Another Humanitarian Crisis in Somalia?: Learning from the 2011 famine - Interim Paper

An interesting paper looking at the 2011 Somalia crisis, Maxwell and Majid's "Another Humanitarian Crisis in Somalia?: Learning from the 2011 famine" goes straight to the point. An easy read that serves also as a primer for humanitarian dynamics in the country.

Intro:
"In July 2014, humanitarian agencies and the government of Somalia warned of a new severe drought in Somalia, only three years after more than 250,000 people died in a deadly famine. In a report for Tufts University and the Rift Valley Institute, Daniel Maxwell and Nisar Majid examine the lessons arising from the international response to the famine in 2011 on how to prevent and mitigate a similar crisis."  

Key points from the paper:
- Early Warning was there despite some recent complains on the quality/reliability of the data. The failure points more to the humanitarian framework, that has a bias towards response rather than prevention/mitigation. How much is forecasting really integrated into planning and funding still remains to be seen, and as we have seen in previous papers, take up of forecasting is different by sector (El Nino oscillations) even if the potential positive preventive effects are quite high (Violence in Liberia). In some cases we do enter in a Chronic Early Warning that lead to an institutionalization of preparedness, mitigation and response (a humanitarian example is Haiti and the Hurricane Season), but I don't believe Somalia is there yet.
- Risk Management, Access... and of course Al-Shabaab. While Al-Shabaab's behavior (internal taxes and blocking of external aid) did exacerbate the crisis, the risk management mechanisms of international actors also limited the response. Reliance on local organizations of variable quality and allegiance coupled with limited direct implementation/monitoring capacity, has meant that many of the interventions were/are shaped by risk management rather that needs and effectiveness. 
- Us, us , us. Most of the discussions around the 2011 crisis are around the international response; yet, in a time were resilience is the buzzword of the day, very little attention is given to the Somali response and how it can be further strengthened. The paper also points out the reinforcement of marginalization for certain clans through the crisis (that further feeds into Somalia's instability by clan competition and Al-Shabaab comparative appeal).
 
Following Amartya Sen's thinking, we can say that famine is not a food failure but rather a political failure (in the wider sense of governance and allocation of resources).

Investigating El Niño‐Southern Oscillation and Society Relationships - Working Paper

"Investigating El Niño‐Southern Oscillation and Society Relationships" by Zebiak et al, looks into climate forecasting on three sectors: water, agriculture and health. It shows the role that forecasting can have as part of planning and response in areas that have large societal impact (beyond the extreme weather events associated in the media with El Niño or La Niña). Abstract:  

"Throughout at least the past several centuries, El Niño‐Southern Oscillation (ENSO) has played a significant role in human response to climate. Over time, increased attention on ENSO has led to a better understanding of both the physical mechanisms, and the environmental and societal consequences of the phenomenon. The prospects for seasonal climate forecasting emerged from ENSO studies, and were first pursued in ENSO studies. In this paper, we review ENSO's impact on society, specifically with regard to agriculture, water, and health; we also explore the extent to which ENSO‐related forecasts are used to inform decision making in these sectors. We find that there are significant differences in the uptake of forecasts across sectors, with the highest use in agriculture, intermediate use in water resources management, and the lowest in health. Forecast use is low in areas where ENSO linkages to climate are weak, but the strength of this linkage alone does not guarantee use. Moreover, the differential use of ENSO forecasts by sector shows the critical role of institutions that work at the boundary between science and society. In a long‐term iterative process requiring continual maintenance, these organizations serve to enhance the salience, credibility, and legitimacy of forecasts and related climate services."

ENSO studies started in the 80's with an important growth in the 90's that led to a better understanding of its consequences and also the development of forecasting tools. How have these tools translated into improved decision-making? By looking at three sectors that are affected by ENSO in a wider geographical area (from the Indian Ocean to the American continent, the authors show the different societal intake of forecasting tools.

Sectors that have previously used weather forecasting (like agriculture and water management) have been the ones that adapted ENSO modelling the most. Previous knowledge of similar tools allowed for users to be comfortable with the models and were conceptually more ready. Entering forecasting in sector that have not previously used it may require a tailored approach with simple quick wins over a longer period of time rather than complex models.

Institutions matter (a much repeated point in development literature), not only their presence (by using already established networks to distribute knowledge like agricultural extension) but also their design (focus on policy, research, awareness raising, best practice implementation, centralized vs decentralized, etc...). Tools have to be seen as relevant to the user, institutional design tell us not only the mandate or objectives but, maybe more importantly, who the user actually is.

Linked to the previous point, forecasts are a decision-making tools should help those making decisions. These seems obvious, but we have to remember that in many instances the users of forecasts are not the final beneficiaries of a decision. For example, a water management board makes decisions that affect water users by managing water flows. The forecasts itself will have to compete with other dynamics within the institution in order to be effective. On the other hand, in the case of agriculture, the farmer may be the recipient of the forecast, the decision-maker and the final beneficiary. That shortened decision-making process may help to explain why it is in agriculture where the uptake of ENSO forecasting tools has been the strongest as the incentives are aligned.

