Friday, October 24, 2014

Understanding the networks of the Kurdish women fighting IS

Understanding the networks of the Kurdish women fighting IS

Meghana Kumar
(not taking the second module)


The YPJ or Women’s Protection Unit, which was formed in 2012, has been making international headlines of late. It appears to be playing a significant role in the fight to protect the Kurdish people from the likes of the extremist group, Islamic State (IS), the Syrian government and the al-Nusra Front, which is affiliated to Al Qaeda.[1]

One of the key reasons behind the attention is the fact that the YPJ is an all female militia group. Its roots lie in the Kurdish resistance movement and it is estimated to number over 7,000 women who fight at the frontline alongside their male counterparts from the YPG (People’s Protection Units) and the Kurdish Pershmerga. [2] A reported recent suicide attack by a 20-year-old YPJ fighter in the Syrian town of Kobane, which resulted in the deaths of IS fighters, has pushed the group further into the spotlight.

Apparently, the group receives no funding from Western nations – its key source of funding and supplies is its community.[3] The militia members are unpaid and the group operates on a volunteer basis – the women can leave when they like.[4]

Primary Question

Given that the group was formed so recently, little is known about the way it operates and the network connections within it. I have a few main questions and would use this project to understand more about the relationships within the YPJ – who are the connectors, where are the key relational ties and what are the factors motivating the girls and women to join. Were these women connected prior to joining or are they only connected by virtue of being part of the YPJ? In addition, I would want to know how many of them have connections to the YPG and the Pershmerga and identify who the boundary spanners are.


In order to conduct the analysis I would need to know if and how the women were connected before they joined the YPJ and who they interact with the most while in the group. I would also want to identify the attributes of those women who display the most connectivity and whether they are the ones who also have the greatest access to resources such as arms and other supplies. It would also be useful to identify whether the women that are better connected to the YPJ’s male counterparts and by virtue of this may have better access to resources, are the ones who are also the ones who are the most connected within the YPJ itself.

The biggest challenge with this project would be collecting the data and gaining access to the thousands of YPJ fighters.

Important Network Measures

Key network measures will include density, eigenvector and degree. It will be interesting to note who the most connected / influential nodes are and who is connected to them. Given that the YPJ are carrying out planned operations against groups, which have more sophisticated as well as a greater number of weapons, it is important that those who do have access to supplies in the group are not operating on its fringes.


A social network analysis of the YPJ is not only interesting from an academic point of view as to the way different militias organize but could also help identify key women in this group that outside actors could work with in the fight against IS.

[1] “YPJ: The Kurdish feminists fighting Islamic State”, The Week, 7 October 2014, available from
[2] Ibid.
[3] Elizabeth Griffin, “These Remarkable Women Are Fighting ISIS. It’s Time You Know Who They Are”, Marie Claire, available from
[4] Ibid.

Funder networks of social enterprise accelerators

I will be taking the second module.


Accelerators play a critical role in the social enterprise value chain. They provide a space and culture for creativity to flourish and for social entrepreneurs to network with each other. Importantly, accelerators also connect social entrepreneurs to funders. Funders of accelerators play a foundational role in enabling this industry of social innovation, but what motivates them to fund the accelerators, and do they fund accelerators because of what the accelerators do, or who else is funding them?

Last year, a fellow SNA Fletcher student sought to explore the connections between boards of funders and social enterprises, with people creating that connection; this analysis hoped to identify potential leverage points for collaboration. (

I hope to take a different view than this work by, instead, mapping the connections between funders and social enterprise accelerators, with funding creating the connections. This identifies the major funders of social enterprise a bit higher upstream in the social enterprise value chain, rather than looking at direct funding of social enterprises. I also hope to explore whether the relationships that funders have with other funders (via mutually-supported accelerators) are leveraged by an accelerator to provide funding to the projects they support.

Research questions 

Who are the primary funders of social enterprise accelerators? What influences the funder's decision to fund an accelerator - relationships with other funders or a theme the accelerator focuses on?

Secondary: If accelerators could connect to other accelerators through a mutual funder, what does that network look like? Do first- or second-degree connections to funders play a role in what funding the accelerators' projects secure?


