Anya Jajodia.

2022

Policy, data, and lived consequences.

Investigating women’s access to enterprise finance through PMMY lending data and public sentiment.

Role
Lead Researcher
Mentor and co-author
Reetu Jain
Publication
International Journal of Software & Hardware Research in Engineering (IJSHRE)

The question

What could lending data and public discussion reveal about an initiative intended to support women’s entrepreneurship?

I examined Pradhan Mantri Mudra Yojana (PMMY), an Indian government initiative supporting access to credit for micro-enterprises. The study combined exploratory analysis of financial-program data with NLP-based sentiment analysis of Twitter discussion.

I led the research, including data preparation, analysis, visualization, and writing, with guidance from my mentor and co-author, Reetu Jain. The work brought programming into a broader investigation of financial inclusion and the barriers surrounding women’s economic independence.

Two sources of evidence

Financial-program data

I examined secondary data covering financial years 2015–16 through 2021–22, drawing on PMMY sources and existing research.

The analysis compared loan accounts and sanctioned, disbursed, and outstanding amounts across the scheme’s Shishu, Kishore, and Tarun categories as defined during the study period. These comparisons examined how lending activity varied across categories and years.

Public discussion

I used Tweepy to collect a sample of 1,000 PMMY-related tweets. The Python workflow used NLTK for text preparation and TF-IDF to represent the relative importance of terms.

A sentiment model trained on movie-review data was then applied to classify the tweets as positive, neutral, or negative. This provided a view of expressed sentiment alongside the financial analysis.

Financial-program analysis

  1. Secondary PMMY dataFinancial years 2015–16 to 2021–22
  2. Exploratory analysisLoan accounts; sanctioned, disbursed and outstanding amounts
  3. Category and year comparisonsShishu · Kishore · Tarun

Twitter sentiment

  1. CollectionTweepy · 1,000 PMMY-related tweets
  2. Text preparationNLTK: segmentation, lemmatization, chunking
  3. TF-IDF featuresRelative importance of terms
  4. Sentiment classificationPositive · neutral · negative
Training materialMovie-review sentiment datasettrained the model applied in classification

InterpretEach strand read alongside the other, as partial evidence

What the study found

Lending activity changed over time

The paper reports an overall increase in disbursement across the study period, with a decline in financial year 2020–21 followed by recovery in 2021–22.

It discusses the pandemic as possible context for the decline, rather than establishing its causal effect.

The sampled discussion was predominantly positive

The paper reports these sentiment classifications within the collected tweet sample:

Sentiment classifications reported in the paper for a sample of 1,000 PMMY-related tweets. Results describe the collected sample, not the wider population.

These percentages describe the model’s classifications of the sampled tweets, not population-wide approval or classifier accuracy.

Credit was one part of a wider question

Alongside the analysis, the paper discussed barriers identified in the surrounding literature, including education, skills, mobility, household responsibilities, social expectations, and confidence.

These considerations framed the interpretation: access to a financial scheme does not, by itself, explain whether someone can make use of it.

Reading the results in context

Twitter users did not necessarily represent the rural women the initiative aimed to serve, particularly those without reliable internet access. The sentiment model also crossed domains, moving from movie-review training material to financial-policy discussion.

The two analyses therefore offered partial perspectives. Lending figures described program activity; tweet classifications described sampled discussion. Neither independently established a causal improvement in financial literacy, independence, or wellbeing.

Beyond the paper

Through a partnership with Saksharta spanning both of my research projects, I helped deliver educational sessions for interested girls and young women in underserved communities in Mumbai.

Alongside classes, we offered assistance, mentoring on request, and practical support for participation and learning. This was related community work, separate from the research datasets and findings.

What I took from it

This was an early opportunity to use programming to investigate a question beyond the software itself.

Working on the analysis alongside community education made me more attentive to who appears in a dataset, who is missing, and what additional understanding comes from speaking with people directly.

Tools and methods

  • Python
  • Tweepy
  • NLTK
  • TF-IDF
  • Sentiment classification
  • Matplotlib
  • Exploratory analysis of secondary financial data

Publication reference

Jajodia, A., and Jain, R. (2022). Analyzing The Impact Of Financial Schemes In Boosting The Lives Of Female Rural Population In India. International Journal of Software & Hardware Research in Engineering, 10(9), 93–101.