 

Whether we are interested in climate events or not, this paper does point out interesting issues on the usage of forecasting in different settings and sectors, and present lessons learnt that could be transferred to other sectors. Maybe the title is over-ambitious with the "society relationships" part.  

Can Civilian Attitudes Predict Civil War Violence? - Afghanistan Working Paper

Another forecasting working paper, this time on Afghanistan :"Can Civilian Attitudes Predict Civil War Violence?" by Hirose, Imai and Lyall. This paper also looks at counter-counterinsurgency (if that word exists), what is, probably, an understudied phenomenon but key in protracted conflicts where each contender keeps learning and adapting to each other strategies.

 Abstract:     

"Are civilian attitudes a useful predictor of patterns of violence in civil wars? A prominent debate has emerged among scholars and practitioners about the importance of winning civilian "hearts and minds" for influencing their wartime behavior. We argue that such efforts may have a dark side: insurgents can use pro-counterinsurgent attitudes as cues to select their targets and tactics. We conduct an original survey experiment in 204 Afghan villages to establish a positive association between pro-International Security Assistance Force attitudes and future Taliban attacks. We then extend analysis to 14,606 non-surveyed villages to demonstrate that our measure of civilian attitudes improves out-of-sample predictive performance by 20-30% over a standard forecasting model. The results are especially strong for Taliban attacks with improvised explosive devices. These improvements in predictive power remain even after accounting for possible confounders, including past violence, military bases, and economic assistance."

The overall finding may be intuitively easy to accept: insurgents and counterinsurgents are trying to gain control over an area, therefore a successful "heart and minds" will trigger a response. However, the study highlights not only the dynamic nature of targeting but also of tactics and how exposure of risk varies (from targeted to indiscriminate attacks).While a single study a theory might not make, it is clear that insurgency targeting and decision making processes need further analysis and understanding.

  

The findings of the study do have many policy and programming consequences:

In many cases our risk matrix identifies things that could go wrong but rarely the risk of success and how to mitigate it (not the success but the risk). Do No Harm approaches do have some understanding of this but usually as a preventative measure (insurgents may prevent us implementation) and not necessarily as a result.

The forecasting power of the model is also quite interesting. Again, not only for counter-insurgency actors but also for development actors. If we can forecast the use of specific tactics as per changes in attitudes, implementation tools and modalities can vary over time to adapt to the changing risk. I.e. triggering Mine Risk Education activities (focused on IEDs) or switching to small group/household level meetings (to avoid large gatherings) in locations when an attitude indicator reaches certain level. Early Warning systems can also benefit from this model.

On the issue of transferability, I think the concept/methodology does lend itself to be used in other contexts, but of course the need for adapting it to the local insurgency tactics (not all insurgencies may respond the same way to successful hearts and minds due to operational, political or social constrains). It is also important to accept the adaptive nature of tactics and that they don't remain fix in time. Tactics evolve over time, sometimes as a response, sometimes due to technology/knowledge transfer; therefore both the programming and the forecasting teams will need to revise and analyze the model and results regularly.

Finally, development and security need to go together in the cases where development can be understood as undermining insurgencies' hold on a contested location. Many (most) NGOs make the claim of neutrality and also assume that working directly with beneficiaries or local 'communities' is also neutral. However, this may not be the way they are perceived by either local population and insurgents (and counterinsurgents for that matter). Ignorance on how our operations may alter local power dynamics not only puts our staff at risk, but also the wider population, as the study shows.

Predicción de violencias locales en Liberia [Spanish Version]

[Spanish version of previous post/ Versión española de articulo en inglés]

Es interesante que apenas utilicemos modelos de predicción cuando programamos en prevención de conflictos en ámbito comunitario. La mayoría de las veces o bien utilizamos un conjunto de supuestos básicos o tomamos conflictos pasados ​​como indicadores de los futuros (que, la verdad, tiende a ser un indicador bastante fiable). Un paso alentador es el nuevo estudio de Blair, Blattman y Hartman: Predicción de violencias locales en Liberia [en inglés]. A continuación está el sumario:

"Utilizamos los modelos de pronóstico y nuevos datos de 242 comunidades de Liberia para demostrar que es posible predecir los brotes de violencia local con alta sensibilidad y precisión moderada, incluso con datos limitados. Capacitamos a nuestros modelos para predecir la violencia comunitaria y penal en 2010 utilizando factores de riesgo medidos en 2008 comparamos las predicciones a la violencia actual en 2012 y encontramos que hasta un 88% de toda la violencia se predijo correctamente. Verdaderos positivos vienen a costa de muchos falsos positivos, dando precisión global entre el 33% y el 50%. Desde una perspectiva política, los estados, las organizaciones internacionales y las fuerzas de paz podrían utilizar este tipo de predicciones para prevenir y responder mejor a la violencia. Los modelos también generan nuevos datos estilizados que la teoría necesita explicar. En este caso, los predictores más fuertes de más violencia son divisiones sociales (principalmente étnicas), y coaliciones con grupos minoritarios" [Original en inglés, Google Translate con correcciones]

