, range 1-4).
The exact procedure differed per condition (see Table 1 for the procedure per condition). Most importantly, only couples in the collaborative goal-setting condition were encouraged to collaborate with their partners to discuss, select, and write down the two most important goals for their income-generating activity.
After training, all couples engaged in a decision-making task where planning was central. Couples received 54 colour-coded building blocks, equally divided between the wife and husband, to build their dream house. We videotaped their interaction to code how both partners interact with each other while completing the task to assess signs of empowerment. Next, the research assistant assessed participants' personal empowerment through self-2 Women in the collaborative goal-setting condition were less educated than women in both other conditions. Women in the independent goal-setting condition were married longer than were women in both other conditions. Controlling for these differences did not change the results. Thus, we report results without covariates.
report. Couples in the control condition completed the decision-making task directly after a demographic questionnaire and only participated in the training after all dependent measures were assessed.

| Measures
We carefully developed and pretested all dependent measures with native speakers to fit the cultural context (Hopkins, 2015).

Goal setting
All participants were asked twice to write down two goals for their income-generating activity: before and after training. All goal statements were translated by a native speaker and coded by two independent coders on whether the goal was specific, broken down into different steps, concrete, measurable, and linked to a time frame (see Table 2 for the coding system). We calculated the pretraining score for goals before training and the posttraining score for goals after training (ranging from 0 to 7 points). Both coders independently coded all pretraining and posttraining goals achieving excellent interrater reliability (K = 0.88). All remaining discrepancies were resolved through discussion, and full consensus was reached.

Self-esteem
We assessed state self-esteem as one indicator of women's personal empowerment. Five items were adapted to the specific training context (Heatherton & Polivy, 1991; see Table 3 for items). Participants indicated to what extent they believed that their performance in the training was good on a 5-point scale ranging from not at all (1) to extremely (5). The scale showed a good reliability (α = 0.74). 3

Couple interaction
First, together with two Sri Lankan research assistants, the first author developed a coding scheme. Next, two independent coders at the University of Groningen coded the videotaped interactions in the decision-making task.
The first 5 minutes of each video was coded. Previous research shows that brief segments of nonverbal behaviour represent behaviour across longer interactions well (Murphy, 2005). The coders reached acceptable interrater reliability (K = 0.47, p < 0.001, r = 1.00). All remaining discrepancies were resolved through discussion, and consensus was reached. We included three indicators of women's empowerment in the couple interaction. The first is personal initiative by the relative number of building blocks added by female participants. The second is speaking occasions by the number of times women spoke. We coded each verbal contribution when women started talking, including mumbling after 5-s silence or an interruption by the partner. The third is speaking duration by the total time during which women spoke.

Training evaluation
We asked participants whether the training was not at all, a little bit, or very much useful and whether they learned nothing new, some new things, or a lot of new things.  Tables 4 and 5 for an overview of all study variables).

| Training evaluation
Of the female participants, 95.7% said the training was very useful and 63.8% indicated that they learned a lot of new things.

| Discussion
The majority of the couples evaluated the training positively. As expected, training significantly improved participants' goal-setting skills across conditions. However, we did not find any significant impact of condition on empowerment indicators.

| STUDY 2
Study 2 had three alterations. First, we intensified the collaboration in the collaborative goal-setting condition to overcome too similar instructions in both training conditions. Second, we selected a more realistic and relevant decision-making task: investing money. Third, we aimed to increase the sample size.

| Sample and design
Seventy-four couples were randomly assigned to one of three conditions: collaborative goal-setting condition (n = 25), independent goal-setting condition (n = 23), or control condition (n = 26). All women were married (M years = 14.12, SD = 7.09; range 1-29) and were between 25 and 45 years old (M = 36.27, SD = 7.66; range 19-51 years); 76.4% of the couples worked together as a couple in their business (e.g., agriculture, selling products).

| Procedure
The same native research assistant conducted 25 training sessions. For each session, we invited three couples to participate (M #couples = 3.50, SD = 1.00, range 1-5). There were two procedural alterations to Study 1. First, we intensified collaboration in the collaborative goal-setting condition. Couples collaborated both on adapting their individually set goals to the training and on selecting the best two goals out of these adapted goals (see Figure 1 for overview of the training and Table 1 for the procedure). Second, we used a different decision-making task. We asked couples to engage in financial decision making together by investing an imaginary 200,000 RS (±1,200 EUR) in their income-generating activity. The task consisted of two phases. First, we encouraged the couples to generate as many ideas as possible on how to spend the 200,000 RS. The research assistant wrote each suggestion on a Post-it. Next, the couples were asked to invest the money by dividing 40 copies of 5,000 RS notes over the Post-its. We videotaped this interaction.