Accelerators are major actors in connecting social entrepreneurs and funders; one might say they are the ones actively building the social enterprise network. I hypothesize that SNA can 1) uncover the major sources of funding these accelerators are tapping into and 2) show whether accelerators expand the funding available for their projects by leveraging the network of their funders and/or the network of accelerators who share a mutual funder.

SNA methodology

This network is two-mode (Funders and Accelerators). To identify major funders, I will find the highest out-degree funders in the two-mode data. To understand the influences on their decision (relationships with funders vs. theme), I will analyze whether groups of funder's out-degree connections to accelerators parallel each other, or whether funders out-degree connections parallel thematic similarities between accelerators instead. I will compare/contrast different kinds of funders (e.g. foundations, private sector, etc) relationship characteristics to each other.

Secondary: To identify the network of funders available for a particular accelerator, I will convert the two-mode data sets to two one-mode data sets. In one, the accelerator-accelerator connections will be if they share a funder. I will analyze the most connected accelerators – are the funders they are able to secure for their projects the funders from the other accelerator's network?

I will collect all data required by locating popular and commonly recognized accelerators and finding who their founding funders were.  Then, I will identify the most central accelerators (highest degree, eigenvector via funders, etc), identify their 2013 projects' funders and compare this list of funders to the funders of accelerators; are they founding funders or second-degree connections who those founding funders have relationship with?


This analysis aims to uncover the networks of funders that underlie the funding of social enterprise accelerators (and its projects). By outlining patterns of funding flows along particular relational or thematic lines, it may clarify how funders make decisions to fund accelerators. Additionally, the analysis might shed some light to help accelerators better understand how to prioritize relationships with funders if it impacts what funding their projects are able to receive.

Thursday, October 23, 2014

Social Network Analysis for Internal Organization

Shalini Sharan (Not taking the second module)

Effective internal organization is vital to success and growth of a firm. Innovation, collaboration, and cross-fertilization are becoming increasingly essential for firms in their efforts to separate themselves in a highly competitive landscape. Employees are key to attaining the aforementioned goals.

In order to create such an atmosphere there has been a popular step by many industries to make “open spaces” in workplaces where interaction is facilitated through thoughtful interior designing. While such a shift is still debated by many I will look at an example of a workplace where this idea was implemented in order to discern any measurable changes. My focus of study will be my previous employer, a think-tank in DC, which implemented such a change a year ago. Since the think-tank has multiple programs spanning several regions and topic areas it was envisioned that a new office building should have open spaces where mixing of program offices would encourage cross-fertilization of ideas and inspire collaboration among programs.

Hypothesis: Smart designing of office spaces with open areas that facilitate interactions will create new ideas, promote more inter-program connections, and increase the overall creative output of the firm.

Data : My main tool of gathering data will be a simple survey. The employee attributes would be collected as 1) position: research assistant, research associate (junior staff) program fellow, program manager (senior staff) 2) the program to which they belong (region such as Europe, East Asia etc., or topic such as non-proliferation, trade etc.)

The survey will ask a few key questions: 1) How often they interacted with each employee on work related issues before the change (rarely, sometimes or frequently) 2) How often they interacted after the change on non work related issues (rarely, sometimes or frequently). I will repeat these questions for non-work related interaction as well.

In addition, I will use the floor plan from previous building and new building to add an attribute to each employee by the floor (number) they worked on. I will also create a database of reports and conferences in the last year and the reports and conferences after the change to add another layer of complex analysis. In my previous observations it was not uncommon for each floor to have a tight knit group and limited interaction with programs on other floors.

This data help me observe how clique-y the old setup was and if much has changed since the move to the new building. Network density before and after the change would offer insights on how well utilized the connections are due to the change. Some of the questions I will address through social networks analysis: Are the employees interacting more with other programs now that the floors have been shuffled or are they still maintaining their old relationships? Is the interaction largely work-related or not? Such patterns will help me discern if there was a strong sense of ‘familiarity’ that has been broken or now and if there is indeed a meaningful and substantive exchange of ideas related to work among new connections. It would also be interesting to study if junior staff have been able to increase interaction with senior staff, especially in other programs. A general increase in inter-program interaction will be useful. We may find that there is only an increase in non-work interaction but that may be step towards more professional collaboration. Subsequent studies can map the same data to see if that occurs.

