Measuring Women’s Empowerment with Just Five Questions · The MASI Method

Project Date
2023

Theme
Novel Methods and Measurement Approaches

The Challenge · Capturing the Unseen

How do you accurately measure a concept as subtle and layered as "empowerment"? Traditional large-scale surveys rely on short, standardized questions that often miss the nuances of real life—the quiet acts of resistance, the context of a decision, or the subjective feeling of choice. What tools can social scientists use to create appropriate metrics? The challenge was clear: How do we create a survey that captures the complex, human truth of a woman’s life in just a handful of questions?

The Mission · Bridging the Depth-Scale Divide

Our mission was to pioneer a new approach that honors the richness of in-depth qualitative data while delivering the efficiency and scalability of quantitative measurement.

Our Approach · The MASI Method

To achieve this, we developed the MASI Method (Machine-learning and Semi-structured Interviews). The strategy involved three critical steps:

Establishing the Gold Standard · We conducted in-depth, open-ended interviews with 209 married women in Haryana, India. Our qualitative team meticulously analyzed these narratives to create a nuanced empowerment score (1-4) for each woman, based on her context, attitude, and resistance. This became our "gold standard."

Conducting the Survey · We simultaneously asked the same women a battery of 63 standard survey questions covering various aspects of agency (e.g., finance, mobility, household decisions).

Translating with Machine Learning · We used a Machine Learning algorithm to analyze the data. The algorithm's task was to identify which subset of the 63 survey questions could most accurately predict the "gold standard" empowerment score we derived from the deep qualitative interviews.

Key insights · Why Qualitative Depth Matters

Our in-depth, narrative analysis revealed that empowerment isn't just about what a woman can do, but about her attitude, resistance, and subjective choice—nuances that traditional surveys simply cannot capture. We had to understand the story to get the score right.

1. Beyond the Numbers: The Discovery of Multiple Realities

These two images capture the complex reality of women in Haryana, one of India’s most conservative states: women veiled in the presence of men and elders coexist with female wrestlers who compete in traditionally masculine spaces.

This juxtaposition reveals that empowerment cannot be measured by a single metric: some women must exercise agency within traditional roles while others can challenge norms directly. This coexistence reminds us that true understanding requires acknowledging multiple realities rather than reducing complex social dynamics to simple narratives.

The Discovery
We realized that empowerment is not a monolithic concept; it exists as a set of multiple, coexisting realities. Our methodology had to be sensitive enough to validate both the woman who successfully resists social restrictions and the woman who skillfully exercises power within them.

2. Hidden in the Details: Resistance Within Restriction
In-depth interviews allowed us to capture something surveys could potentially miss: the difference between a woman who accepts restrictions and one who quietly resists them. To understand how this played out, let’s look at mobility as an example.

Women were ranked from 1 to 4 based on their level of empowerment, with 1 being the lowest and 4 the highest.

Score 1: The woman required explicit permission to leave the house and was always accompanied—whether visiting a village shop, the hospital, or her parents’ home.
Score 2: The external restrictions remained the same, but her transcript showed her pushing back against them or actively trying to resist—questioning rules, negotiating outings, or expressing frustration. This qualitative evidence of agency shifted her score upward.
Score 3: She had gained partial freedom, such as walking alone within the village, but still needed accompaniment for trips involving transportation.
Score 4: She had full mobility and could travel anywhere unaccompanied.

These gradations matter because they reveal a woman’s level of empowerment that might otherwise remain invisible.

The Discovery
We learned that agency is an attitude. A standard survey only measures the external result ("Did she go alone?"). Our qualitative data measured the internal process ("Did she fight to go alone?"), allowing us to validate a woman's true empowerment despite her external constraints.

3. Choice Within Constraint: When Compliance Isn’t Powerlessness

A standard financial empowerment survey asks, "Do you control your own income?" Our qualitative data showed that even when control is absent, choice may still be present.

We interviewed a working woman whose income went directly to the family head. Objectively, she ranked as the lowest (Score 1). However, her full story revealed a respectful joint family where her opinions were highly valued, and she willingly contributed her money to the family pool. We coded her as a Score 3—someone exercising choice within a constraint—because her participation was based on respect and inclusion.

The Discovery
We learned that willing participation, grounded in respect, is a form of empowerment. Rigid, objective categories would have missed this complex, personal reality entirely.

The Impact · A Practical, Scalable Gold Standard

The real innovation wasn’t just in understanding women’s empowerment more deeply—it was in figuring out how to translate that understanding into something practical and scalable.

After analyzing all 209 rich interviews to establish our "gold standard" empowerment scores, we took our comprehensive set of 63 standard survey questions and posed the crucial question to machine learning: Which questions truly matter?

We turned to machine learning for the answer. The algorithm analyzed patterns, identified correlations, and pinpointed which questions mattered most. The finding was striking: just five questions performed nearly as well as all 63 combined.

The Breakthrough
The machine learning identified just five key questions that performed nearly as well as all 63 combined.

This is the MASI method in action—using in-depth interviews to establish what we should be measuring, then using machine learning to identify the most efficient way to measure it at scale. The result is a new approach that honors the complexity of concepts like empowerment while making them practical to measure across large populations.

Publication

Jayachandran, S., Biradavolu, M., & Cooper, J. (2023). Using machine learning and qualitative interviews to design a five-question survey module for women’s agency. World Development, 161, 106076.

Read Full Paper

The Team & their Contribution

Seema Jayachandran
Principal Investigator, Princeton University

Key Contribution
Led the overall study design and quantitative analysis.

Monica Biradavolu
Principal Investigator, QualAnalytics

Key Contribution
Led the qualitative research as Principal Investigator, designing the interview methodology and analytical strategy that underpin the MASI approach. She also trained Anubha Aggarwal and Ambika Chopra to conduct interviews in Haryana.