| Goal setting
Again, the participants wrote down two goals for their income-generating activity before and after training. We used the same coding method (see Table 2; range 0-7), showing excellent interrater reliability (K = 0.86). All remaining discrepancies were resolved through discussion, and full consensus was reached.

| Self-esteem
Five items assessed women's state self-esteem in the current context (Heatherton & Polivy, 1991; see Table 3 for items). Participants indicated to what extent they believed that they could plan their income-generating activity on a 5-point scale ranging from not at all (1) to extremely (5). We could not construct a reliable scale from the five items (α = 0.14). On the basis of factor and reliability analyses, we constructed a three-item scale (α = 0.55, Table 3). 4

| Couple interaction
We used an almost identical coding scheme to Study 1. Two independent coders coded the first 4 min of each phase of the videotaped interactions in the decision-making task, thus a maximum of 8 min. They reached acceptable interrater reliability (K = 0.39, p < 0.001, r = 0.92) in a first coding round. All remaining discrepancies were resolved through discussion, and full consensus was reached.
We included four indicators of women's empowerment in the couple interaction: first, personal initiative by the percentage of the total 200,000 RS allocated by women; second, first say by assessing who was the first to act or speak; third, speaking occasions by the number of times women spoke; and fourth, speaking duration by the total time during which women spoke.

| Task duration
We calculated the length of the decision-making task.

| Training evaluation
Again, we asked participants whether the training was not at all, a little bit, or very much useful and whether they learned nothing new, some new things, or a lot of new things. 4 Due to low scale reliabilities for several personal empowerment measures (α = 0.12 to α = 0.43), we decided to focus on the observational indicators of empowerment in the videotaped couple interaction.

| Couple interaction
There was no statistically significant multivariate effect in women's empowerment in couple interactions between the

| Total duration of the task
The duration of the decision-making task did not differ between conditions, F (2, 64) = 2.97, p = 0.058, η p 2 = 0.09.

| Training evaluation
Of the female participants, 92.9% said the training was very useful for them and 52.2% indicated that they learned a lot of new things.

| GENERAL DISCUSSION
This research describes a first attempt to systematically investigate how goal-setting training for female borrowers and their husbands (based on women's needs) may stimulate women's empowerment. We describe the adjustment and implementation of a goal-setting training in Sri Lanka. Next, we discuss the impact of this training on goal-setting skills, women's self-esteem, and empowerment in couples' interaction. In two studies, confirming Hypothesis 1, women developed SMARTer goals for their income-generating activity across conditions. We observed no significant impacts of condition on women's self-esteem (Hypothesis 2) nor on signs of empowerment in couples' interactions (Hypothesis 3). However, on a descriptive level, we see first possible signs of change in Study 2, albeit marginally significant at most. Interestingly, couples in both training conditions engaged in longer decision making than did couples in the control condition who had not participated in training yet. Moreover, women in the independent goal-setting condition showed more personal initiative and more often had the first say in decision making than did women in both other conditions. As these findings are not conventionally significant, they should be interpreted with caution, although they may signal first change.

| Gendered power
Previous research has stressed the need to address the structural base of gender inequities and to take cultural norms into account (e.g., Connell, 1987;Dutt et al., 2016). In the current research, the training aimed to stimulate joint goal setting among couples with the aim to change gender roles with respect to their income-generating activity. The underlying gendered power imbalances were not addressed. To stimulate gender equity, attention should be given to the root sources of the gendered power imbalances and manifestations of gender roles in local contexts (e.g., Dutt et al., 2016;Pratto & Walker, 2004). To the best of our knowledge, only one longitudinal field experiment investigated the impact of a gender and business training (focussing on the specific role of female entrepreneurs [gendered power imbalances, gender roles] and business skills) and the impact of inviting husbands to join this training on female borrowers' empowerment (Huis et al., 2018). Results provided evidence for increased women's empowerment.
Importantly, research suggests that men's involvement should not come at the cost of women's autonomy to achieve egalitarian decision making (Allen, Armendáriz, Karlan, & Mullainathan, 2010). Engaging men only to some parts of a training programme while also keeping an exclusively women's space, in which women can discuss daily challenges and problems, might be a promising way for social change.

