Saturday, October 22, 2016

Building Better Mental Health Support for Peace Corps Volunteers

Kelsey Goodman
Not taking the 2nd module

Can the Peace Corps use social network analysis to provide better mental health support to volunteers? 

The Peace Corps is an international volunteer program, funded by the US Congress, that sends American volunteers to host countries abroad to provide technical assistance and to promote global friendship and cooperation. Since its founding in 1961, over 220,000 volunteers have served in 141 countries in program sectors like Education, Small Business Development, Environment, Youth Development and more.[1]

The Peace Corps Morocco program – where I served from 2013-2016 – transitioned in 2012 from a multisector program with Environment, Small Business Development, Health, and Youth Development volunteers to a single sector Youth Development program. Under this new arrangement with the Moroccan government, Peace Corps Morocco was officially hosted via formal partnership with the Moroccan Ministry of Youth and Sports (MYS).

The YD transition resulted in several programmatic changes:

1.     Moroccan program managers who oversaw volunteers based on sector were transitioned to “regional” managers who oversaw volunteers based on geographical location.
2.     Two to three intakes of 30-40 volunteers a year became one intake a year of 90-100 volunteers
3.     New sites were developed to accommodate larger intakes
4.     New sites were developed in bigger cities/towns because more volunteers needed to be in sites with MYS community centers (In towns < 5,000 inhabitants, community centers are typically either run by local civic associations or do not exist.)

I arrived one year after the youth development transition was implemented. My intake started off as 95 volunteers. Over the course of our two year service[2], about 21% of my intake (20 volunteers) terminated their service early, either as an “early termination” (volunteer chooses to leave) or a “medical separation” (volunteer develops a medical condition that Peace Corps cannot medically accommodate; mental health issues are considered a medical condition.)

This termination rate is consistent with other the post-YD transition intakes, however it is a higher termination rate than the multisector intakes from before the YD transition[3]. While some of these termination incidences were due to unforeseen family or medical emergencies, many were related to termination related to mental health reasons.

According to Peace Corps Morocco staff, correlated with the higher rates of termination, were also increased reports of unwanted attention, harassment, sexual assault[4],[5], as well as requests for mental health counselling[6].

Combined, these three developments signaled to Peace Corps that something was amiss with Peace Corps Morocco program.

Peace Corps Morocco attempted to address these issues in order to reduce termination rates, protect volunteers, and improve the resiliency and mental health of volunteers. First, they allocated special funds for an in-country therapist. However, there was difficulty finding a therapist in Morocco who was willing to accept volunteers as patients who could commit to working with Peace Corps in the long-term (American expatriates in Rabat are often transitional and temporary.) Second, they tried to re-transition to smaller sites with the hopes that volunteers would be exposed to less unwanted attention and harassment. Once again, Peace Corps was constrained by the resources in Morocco. The partnership with MYS states that volunteers must be supervised by a MYS employee, which limits volunteers to sites with a MYS community center.

While good site development and accessibility to mental health services are important, it is equally important to develop the resiliency of volunteers by strengthening their volunteer support networks.

Existing mental health research finds a correlation between weak support network of friends and higher occurrences of anxiety and depression.

A study of mental health in post-conflict Kosova demonstrated that only one type of social network – contact with friends – was associated with mental health outcomes. On a multivariate logistic regression, only those who “sometimes/never contacted friends” had a statistically significant association with anxiety and depression.[7]

In a study of a different population, researchers found that weak social support networks among older women were associated with feelings of loneliness.[8]

Furthermore, Peace Corps knows that volunteers with robust support networks are happier, more resilient volunteers. Peace Corps data shows that globally 78% of volunteers worldwide list “Spend time with friends” as an activity they do to manage stress and 77% “Contact others by phone, text, email, etc.” These activities are second only to “Read” and “Listen to Music”.[9]

This is why Peace Corps Morocco pays for a cellphone plan that includes free calling between volunteers and schedules all volunteers’ mandatory yearly flu shot in Rabat the day after Thanksgiving (so that the volunteers can all gather for Thanksgiving, and travel expenses can be covered by Peace Corps’ medical transportation budget.)

