Positive Deviance: Positive Outliers Matter
By Jeremy Boy, Data Scientist, UNDP Accelerator Labs together with Andreas Gluecker, Project Manager, GIZ Data Lab.
The following piece was originally published by the UNDP Accelerator Lab Network on August, 21, 2020. For more details on the project background, please read our first blog on this series “Launching the Data Powered Positive Deviance Initiative”.
At the beginning of 2020, the UNDP Accelerator Labs Network kicked-off a set of exciting pilots with the GIZ Data Lab of the Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ) to explore the potential of Data Powered Positive Deviance (DPPD) for sustainable development.
Together with the UNDP Accelerator Labs and GIZ projects in Ecuador, Mexico, Niger, and Somalia/ Somaliland, we are using combinations of administrative data, satellite imagery, urban data, social media data, and mobility data to identify positive deviants in such diverse contexts as the borders of the Amazon forest, public spaces in Mexico City, and pastoralist villages in Somaliland. Our goal is to identify locally-developed and well-adapted solutions that can help broader communities overcome the development challenges they face.
A “Positive” Case Study
The Positive Deviance approach assumes that in every community, there are individuals or groups whose uncommon behaviors or coping mechanisms enable them to find better solutions to problems they face than their peers while having access to the same resources. Positive Deviance interventions aim to identify those individuals and groups and understand what attributes or traits differentiate them from others. These are then mobilized to help fellow community members overcome their problems together, by using locally accessible assets and resources. In short, instead of looking at the bad and trying to correct it, a Positive Deviance approach looks at the good and tries to collectively replicate it in situ.
Perhaps the most cited Positive Deviance story comes from Save the Children’s action in Northern Vietnam in the 1990s. At the time, 65% of all children between the ages of one and four living in Vietnamese villages were malnourished. Save the Children had a very clear mandate: find a sustainable, large-scale solution based on local resources that would show results within six-months.
The NGO’s staff on the ground set out to find families in which children showed no sign of malnutrition. The team identified several of these “positive deviants”, and discovered that caregivers would add sustenances like shrimp, crab, and fish to their children’s meals — sustenances that were available to everyone, but generally deemed inappropriate for young children. Acting on this insight, Save the Children developed activities with local communities to rehabilitate hundreds of malnourished children, and to teach families how to sustain their children at home, and on their own.
Since the 1990s, Positive Deviance approaches have been applied to a number of other development issues like school retention, neonatal mortality and pregnancy outcomes, salesforce productivity, violence against women, health care services, HIV transmission, and antibiotic-resistant bacteria transmission in hospitals.
Big Data, Positive Deviance, and Development
Traditional Positive Deviance initiatives rely heavily on primary and secondary sources of data, i.e., data collected directly for the purpose of the study, and repurposed data (like census data), respectively. Unfortunately, these data are either very costly (primary data), too highly aggregated, or focused on variables irrelevant to positively deviant behaviors of interest (secondary data). They may also be too limited in time (a snapshot rather than an evolving view of a situation).
Recent work by the GIZ Data Lab, UN Global Pulse Lab Jakarta, and the University of Manchester shows it is possible to leverage big data, i.e., high-volume, high-velocity, and high-variety data mainly generated by individuals as they go about their daily activities, to discover positive deviants in farming communities, and to isolate some of the factors that differentiate them from their peers.
Inspired by this, the UNDP Accelerator Labs and the GIZ Data Lab with technical support from the University of Manchester, kicked off a series of pilots within the Data Powered Positive Deviance (DPPD) initiative early 2020, with a call for proposals from interested Labs.
17 Accelerator Labs and 2 GIZ projects responded with 23 proposals. Four proposals were retained from Ecuador, Mexico, Niger, and Somalia, based on alignment between the local Accelerator Labs and GIZ portfolios, and their potential for using big data to identify positive deviants facing the challenges of deforestation, violence against women, sustainable agriculture, and water management.
Here’s what we are working on:
Ecuador: Mitigating Deforestation through Sustainable Cattle Raising
In 2017, a coalition of stakeholders including UNDP and the Ecuadorian Ministries of Environment and Agriculture launched the PROAmazonia initiative, an ambitious, five-year project that aims to transform the agriculture and forestry sectors in the Amazon region. Part of this ambition is to reduce CO2 emissions by mitigating deforestation in the Ecuadorian Amazon through the promotion of sustainable land-use that does not require cutting down additional forest for local agriculture. Our DPPD team members from GIZ Ecuador are working with the Ecuadorian Ministry of Environment and Water to implement its national bioeconomy strategy for the protection and sustainable use of biodiversity by developing socially, economically, and ecologically sustainable solutions.
Agricultural practices are one of the main causes of deforestation in the Ecuadorian Amazon. 99% of deforested areas are transformed into agricultural fields, and 64.9% of these are used as grassland for livestock.