| The promising impact of training
In this article, we focussed on the impact of training and not microcredit. We show that offering training-even a single 45-min training session-to female microfinance borrowers and their husbands improved goal-setting skills.
However, we found no significant impact on women's empowerment. Other field research also showed that more frequent training participation was related to stronger feelings of empowerment (i.e., locus of control) among conditional cash transfer participants in Panama (Alvarez, van Leeuwen, Montenegro-Montenegro, & van Vugt, 2018) and microfinance borrowers in northern Sri Lanka (Hansen, 2015).
We speculate that there might be five reasons why training did not strengthen women's empowerment in our research. First, the training did not focus on the discussion of gendered power imbalances and existing gender roles.
Second, we only focussed on a single training session. It may take longer before interventions can increase women's feelings of empowerment. Third, our participants had been married for approximately 15 years. They may have established communication and decision-making rituals in their relationship (Pearson, Child, & Carmon, 2010), which are not influenced by one training session. Fourth, the presence of husbands may have influenced women to confirm to existing gender norms of being agreeable and supportive (i.e., stereotype threat; e.g., Logel et al., 2009), thereby hindering the expression of signs of empowerment. Fifth, the sample size in our studies was likely too small to observe subtle behavioural changes signalling empowerment in couples' interaction.

| Studying signs of women's empowerment in a collectivistic society
Relationships are an important aspect of daily life and perhaps even more central in societies adhering to a cultural ecology of interdependence (e.g., Adams, Anderson, & Adonu, 2004), such as Sri Lanka. We assessed empowerment in couples' interaction by coding couples' nonverbal behaviour to gain insight into couples' decision-making dynamics.
Observing nonverbal behaviour is a promising way to assess empowerment (Malik & Lindahl, 2000). We developed the coding system with native speakers to take the cultural specificity of women's empowerment into account (see Hopkins, 2015). Although we only found first, nonsignificant, signs of women's relational empowerment, we nonetheless hope to inspire future research to consider using observational data.

| Limitations and future research
There are three important limitations. First, the training focussed on teaching female entrepreneurs and husbands to set goals for their income-generating activity as a first step to renegotiate gender roles with respect to the income-generating activity. The training did not focus on gendered power imbalances directly, which might be a prerequisite to stimulate sustainable change. Second, due to reality constraints in the field work (e.g., number of clients per women organization office), we only managed to include a small sample and thus do not have enough statistical power to observe subtle changes in couples' interaction patterns. Furthermore, in addition to randomly selected participants, WDF's field officers invited additional couples to participate. Thus, self-selection may be a problem and should be considered in interpreting our findings.
Third, we investigated a single 45-min goal-setting session from the GET-Ahead training programme, which in total lasts approximately 36 hr (Bauer et al., 2004). Other research observed increased empowerment effects after an adapted training lasting approximately 9 hr (Huis et al., 2018). Future research should examine the impact of a longer (multiple session) training programme focussing on gendered power imbalances and business skills and should include a larger sample to observe subtle changes in couples' interaction.

| Microfinance approach to women's empowerment
The underlying assumption of microfinance service is that women's engagement in the financial market should stimulate women's personal growth (e.g., Armendáriz & Morduch, 2010). However, individual agency may not be the only pathway towards women's empowerment (see Estrada-Villalta & Adams, 2018). Previous research examining empowerment through microcredit in Tanzania suggests that cooperative business ownership more than independent business ownership strengthens women's empowerment (Dutt et al., 2016). Thus, in interdependent cultures, interventions focussed on women's cooperation rather than individual growth may offer more promising pathways to empowerment.

| CONCLUSION
In this research, using self-reports and observational data, we describe and examine the empowerment impact of a goal-setting training for female microfinance borrowers and their husbands in the field. We provide first evidence that a short goal-setting training improves participants' goal-setting skills and may strengthen women's empowerment. We critically discuss that empowering women requires a focus on existing gendered power imbalances as well as gender roles. We hope that this research inspires new training initiatives and innovative monitoring and evaluation designs. In sum, we propose that research conducted in real-world settings can inspire theoretical and practical discussions to increase gender equality through training.
to thank all field officers, female microloan holders, and their husbands who voluntarily participated in this research.
The first author was supported by a travel scholarship by the University Meets Microfinance Fund.