Can household loans boost women’s agency without targeting women?

This is our 6th blog post for our Job Market Paper Series blog for 2025-2026.

Tarana Chauhan is a Postdoctoral Research Associate at Brown University. She is an applied microeconomist studying topics at the intersection of Development and Labor Economics, with a focus on Gender, and teaches an upper-level economics elective course on Gender and Development to undergraduate students at Brown. She is on the academic job market 2025-26. 

An important question for effective delivery of social protection and financial capital is whether we should target resources to the household or individual. Programs transferring resources to women may improve household profits but don’t usually succeed in improving women’s income relative to household members or their say in household decisions (Banerjee et al. 2015; Green et al. 2015; Attanasio et al. 2015; Angelucci et al. 2015; Fiala 2018; Said et al. 2019; Bernhardt et al. 2019). Providing bank accounts to women has not consistently increased their role in household decisions (Field et al. 2021), although more households saved and borrowed from formal institutions (Chauhan 2025). While cash transfer programs have improved women’s decision-making in some contexts (deBrauw et al. 2014, Bergolo and Galván 2018), some studies document increases in threats and aggressive behaviors (Angelucci, 2008; Bobonis et al. 2013). 

The evidence of pecuniary and social costs of targeting raises a practical question: can easing household liquidity constraints – without explicitly targeting women – indirectly strengthen women’s control over resources and financial inclusion?

Conceptual frame and outcome variables

Adapting Bobonis (2009), we view the household as a collective decision-making unit. A short-term loan increases available resources (“non-labor income”), relaxing liquidity constraints for private and composite goods (consumed both privately and jointly) such as productive capital and consumer durables. This can raise women’s utility via greater consumption and shift labor supply through an income effect.

Naila Kabeer (1999)’s definition of agency as “the ability to define one’s goals and act upon them” has been widely used in the literature studying women. In our expanded definition, we measure women’s agency across three domains that are closely related with the loan program:

1. Financial inclusion: account use, debit-card use, and mobile banking knowledge/use.
2. Time allocation: hours in farm work, non-farm work, domestic tasks, and personal/leisure.
3. Decision-making: involvement in production and consumption choices (adapted from pro-WEAI).

Intervention and research design

The paper uses data from a randomized loan program^[1] implemented in 80 villages of India (40 treatment, 40 control). These villages were split equally between the rice growing state of Odisha and cotton-cultivating Maharashtra. In treatment villages, small and marginal farmers were offered collateral-free loans of USD 300–600 before planting, at interest rates 10 pp below market alternatives, with timely disbursal. Eligibility required minimum residency and cultivation experience, landholding ≤ 3 acres, and a credit score that estimated plot viability using remote sensing tools. The loan was serviced through Dvara E-Registry (DER) who collaborated with local commercial banks.

Before randomization, DER held village information sessions on loan requirements and compiled a roster of farmers who attended. Loan offers followed these sessions. Baseline and endline surveys allow constructing a balanced panel of 1,329 households where baseline demographics and economic characteristics were statistically similar between treatment and control villages. Attrition was 15% and balanced across arms. We estimated the intent-to-treat (ITT) effects of the loan program on household investments and income using an Analysis of Covariance (ANCOVA) model. The endline survey interviewed both the respondent who attended the information session and another household member. This yields a sample of 1,028 women. We examined estimated Local Average Treatment Effects (LATE) of the loan on women’s outcomes using village-level assignment as an instrument for receiving a loan. We report sharpened q-values of the treatment coefficient that adjust biases from multiple hypotheses testing in the paper.

What changed at the household level?

Households used the short-term liquidity to smooth consumption and purchase durables rather than expanding farm production. They were 10 percentage points (pp) and 9 pp more likely to own non-mechanized farm equipment (e.g., hand tools, animal-drawn ploughs) and large consumer durables (e.g., refrigerator, washing machine, furniture), respectively. Applying World Food Program’s measure, we find that household food security score increased by 21% (control mean 44.7). The loan did not significantly expand cultivated area, agricultural input spending, or farm income over the 2023 summer planting season. A potential source of increased household welfare is savings from opting for the cheaper loan offered under this program.