Research Questions

Is there an association between weak social support networks and terminations and/or mental health issues?

If so, how can Peace Corps use social network analysis to strengthen volunteers’ networks?

How diverse are volunteers’ networks? Are cliques formed by training location, site, or another attribute like age or race?

Who are the bridges between cliques? Could Peace Corps use these volunteers to build stronger networks?


Disregarding unexpected family and medical emergencies, volunteers who are more connected within the Peace Corps community are more likely to complete their entire 27 months of service.

Friendships between Peace Corps volunteers are formed under 3 circumstances: 1) Community Based Training (CBT) 2) Regional proximity 3) Similar backgrounds (homophilies)

Members of committees are more likely to be trusted for advice/seen as leaders than volunteers not involved in committees.

New volunteers trust the advice of old volunteers, but not the other way around.

Data Collection

In a perfect world without the constraints of things like the passage of time, I would be able to conduct an analysis of the social networks of pre-YD transition and post-YD transition volunteers, and compare them for connectivity as well as mental health

Since that is not possible, instead, I would like to conduct a survey of all volunteers, 7-9 months into the service of the most recent intake groups (i.e., if we survey volunteers in May 2017, one intake will have been in the country for 21 months, and one the other 9 months.) This time frame is optimal because it is after the newest volunteers have settled into a routine in their permanent site, but not so late that the older volunteers are busy closing their service and non-response rates would most likely be higher.

Because of HIPAA and Peace Corps confidentiality regulations, most of the data would need to be collected via survey. In my personal experience, volunteers have regular access to the internet and a lot of free time. Volunteers have the capability to complete the survey, 
however achieving high response rates on the survey requires buy-in by volunteers to the importance of this type of research. Respondents will be assigned pseudonyms to encourage truthfulness in their responses.

The one data point we will not collect through data is termination rate. That will be collected via Peace Corps Morocco simply by comparing the data from the survey with terminations as they happen. Peace Corps staff will be able to determine if the hypothesis is correct and if isolates are more likely to terminate than connected volunteers.

Network Questions
Question 1: Friendships – On a social basis, I talk to this person regularly (4); On a social basis, I talk to this person occasionally (3); On a social basis, I rarely talk to this person (2); I have never talked to this person (1), (valued)

Question 2: Leadership – I trust this person’s advice relating to Peace Corps policies or activities (4); I somewhat trust this person’s advice relating to Peace Corps policies or activities (3); I do not trust this person’s advice relating to Peace Corps policies and activities (2); I have never talked to this person (1). (valued)

Question 3: Collaboration – I worked on a work-related project, camp, or program with this person. (y) (n) (binary)

Attribute Questions
Question 3: What year did you begin your service? (Which intake)

Question 4: Where was your community based training site?

Question 5: What is your region?

Question 6: Of which of the following committees are you a member: GAD, VSN, VAC, SPSN, SHC, Wellness Retreat

Question 7: Diversity: Do you self-identify as minority in terms of race, color, national origin, religion, sex, gender identity (including gender expression), sexual orientation, disability, age, marital status, family/parental status, income derived from a public assistance program, political beliefs, or reprisal or retaliation for prior civil rights activity.[10] (y) (n)

Question 8: Throughout your service, have you experienced feelings of loneliness, isolation, anxiety, or depression?
·      Yes – intense/severe
·      Yes – moderate
·      Yes – mild
·      No


To perform the social network analysis, first we would visualize the data. Looking at how the friendship, leadership, and collaboration networks compare and looking for any correlations between attributes and clustering. We would also compare networks by intake and across intake.

Looking at whole network measures, we would analyze density and average distance of the networks. This gives us a feel for the level of interconnectedness of the intake (this is where it would be great to compare to density and average distance of the pre-YD transition intakes to the post-YD transition intakes.)