The project team is using satellite imagery to identify positively deviant cattle farmers who operate in zones of expansion of the agricultural frontier without further contributing to deforestation.
Using additional sources of climate, socio-economic, and ethnographic data, the team will then investigate the contextual factors and characteristics that contribute to these farmers’ forest-friendly cattle raising practices.
In order to extrapolate results for future policy design based on local evidence, two territorial units of analysis were selected (north and south) considering that these regions have gone through different colonization and deforestation processes. The most complex part of the land cover classification analysis is to distinguish grassland with cattle from other types. We are hoping to use satellite images of methane realized by cows whilst eating and digesting to overcome this challenge.
Mexico: Addressing Violence Against Women In Public Spaces
In November 2019, Mexico City’s mayor issued a gender-based violence alert, effectively activating a series of measures to reduce violence against women. Some of the measures aimed at public spaces include renovating public corridors; placing panic alarm buttons and other sensory devices throughout the city’s streets to encourage women and bystanders to report incidents.
The project team is using the data collected by these devices to identify positively deviant public spaces (e.g., public transports, parks, corridors) where women are safest.
Using additional sources of data — such as, sentiment analysis of social networks, socio-economic census, perception of public security, urban infrastructure — the team will then investigate the contextual factors and characteristics that contribute to lower assault rates against women in these spaces.
Niger — Ensuring Food Security Through Sustainable Agriculture
Sustained agriculture in Niger is under tremendous pressure. On the one hand, prolonged land occupation for growing crops is a direct source of inter-community conflict. For example, deadly altercations have been reported between breeder and cultivator communities in the Tahoua and Maradi regions. On the other hand, climate change and the reduction of rainfall affect longer agricultural cycles, like that of sorgho which has an average cycle of six months. Crops are of lesser quality as they mature in a drier context, which in turn aggravates food insecurity.
The project team will use remote-sensing data to identify positively deviant cultivator communities that manage to accelerate their agricultural cycles, for instance by using soaking techniques, thereby shortening their land occupation and improving the quality of their crops.
Using additional sources of big data, the team will then investigate the contextual factors and characteristics that contribute to these communities’ ability to accelerate their practices, and whether this can help mitigate conflict.
Somalia/Somaliland: Supporting Pastoralist Communities Amidst Environmental Challenges
Between 2010 and 2012, and again in 2017, repeated droughts in Somalia caused more than a quarter of a million deaths and contributed to the displacement of roughly 4.2 million people. The country is still recovering from these events, and many Internally Displaced Persons (IDP), most of whom are from nomadic pastoralist communities, continue to gather around major urban areas.
Pastoralists represent around 55% of the Somalia/Somaliland population and they are an essential part of the region’s economic lifeblood. However, these communities are extremely vulnerable to environmental degradation and changes in water supply, and therefore are highly prone to internal displacement.
The project team will use remote sensing and mobility data to identify positively deviant pastoralist communities that continue to flourish despite the effects of climate change and avoid further congesting IDP settlements.
Using additional sources of official data from national and international organizations, the team will then investigate the contextual factors and characteristics that contribute to these communities’ ability to sustain their way of life.
A challenge we are facing with our ongoing work on positive deviance is acquiring large scale data on a wide geographic area. We are also exploring multiple data sources to give us a better understanding of the current landscape. Take, for example, using vaccination data as a proxy for the changes in livestock herd sizes before, during, and after the droughts. We are also reaching out to our local telecom provider to supplement this with CDR data and take into account the changes in location and mobility of the sample in our area of study.
What To Expect Next
The UNDP Accelerator Labs are starting to implement their pilots and are doing preliminary analysis to strengthen the implementation of their experiments. Their learnings, challenges, insights, and feedback will then be shared across the UNDP Accelerator Lab Network, and beyond.
Longterm, these four projects aim to inform targeted interventions to help communities respond to the challenges they face, using their “neighbors” as examples and mentors. However, for now, the ambition of the DPPD initiative is to identify what characterizes positive deviants in these different contexts with regard to the challenges people face, and how big data might contribute to this effort.
— — —
Watch this space for updates or follow along with the conversations online.
● Jeremy Boy @myjyby
● UNDP Accelerator Labs Network @UNDPAccLabs
● GIZ Data Lab @giz_data_lab
● Catherine Vogel: Catherine.firstname.lastname@example.org
With thanks to our UNDP Accelerator Lab colleagues for their contributions:
●Ana Grijalva, Head of Exploration, UNDP Ecuador Accelerator Lab
●Hodan Abdullahi, Head of Exploration, UNDP Somalia Accelerator Lab
●Gabriela Rios, UNDP Mexico Accelerator Lab
●Moustapha Sahirou Yacouba, UNDP Niger Accelerator Lab