Study Overview This sub-study is nested within the ECOVI trial (Disentangling and Preventing Economic Violence against Women), a two-arm cluster RCT evaluating a couples-based financial literacy and gender-transformative intervention ("Let Us Grow Together: Economic Wellbeing for Families") across 150 clusters in Maharashtra, Andhra Pradesh, and Rajasthan, India. The sub-study examines parenting practices as an understudied downstream outcome. Parenting questions (s-EMBU-P) are added to the ECOVI endline survey prior to data collection, providing a fully prospective analysis plan. All mediator measures, including IPV types and mental health (GHQ-6), are collected at baseline and endline as part of the main trial. The sub-study pursues four objectives: (1) to estimate whether the intervention effect on parenting operates through relational quality (Relationship channel); (2) through a reduction in IPV (IPV Reduction channel); (3) through improvement in parents' mental health (Mental Health channel); and (4) through financial distress reduction (Financial Distress channel).
Background and Motivation Parenting practices fall into two broad dimensions: negative practices (corporal punishment, inconsistent discipline, poor monitoring) and positive practices (warmth, responsiveness, emotional involvement, active engagement in learning). Negative parenting, particularly harsh physical discipline, is associated with elevated child anxiety, aggression, and long-term deficits in cognitive and socio-emotional functioning. Positive parenting serves as a protective factor, buffering children from the adverse effects of household poverty, conflict, and other adversities.
IPV is well-documented as an important driver of parenting quality. Exposure to IPV is associated with declines in nurturance and communication between parents and children and with increases in aggression toward children. Effects of IPV on harsh parenting are partly mediated by parental depression, establishing mental health as a key mechanism linking couple-level violence to parent-child interactions. Notably, most existing evidence focuses on mothers; relatively little is known about how paternal involvement in IPV affects men's own parenting.
Mental health is a central mediating pathway: depression, anxiety, low self-esteem, and parenting stress are the psychosocial risks most consistently linked to child maltreatment. Experimental evidence indicates that interventions reducing maternal depression can produce lasting improvements in parenting investments and reductions in child physical abuse, and that mental health fully mediates associations between economic empowerment and child maltreatment.
Gender-transformative couples' programmes have demonstrated reductions in both IPV and violent discipline of children. Key programme features include engaging fathers alongside mothers, addressing marital conflict, and transforming attitudes condoning gender inequality. Evidence from multi-site RCTs demonstrates that such programmes even when parenting is not the primary target can produce lasting downstream improvements in parenting, suggesting that skills in equitable communication and joint decision-making transfer to parent-child interactions.
Conceptual Framework and Proposed Pathways
Four pathways through which the intervention may affect parenting practices are proposed:
Relationship channel: Improved communication, gender norms, and relationship quality may translate directly into better parenting behaviours, independent of changes in IPV or mental health.
IPV Reduction channel: Reductions in household violence and conflict reduce parenting stress and emotional exhaustion, enabling more consistent, warm, and non-violent discipline.
Mental Health channel: Improvements in mental health - driven partly by IPV reductions and improved relationship quality - can increase parenting capacity.
Financial Distress Reduction channel: Reductions in financial stress, a direct goal of the financial literacy pillar, may lead to more positive and non-violent discipline.
Research Questions and Hypotheses
Research questions addressed by the sub-study:
RQ-1: Does assignment to the intervention improve parenting practices at endline, relative to the control group? RQ-2: Is any intervention effect on parenting practices mediated by improvement in the couple relationship (Relationship channel)? RQ-3: Is any intervention effect on parenting practices mediated by a reduction in IPV between baseline and endline (IPV Reduction channel)? RQ-4: Is any intervention effect on parenting practices mediated by an improvement in psychological well-being (GHQ-6) between baseline and endline (Mental Health channel)? RQ-5: Is any intervention effect on parenting practices mediated by an improvement in financial distress between baseline and endline (Financial Distress channel)?
Pre-specified hypotheses:
H1 (Negative parenting): Compared to control group parents, intervention group parents will report lower harsh parenting and lower overprotection/control at endline.
H2 (Positive parenting): Compared to control group parents, intervention group parents will report higher positive parenting at endline.
H3 (Mediation - IPV and Financial Distress channels): A portion of the intervention's effect on parenting practices will be mediated by a reduction in IPV and financial stress.
H4 (Mediation - Mental Health and Relationship channels): A portion of the intervention's effect on parenting practices will be mediated by improvement in GHQ-6 scores and better relationship quality.
Measurement and Analytic Sample Parenting outcomes are observed at endline only, having been added to the endline survey before data collection commenced. The proposed mediators - IPV, mental health, relationship quality, and financial distress - are observed at both baseline and endline as part of the main ECOVI trial.