Within the interactions before and after the reshuffling I will also identify the outliers, connectors, and boundary spanners of the organization. Sub grouping by programs will help me identify the leaders in each program and the connectors to other programs. Ego networks within programs will help identify the key staff members who should be interacting with other actors within different programs. The boundary spanners will be important to identify because they are actors who can create overall cohesion within the firm. Identifying these actors by their job category will help determine if the junior staff is more engaged than the senior staff. In degrees and out degrees will also identify how interactions are happening within programs in order to facilitate interaction between the “influencers” of each program.

Looking at the patterns of actual events and programs by simple graphs will help put these interactions in perspective and see if joint conferences, events and reports were conducted. This is, however, not a definitive set of analysis because not all meaningful interactions materialize in such a way.

Conclusions: Social Network Analysis will help me measure the true changes in the organization and behavior of employees after the shift of office space. It will help to assess if transforming the nature of a workplace leads to employees interaction, learning from peers and exchange of ideas for the betterment of the firm and the firm’s product offering.

Food Insecurity & Trading Networks in Northeastern Uganda: Finding Opportunities for Market Interventions

Eric Jospe (I will be taking the second module)

Food insecurity in rural regions of developing countries is characterized by high prices of staple commodities, high price variation over small distances, and market dependence. Karamoja, the poorest and least developed region of Uganda, is a case in point. The population experiences chronic food insecurity, with high rates of global acute malnutrition across the region, and large differences in market prices within and between districts. Mercy Corps Uganda recently conducted a network mapping of the staple crop food trading system in three districts of Karamoja. Their preliminary analysis shows the most fragmented trading network in the poorest of the three districts, and the most connected trading network in the wealthiest of the three. In order to design an effective livelihoods program, they would like further analyses conducted, particularly breaking down the networks by the type of crop traded.

Research Questions
Some of the crops are not as important as others. Sorghum is the most important, with beans and maize among the other important ones for market-dependent households. If the trading networks are broken down by commodity traded, focusing on these three crops, is it possible to find other features that would give Mercy Corps Uganda leads to make the food trading system more efficient? Are bottlenecks causing price discrimination for one or more crops? What characteristics of traders are the most important in determining who they trade with, and how frequently they trade? Are there individual traders within and between markets that can be connected to reduce price variation?

Social network analysis of the staple crop food trading system in Karamoja, Uganda will reveal the specific traders that can be connected to make the market more efficient. By analyzing the network by each commodity traded, bottlenecks will emerge, along with potential bridges. Attribute data showing the characteristics of traders that make them more likely to trade can be used to identify potential trade partners across markets and districts.

Fortunately, the data has already been gathered for this project. The survey was conducted over a two-month period in 2014. Local staff of Mercy Corps Uganda used a snowballing technique to interview every trader they could find in three districts of the Karamoja region in northeastern Uganda. Overall, 283 traders were surveyed, with 543 trading relationships identified. Respondent information includes name, district, gender, age, ethnicity, religion, birth parish, ownership status, decision-maker status, residence, whether they trade with other traders, if so how many, and whether they sell to consumers of the same parish or other parishes. Data on trading partners includes which individuals/markets they trade with most frequently (1-6), the number of commodities traded (1-4), commodity type, frequency of trade, and quantity traded.

I will transform the data from its current state in an Excel spreadsheet into a one-mode, directed network showing how traders are connected to each other, as well as an attribute dataset. Other possibilities are to create a 2-mode undirected network showing traders by marketplace and a 2-mode undirected network showing traders by commodity. Centrality measures that have been highlighted in the preliminary analysis were clustering coefficient and betweenness centrality. The former did not prove to be an effective measure in this type of network, but the latter did. Additional measures that may highlight important traders are eigenvector and closeness. As the survey asked about frequency and quantity of trade, this can be viewed as stronger ties.


This is a new frontier in the analysis of rural trading networks in the developing world. Social network analysis has the potential to identify structural explanations for inefficient market characteristics that keep the rural poor market-dependent. By identifying potential bridges within and between markets and districts, SNA could help Mercy Corps develop an effective market intervention to reduce prices of primary commodities and improve food security.

