How Attitudes Spread: Using Social Network Analysis to Evaluate Women’s Ideas towards Empowerment

Maya Ranganath and Aditi Patel. We will both be taking the second half of the course.

Abstract:

We will perform a social network analysis using already existing data from rural India to show how ideas of women and girls’ aspirations are spread throughout networks. We will do this by creating network datasets based on selected characteristics (living in the same village, attending the same school, frequenting the same health centre, participating in the same village meetings), and attribute datasets based on general attitudes regarding women’s role in society, information about their physical mobility, and participation in public life. We hope to see how girls’ networks affect these attributes, and to provide a model for how organizations could employ a similar methodology to effectively and positively influence women’s attitudes  about gender relations to empower themselves.

Question:

How do social networks influence women and girls’ aspirations? How can social networks be leveraged to encourage women and girls to have greater aspirations for themselves/their children? What implications do these questions have for development program design, monitoring and evaluation?

Context:

In recent years, India has made international headlines for it’s violent discrimination against women and girls. As women have gained increasing power and more prominent roles in Indian public life, the patriarchal backlash has ranged from a constant low-level of discrimination and harassment, to outright violence and rape. In this context, the question of how to empower women and girls to thrive in the face of these challenges is a topic that has increasingly been brought to the forefront of the global consciousness. The solution to this problem is not a neat one, and academics and practitioners have attempted to address it using various research methodologies, from feminist inquiry to anthropology, from sociology to psychology.

A team of researchers, led by Esther Duflo at the Massachusetts Institute of Technology  and Rohini Pande at Harvard University, chose to take an econometric approach to arrive at the conclusion that “exposure to a female leader seems to improve perceptions of female leader effectiveness and to weaken stereotypes against gender roles in the public and domestic spheres”. This study is particularly relevant as India experiments with “reservations”, or quotas, for elected female representatives across levels of government. Some districts have already been implementing this at the local level, since in 1993 India passed  a constitutional amendment for a random one third of village council leader, or pradhan, positions to be reserved for women.

Duflo et al surveyed both parents and adolescents in 495 villages of West Bengal, India. The study was carried out in Birbhum District, in which most people reside in villages and work in agriculture. At the time of the study, the sex ratio in the district was above average for the country, at 950 females per 1000 males, as opposed to 946 females per 1000 males for the average Indian rural population. The female literacy rate was low, though above the 2001 national rural average, at 51.55 per cent of women, as compared to the national rural average of 46.1 per cent. These characteristics--of a predominantly rural, agricultural, district where women on average are disadvantaged--make Birbhum a particularly relevant district to study, as lessons from Birbhum can be carried to several other districts in India with very similar demographic profiles.

Of the 495 villages in Birbhum District, the researchers randomly assigned a proportion to have women leaders in two election seasons, one election season, or never. Both parents and children in these villages were then surveyed on the topics of, “desired educational attainment, desired age of marriage, preferred occupation at the age of 25, and if the parent wished for their child to become a village leader or the child herself hoped to become a village leader” to see if the various levels of exposure to female leaders affected their responses to these questions.

In their study, they found that compared to villages that were never reserved, the gender gap in aspirations closed by 25% in parents and 32% in adolescents in villages assigned to a female leader for two election cycles. The gender gap in adolescent educational attainment is diminished and girls spent less time on household chores.

Objectives:

We plan to use the publicly-available data from the above project, but to a different end than the original researchers employed. We believe that social network analysis can be used effectively to understand how relationships between women and girls in these villages influenced their aspirations and beliefs about women’s roles in society.

The aforementioned study examined the vertical effect of how leaders influenced girls’ and parents’ feeling on several issues. We plan to use social network analysis to answer the question of how, rather than the leaders affecting the perceptions among the various individuals, these ideas were disseminated through the women’s social networks (the horizontal effect.)

Hypothesis: 

Attitudes about women’s roles in society, aspirations for children, and political participation are influenced and reinforced by women’s social networks within the villages of Birbhum District.

Although the study we are referencing demonstrates that a woman leader is a strong role model for women and girls in her geographic region, we will use social network analysis to show that for the same women and girls, shared attitudes and values about women’s empowerment are influenced by their relationships with one another.

Project Plan:

We will perform a social network analysis to investigate these issues by taking the following steps:

Building datasets:

1. Using the data available to put together a one-mode network to show which women knew each other. We will do this by using the following categories as proxies for social affiliation:
1. Women and girls who live in the same village
2. Girls attending the same schools
3. Women attending the same meetings at the same time
4. Households who go to the same health centres

If they fulfill any two of the above four criteria, we will categorize them as a ‘1’, meaning that they know each other. If they do not fulfil any of the above criteria, they will be categorized as ‘0’, meaning that they do not know each other.