After looking at whole network measures, we would analyze the networks by subgroups. Analyzing subgroups by faction or Newman Girvan, we could draw out clusters of friends or colleagues and see if there is a correlation between certain attributes and ties. With a deeper dive, we could eyeball the factions for cliques.

Are there friendship cliques? Are cliques regional? Formed by training site? By minority status?
Are there cliques formed by happiness (i.e., do volunteers experiencing no or mild feelings of loneliness, isolation, anxiety and depression flock together vs. volunteers experiencing moderate or intense/severe feelings of loneliness, etc.?)

Comparing the friendship network to the collaboration network, we could see if volunteers choose to collaborate on project based on friendship or another factor (such as regional proximity.)

By comparing in-degrees across the friendship network and the leadership network we can see if volunteers’ perception of leaders is correlated with their friendships, or if there are certain individuals who are seen as leaders regardless of whether or not they are friends with each other. Looking at eigenvector we can determine the most influential volunteers to the network. These are the volunteers that Peace Corps could target to lead diversity trainings, wellness retreats or could encourage volunteers to embrace new policy or programmatic changes.

Looking at node centrality measures, we would look at whether or not a lower number of connections is correlated with more intense feelings of loneliness, isolation, anxiety, or depression. If so, is there any other patterns we can see in the data? Do these volunteers have certain similar attributes? Perhaps they are located in a region with poor transportation, making it hard to visit friends in other sites.

Do nodes with high eigenvectors have no or mild feelings of loneliness, isolation, anxiety, or depression? If so, this would support the hypothesis that well-connected volunteers are more likely to complete their service (absence of unforeseen circumstances).

Are there any nodes with high eigenvectors that have severe/intense feelings of loneliness, isolation, anxiety, or depression? If so, are they directly connected to other nodes with severe/intense feelings of loneliness? Does this suggest a “negativity” path or that there are certain “negativity influencers”?

Analyzing subgroups based on attribute data can illuminate areas of focus in which Peace Corps can improve its inclusion and training. If we compare the density of networks by region, we can see if certain regions have greater levels of support and collaborations than others. Does this correlate with better mental health?

Similarly, we could look at the network of volunteers who are self-identified as belonging to minority groups as a subgroup and as part as the network whole. By analyzing by Newman Girvan faction the friendship, leadership, and collaboration networks with a focus on the attribute of diversity, we can see if there are clusters formed by diversity. If these clusters appear in the friendship network, but not in the leadership or collaboration networks, this suggests that volunteers who self-identify as belonging to minority groups are well-connected within the whole network, but perhaps draw their emotional support from volunteers who have experienced similar challenges as them. If clusters appear in the friendship network and the leadership and/or collaboration networks, then this suggests that there are much more serious problems with diversity and inclusion in the Peace Corps network.

Finally, to look at the correlation between social network density, mental health, and termination, we would track the volunteers who terminate after the survey is complete. Who was in their egonet? Did they have ties to other volunteers who terminated? Is there a path of volunteers who terminated (domino effect?)  


Since this is an observational study, we cannot make any claims of causation. However, by analyzing the social networks of Peace Corps volunteers, we can make recommendations to Peace Corps concerning ways to improve the mental health support of volunteers which would hopefully reduce rates of termination and make volunteers happier, more productive, and more effective to their communities and to the mission of the Peace Corps.

For instance, we if can determine that isolates or volunteers with low degrees of connectivity are more likely to have severe/intense feelings of loneliness, isolation, anxiety, or depression then Peace Corps could consider using its discretionary programming funds to subsidize the wellness-focused social gatherings, like reimbursing volunteers for travel to the volunteer-run Wellness Retreats.

If we can identify the volunteers with the highest eigenvectors on the leadership network, then Peace Corps will know who they could ask to lead diversity and inclusion trainings at pre-service trainings and regional meetings so that volunteers will be more engaged in the content of the trainings because they will trust the importance of it. These are also the volunteers who are important when enacting program changes. For instance, if these volunteers embraced changes in policies for working in summer camps outside of their regions, then other volunteers perhaps would be more likely to follow their lead and comply instead violating the rules.