Did women’s financial inclusion improve?

We find suggestive evidence of resource sharing and greater financial access in Figure 1: significantly more women in loan receiving households reported depositing cash in their bank account every week (these are not earnings or government transfers). Conditional on having an individual bank account, women in loan receiving households were more likely to know and use mobile banking services (over 100 pp). These linear estimates are outside [0, 1] as the dependent variable is binary. Gains in women’s uptake of digital financial services are robust to non-linear estimation using bivariate probit models although effect sizes are smaller than the LATE estimates. The significantly greater uptake of digital financial services in the treatment arm is driven by smartphone ownership (30 pp) and household’s preferences for financial autonomy.  Both men and women in households that received the loan product were more likely to use debit cards independently than in the control group (33 and 87 pp, respectively).

About 70% of households owned at least one smartphone at baseline with no statistical differences between the intervention arms. Since real household income did not change, the loan is a likely source of financing for women’s own smartphones consistent with the view that liquidity helps overcome lumpy purchase constraints. These findings contrast with the evidence from smartphone distribution programs. In a state-run program (Chhattisgarh), women’s mobile-banking knowledge and use did not increase, and two in five women transferred the phone to another household member(Barboni et al. 2024). In our setting, untargeted household liquidity coincided with women’s autonomous use of digital finance, highlighting that direct resource transfers do not guarantee impact.  

How did the loan affect women’s time use?

On the busiest agricultural day, Figure 2 shows that women in treatment households spent less hours on farm work (1.6 vs 3.1 hours) and more time on personal activities (2.5 vs 1.6 hours). Agricultural income or expenditure on hired labor did not change in response to the loan. These patterns, reflecting household preferences for leisure, are consistent with an income effect: liquidity plus investment in productive capital reduced female labor requirements, freeing time for rest or other activities. Cultural acceptability may also play a role: cash transfers in rural Egypt reduced women’s non-farm work but did not shift farm labor (El-Enbaby et al. 2025).

What about decision-making power?

We find no overall difference in women’s or men’s involvement in decisions across production and non-production domains in the last 12 months. This aligns with mixed results in the literature: many cash transfer programs raise decision-making (deBrauw et al. 2014, Bergolo and Galván 2018), while social protection and financial inclusion interventions did not find similar improvements (Roy et al. 2014, Field et al. 2021, Chauhan 2025). However, heterogeneity matters: using the baseline share of joint decisions (out of 10 categories) as an ex-ante measure of bargaining power, women in the treatment arm and households with higher spousal coordination were more likely to be involved in livestock/poultry management and non-farm business decisions.

Two broader implications follow:

1. Liquidity can be gender-salient without gender targeting. When capital and durables are lumpy, easing liquidity constraints may facilitate purchases that enable women to transact independently (e.g. phones), increasing the intensity/ frequency of decisions made even when the prevalence of decision making did not change.
2. Expect short-run consumption and asset smoothing, not immediate production expansion. Over one season, households behaved like recipients of a transfer raising food security and durables. 

Takeaways

• Collateral-free loans to farming households did not change agricultural yield or income in the short run, but increased productive capital, consumer durables, and food security.
• Women in loan receiving households reported greater financial inclusion with autonomous debit-card use and mobile-banking knowledge/use, likely facilitated by greater smartphone ownership.
• Women also spent less time on farm labor and more on personal activities, consistent with income effects and capital substitution.
• Decision-making did not change overall, but women in households with higher baseline involvement were more likely to participate in livestock and non-farm business decisions.

Link to paper: https://github.com/TaranaChauhan/Agricultural-Credit-and-Women-s-Agency

Featured image: Dvara E-Registry (URL: https://www.youtube.com/watch?v=P7nu86jM_7U)

---

^[1] Kramer, Berber and Patrick Ward. 2023. “Agricultural credit, insurance, and over-indebtedness among smallholder farmers.” AEA RCT Registry. June 28. https://doi.org/10.1257/rct.11616-1.0. The RCT was funded by CEGA’s Digital Credit Observatory, a Gates Foundation-supported initiative.