Network Analysis for Conflict Prevention: Examining Ex-Combatant Networks and Political Actor Influencers in Burundi

Sarah Collman (will not be taking 2nd module)

Network Analysis for Conflict Prevention: Examining Ex-Combatant Networks and Political Actor Influencers in Burundi


After gaining independence from Belgium in 1962, Burundi experienced several waves of violence and mass killing between two ethnic groups, Hutus and Tutsis. Prior to the civil war, the two most extreme events occurred in 1972 and 1993. In 1972, Tutsis were systematically targeted by Hutus, which led to massive reprisal attacks on Hutus (estimated 100,000 killed). Subsequently in 1993, following the first democratic election of a Hutu-led government, Tutsis assassinated Ndadaye, the leader of the Hutu government. This in turn spurred mass killing of Tutsis by Hutus. It is estimated that 250,000 people died between 1972 and 1993.

In April 1994, a plane carrying the Hutu president of Burundi, Cyprien Ntaryamira, and Juvenal Habyarimana, president of Rwanda, was shot down. For the next 11 years, Burundi spiraled into a bloody civil war, in which Hutu rebel groups fought the Tutsi-dominated national army. The two prominent rebel groups, which eventually formed political parties once a peace agreement was reached in 2005, National Council for the Defense of Democracy – Forces for the Defense of Democracy (CNDD-FDD) and the National Forces of Liberation (FNL).

Although the war officially ended in 2005, even though the FNL did not sign the peace agreement until 2009, Burundi has experienced a fragile peace that has been colored by political violence (surrounding the 2010 election), a tightening of the government’s grip on power, and continuing ethnic tensions. In the lead up to the last election, the ruling CNDD-FDD’s youth wing, the Imbonerakure, allegedly committed political violence, which spurred reprisal attacks. Due to its violent past, Burundi has the potential for further conflict ahead of and during the upcoming June 2015 election.

Primary Questions

In Burundi’s transition from civil war to a post-conflict state, were connections and networks of armed group ex-combatants maintained and do they remain strong today? 

Are there certain political figures that have the capacity to influence members through these underlying networks? 

In addition to networks within armed groups (or among ex-combatants in former armed groups), is the CNDD-FDD in Burundi connected to any of the more than fifty armed groups in neighboring Democratic Republic of the Congo? 

If so, can an analysis of these networks be useful in a conflict prevention approach leading up to the June 2015 election?


If underlying networks from armed groups during the civil war exist, political actors with influence may be able to recruit ex-combatants to join the CNDD-FDD and the associated youth wing, which is alleged to have committed political violence during the last election in 2010.


Obtaining credible and easily accessible data for this project may be an obstacle. The currently sensitive nature of the conflict could prove difficult for data collection. Ways to obtain information about networks of armed group ex-combatants include interviews, surveys, and observation of communication patterns. In order to identify powerful political actors and levels of influence, a stakeholder analysis must be conducted. It would be useful to gather two sets of data: one set to gauge networks and connections of armed group networks during the war (around 2000 or so) and the other set in present day. It would be interesting to track how the networks changed and identify underlying networks that have been maintained in the transition from ongoing conflict to a post-conflict state.

In assessing the data, important measures to consider include subgroups to identify factions, and centrality measures such as in-degree and betweenness. Identifying key people that act as silos to other groups will be important to assess power and influence. A consideration of the attribute data may be interesting to determine who the ex-combatants are and what similarities they have in common with those they are connected to. This information will be useful in terms of conflict prevention, in order to identify certain groups to target.


Social network analysis can help map the network and connections between ex-combatants from the civil war and former rebel leaders that now hold positions in the government now. Conducting a stakeholder analysis and gaining insight into power and influence may be useful in predicting mobilization techniques for future violence. While obtaining data may be somewhat of a challenge, understanding the networks that have been maintained since the end of the civil war may prove to be useful in terms of conflict prevention.


·      Insight on Conflict, Burundi Conflict Profile, March 2014
·      International Crisis Group, Burundi: Ensuring Credible Elections, February 2010
·      Human Rights Watch, We’ll Tie You Up and Shoot You, May 2010

·      IRIN, Analysis: Upcoming polls to test Burundi’s fragile peace, November 2009