For the parenting analysis, the analytic sample is restricted to couples with at least one child residing in the household at the time of the endline survey. Where a household includes multiple children, respondents are asked to answer parenting questions regarding the youngest school-age child (the youngest child aged 5-18 currently enrolled in school). Parenting questions are administered separately to mothers and fathers; primary analyses are conducted separately for each, generating gender-specific ITT estimates.
Outcome Measures
Parenting practices are assessed using a 10-item subset of the validated short-EMBU Parent version (s-EMBU-P), covering three subscales:
- Rejection/Harsh Parenting (HP): Four items assessing physical punishment, unwarranted anger, criticism, and disproportionate punishment. Higher scores indicate worse parenting.
- Emotional Warmth/Positive Parenting (PP): Four items assessing comfort, affection, praise, and pride. Higher scores indicate better parenting.
- Overprotection/Control (C): Two items assessing strict limits and accountability demands. Higher scores indicate worse parenting.
Items are rated on a four-point frequency scale (1 = No, never; 4 = Yes, most of the time). Each subscale score is computed as the mean of its constituent items. For composite analyses, the negative parenting subscales (HP and C) are reverse-coded prior to averaging, so that higher values across all composite indices indicate better parenting outcomes. Items are translated into Marathi, Telugu, and Hindi/Rajasthani using standard forward-back translation procedures.
Mediator Variables Intimate Partner Violence (IPV): IPV is measured using the instrument employed in the main ECOVI trial, including physical, sexual, economic, and emotional/psychological IPV subscales consistent with WHO measurement standards. Women's experience of IPV is assessed. The mediating variable is the change score (endline IPV minus baseline IPV), with negative values indicating a reduction in IPV.
Mental Health (GHQ-6): Psychological distress is measured using the General Health Questionnaire-6 (GHQ-6), scored 1-5 with higher scores indicating greater distress. The mediating variable is the change score (endline minus baseline GHQ-6), with negative values indicating improvement.
Relationship Quality: Change in a composite Relationship Satisfaction Index measuring perceptions of the couple as a team, happiness in the relationship, and relationship strength. Scored 1-5, with higher scores indicating a stronger perceived relationship. Positive change scores indicate improvement.
Financial Distress: Measured as the frequency and severity of household economic hardship over the past six months (scale 1-6, with higher values indicating lower distress), capturing perceived financial stress, food insecurity, difficulty meeting basic needs, reliance on borrowing or asset depletion, and indirect impacts on children's well-being. Positive change scores indicate reduced distress.
Covariates Baseline covariates included in all primary regression models: respondent age; respondent education; respondent socioeconomic status; residential type (rural vs. urban); living arrangement (co-residing with in-laws vs. not); gender norms; and state fixed effects (Maharashtra, Andhra Pradesh, Rajasthan). Baseline values of each mediator are also included in both mediator and outcome models to account for regression to the mean and strengthen identification.
Analytical Strategy Primary ITT Analysis
The primary analysis estimates the Intent-to-Treat (ITT) effect of the intervention on each s-EMBU-P subscale score and on composite indices using OLS regression:
Yᵢ = α + β·Tᵢ + γ·Xᵢ + εᵢ where Yᵢ is the standardised parenting score of individual i at endline; Tᵢ is a binary treatment indicator (1 = intervention, 0 = control); Xᵢ is a vector of pre-specified baseline covariates; and εᵢ is the error term. Standard errors are clustered at the community level. The coefficient β is the ITT estimate of interest.
To address multiple testing, a composite Negative Parenting Index (mean of standardised HP and C scores, reverse-coded so higher = worse) and a Positive Parenting Index (standardised PP score) are pre-specified as primary outcomes. Individual subscale regressions are reported as secondary outcomes.
Mediation Analysis Causal mediation analysis decomposes the total treatment effect into: (a) the Average Causal Mediation Effect (ACME) - the indirect effect operating through the mediator - and (b) the Average Direct Effect (ADE) - the effect not operating through the mediator. All mediators are defined as change scores between baseline and endline. Each mediator is analysed separately in the primary mediation models.
Because mediators are not randomised, identification relies on the sequential ignorability assumption. Mediation estimates are therefore interpreted as indicative of plausible pathways rather than definitive causal mechanisms. Sensitivity analyses estimate the sensitivity parameter ρ (the correlation between error terms of the mediator and outcome models) and report the value of ρ at which the ACME would equal zero.
An additional exploratory structural equation modelling (SEM) analysis examines multiple mediating pathways simultaneously, including indirect and sequential pathways. SEM results are interpreted as descriptive of pathway structure rather than causal identification.
Ethical Considerations This sub-study is conducted under the ethical approval obtained for the main ECOVI trial by the Technical University of Munich and PCI India. The addition of parenting questions to the endline survey constitutes a minor amendment to the data collection instrument; relevant ethics board approval is obtained prior to fielding the amended endline survey. All participants provide written informed consent, are informed of their right to withdraw at any time, and are assured of the confidentiality of their responses.