2. We will also put together a one-mode network to show which girls knew each other, by following the same process.

3. We will create two attribute datasets, one each for women and for girls, and assign it to the above datasets respectively. The attributes we will use (found in the survey data from the Duflo et al study)  include:
1. General attitudes, such as the following, measured on the Lickert Scale (strongly agree, agree, disagree, strongly disagree, neither agree nor disagree):
1. A man is never justified in hitting his wife
2. For the most part it is better to be a man than to be a woman
3. It would be a good idea to elect a woman as the president of India
4. A wife shouldn’t contradict her husband in public
5. Women should be cherished and projected by men
6. Men should be willing to sacrifice their own well-being in order to provide financially for the women in their lives
7. Preschool children suffer if their mother works
   
   
   We will dichotomize the data by assigning 1 or 0 values to a pre-defined subset of the criteria above. First, we will convert the statements into a form that implies misogyny (changing “A man is never justified in hitting his wife”, to “A man is justified in hitting his wife”). Then, we will combine all answers as either agree or disagree. Next, if a respondent agrees with 4 out of 7, we will assign her a ‘0’ value, meaning that she has negative attitudes about women.
   
   
   Similarly, we will create include the following attribute data in the dataset (created the same way as outlined above, with more female-positive statements ranked higher than lower, and the higher statements assigned a 1):
   
   
2. Aspirations for their children (for women), for example:

1. What is the highest educational level that you would like your child to complete?
2. At what age would you like your child to marry?
3. Would you like your child to be the GP Pradhan (leader) of the village?
   
   

3. Aspirations for themselves (for girls)- using the above examples, but rephrased.
   
   
4. Physical mobility (for women), for example:

4. How often they traveled outside their village in the past month
5. How often they took public transport in the past month
6. Whether they are allowed to travel outside their village unescorted
7. Whether they participate in self-help groups/savings groups
   
   

5. Physical mobility (for girls)- using the above examples.
   
   
6. Women’s participation in public life: we will determine women as ‘involved =1’ or “not involved = 0’, based on the following criteria:

8. Are they the main financial decision makers in their households?
9. Do they regularly attend village government meetings?
10. Do they often speak up at village government meetings?
11. Have they personally approached local village authorities to get any issues resolved?
12. Do they vote in elections?
13. Do they know the names of all their local political/government leaders?

Questions we want to answer using social network analysis

How do social networks influence women and girls’ aspirations? How can social networks be leveraged to encourage women and girls to have greater aspirations for themselves/their children? By using social network analysis, we will aim to answer these questions.

1. What are the network patterns formed between women who share gender attitudes?
2. What are the network patterns formed between women who share aspirations for their children?
3. What are the network patterns formed between women who are allowed physical mobility?
4. Do they share geographic locations? Children’s schools? Health centers?
5. Who are the most well-connected women, using various centrality measures? Do women with certain attitudes or aspirations tend to be clustered around a particular woman?
6. Do women with similar attitudes and aspirations also have children who share attitudes and aspirations?

For each of the above questions we will see if the women share geographic locations, children’s schools, and health centres.

7. If a woman holds more decision making power in her household, and participates regularly in village governance and political life (i.e., she has the attribute of being “involved”), then does she also have attitudes and aspirations that are progressive? If so, are these attitudes shared by other woman in the village? Where is she located on the village network map? Do women who are more “visible” in the public life of a village lead to the diffusion of progressive ideas?

So what?

By performing this analysis, we can see how women’s aspirations diffuse throughout networks and become widespread. If our hypothesis proves correct, the data from this information could be used to identify which girls and women are key people to spread messages of empowerment, and programs that aim at women’s empowerment can more accurately target how to achieve this goal. Furthermore, targeting such programs to the key players in women’s social networks is a more cost-effective method  of disseminating ideas. We hope to show that social network analysis has an important part to play in the design, monitoring, and evaluation of women’s empowerment programs and beyond.

Potential Limitations:

The data we are using is of very high quality, given that it was collected by researchers at the Abdul Latif Jameel Poverty Action Lab (J-PAL), an organization that employs rigourous survey and data collection standards. However, given that the surveys do not directly ask questions about inter-personal relationships, a large part of our study is based on assumptions and conjecture about relationships between women and girls. We believe that given the realities of village life in India, these are very valid assumptions to make (for example, that women residing in the same village know one another)--however, we think it important to highlight that these are nonetheless assumptions that we are making from the data provided.

We are also conscious of the limitations of including too much data, and over the next few weeks, will be refining the attributes we choose to include. We want to be selective with the attributes we include, as coding them can become too unwieldy.

Data Source:

Lori Beaman; Raghabendra Chattopadhyay; Esther Duflo; Rohini Pande; Petia Topalova, 2011, "Powerful women and aspirations in India", http://hdl.handle.net/1902.1/17154 UNF:5:3R+IJ7hWkUcJ6SzNc3s2eA== V2 [Version]HOw