I am particularly interested in the support networks of volunteers who self-identify as members of minority groups (people of color, the LGBTQ+ community, religious minorities, senior citizens, people with special needs.) While all Peace Corps volunteers tend to get a certain degree of unwanted attention within their host countries simply because they are foreigners, volunteers who belong to certain minority groups often experience different forms unwanted attention or harassment and often deal with different stereotypes than volunteers who fit host country nationals’ perception of “America". While I believe Peace Corps staff in Washington and Morocco recognize the specific challenges that volunteers from different backgrounds face, in my experience volunteers were not always supportive or empathetic to the challenges that other volunteers faced (for instance, male volunteers saying female volunteers are overreacting when complaining about sexual harassment or volunteers minimizing the challenges that Muslim volunteers face because Morocco is a Muslim country.)

Challenges & Limitations

There are two missing pieces of the puzzle in terms of evaluating social support networks: social network within friends and family in the United States and social network within their communities with Moroccan friends and colleagues.

Unforeseen changes with friends and family in the United States (births, deaths, ultimatums issued by long-distance girlfriends) usually are factors in volunteers’ decisions to terminate. It would be interesting to analyze the data on those network, however for the scope of this study, it would not be feasible.

Relating to the other social network not studied, I hypothesize that the happiest volunteers are actually the ones whose main support network is located in their Moroccan communities, not necessarily to other volunteers. In fact, being too well-connected to the Peace Corps community might indicate that a volunteer travels frequently out of site and is not well-connected within their community.

Finally, to better understand the correlation between the density of social network and certain attributes, we would need to do a regression analysis on the node centrality measures and the attribute dataset. I do not believe this is possible to do in UCINET, thus for a more robust study, we need to utilize other analytical tools.

The content of this website is mine alone and does not necessarily reflect the views of the U.S. Government, the Peace Corps, or the Moroccan Government.

[2] For those who noticed an inconsistency in timeline: a normal Peace Corps service is 27 months. I extended for a year (including one-month home leave), so my total service length was 40 months.
[3] While statistics on termination rates (as well as statistics on crimes and unwanted attention) are monitored by Peace Corps, the public available information is not broken down by country in publicly available information. I am aware of the termination rates from the 2011, 2012, 2013, and 2014 intakes because my service overlapped with those intakes and, out of curiosity, track of the termination rates. These numbers are confirmed by Peace Corps Morocco staff.
[4] The lurking variable in these statistics is the passage of the Kate Puzey Volunteer Protection Act of 2011 which changed the way sexual assault was reported and processed by Peace Corps security and medical staff – specifically that groping and fondling were now categorized as sexual assault, not sexual harassment. With the change in sexual assault reporting, Peace Corps saw increasing rates of sexual assault worldwide, not just in Peace Corps Morocco. (Statistical Report of Crimes Against Volunteers 2014, pg 7) However, this does not take into account the increasing reports of unwanted attention and sexual harassment as the Kate Puzey act actually narrowed the definition of unwanted attention and sexual harassment by reclassifying groping and fondling as sexual harassment.
[5] This information is not publicly available, but is confirmed by Peace Corps Morocco staff
[6] Once again, this information is not publicly available, but is confirmed by Peace Corps Morocco staff

[7] Nakayama, Risa et. al. “Social networks and mental health in post-conflict Mitrovica, Kosova.” BMC Public Health. 17 November 2014.

[8] Gibney, S and M McGovern. “Social support networks and mental health: evidence from share.” J Epidemiol Community Health August 2011 Vol 65 Suppl 1

[10] This data could also be collected with a demographics survey; however, I would be concerned that volunteers may be sensitive about responding with information that will make them too easily identifiable.

1 comment:

Christopher Tunnard said...

This is an excellent and thoughtful idea, one that could have a very big impact on the Peace Corps if it could ever be accomplished. You demonstrate a very good knowledge of SNA, its methodology, and what it can yield. WOuld that this project cold actually be achieved...