Los puntos principales, en mi opinión, son los siguientes:

- Han identificado como relevantes un número limitado de indicadores clave. Esto puede significar que incluso pequeños programas/organizaciones serian capaces de reunir los datos necesarios. Puede que los resultados sean específicos de Liberia, o incluso a las regiones bajo estudio (la pregunta habitual de: ¿Es transferible a otros contextos?), pero la metodología parece prometedora.

- ¿! Qué pasa con el reparto del poder?!?! Es posible que tengamos que investigar más en detalle sobre la "calidad" de coaliciones y acuerdos, y en qué manera afectan las dinámicas de poder. ¿Elevan las expectativas de los grupos minoritarios y dan lugar a más conflictos? o ¿los grupos mayoritarios no terminan de aceptar la presencia de los grupos minoritarios? ¡Tal vez deberíamos enfocar el problema desde la teoría de juegos! 

- ¿Son más precisos nuestros poderes de predicción (intuición) que el modelo a la hora de seleccionar los lugares de trabajo en las etapas de diseño e implementación? Las herramientas de predicción puede ser especialmente útiles cuando hay nuevo personal con conocimientos conceptuales o contextuales limitados (facilitando así el proceso inicial de aprendizaje) y en el proceso de diseño

- La idea de que el modelo auto-aprenda o sea capaz de generar nuevos indicadores o datos. Esto puede ayudar en la identificación de nuevas tendencias (no teorizadas) a lo largo del tiempo. En cierto modo esto está conectado con el hecho de que data-mining parece tener una mala imagen y, sin embargo, se las arregla de vez en cuando para sorprendernos y plantear nuevas preguntas.

- Y, por supuesto, la pregunta que está viene de mi lado practico/terreno: ¿cómo puedo utilizar la previsión como base de referencia para la medición del impacto en la prevención de conflictos? Aparte de la cuestión de atribución, aceptaran los donantes los conflictos inexistentes (¿prevenidos?) pero que fueron pronosticados anteriormente? ¿Van a aceptar realmente la previsión como herramienta de planificación? Como hemos visto en Somalia, en el caso de la sequía y la seguridad alimentaria, las atenciones se agudizan en las primeras etapas de las crisis pero no cuando se pronosticó o incluso el inicio mismo.

Predicting Local Violence in Liberia Working Paper

It is interesting to note that we barely use forecasting models for conflict prevention programming at community level. Most of the time we either use a set of basic assumptions or chasing past conflicts as predictors of future ones (that, to be fair, tends to be a reliable one). An encouraging step is the new working paper from Blair, Blattman and Hartman: Predicting Local Violence. Below is the abstract:


"We use forecasting models and new data from 242 Liberian communities to show that it is to possible to predict outbreaks of local violence with high sensitivity and moderate accuracy, even with limited data. We train our models to predict communal and criminal violence in 2010 using risk factors measured in 2008. We compare predictions to actual violence in 2012 and find that up to 88% of all violence is correctly predicted. True positives come at the cost of many false positives, giving overall accuracy between 33% and 50%. From a policy perspective, states, international organizations, and peacekeepers could use such predictions to better prevent and respond to violence. The models also generate new stylized facts for theory to explain. In this instance, the strongest predictors of more violence are social (mainly ethnic) cleavages, and minority group power-sharing"

Main takeaways, from my point of view, are:
- They have identified a limited number of key indicators (so much for "complex root causes") as relevant. These may mean that even relatively small programmes/organizations could gather the data. The results may be Liberia specific, or even limited to the areas of the study (the usual How transferable this thing is? question), but the methodology seems promising
- What is it with power-sharing?!?! we may need to look further into the "quality" of the sharing and how that affects power dynamics. Does it raise expectations of minority groups and lead to more conflict? or majority groups do not accept the influence of minority groups? Maybe we should throw in some game theory!
- How much more accurate are our 'gut-feeling' predictive powers when selecting locations at the design and implementation stages? Forecasting tools may be specially useful when new staff with limited conceptual or contextual knowledge joins (it will help with the steep learning curve of the first months) in the design process
- The idea of machine learning or generation by the models of new facts. It may also help in identifying new trends over time. In a way this is connected to the fact that data-mining seems to have a bad image, yet it manages from time to time to surprise us and lead us to new questions.
- And of course, the question that is in the practitioner in me: how can I use forecasting as a baseline for impact measurement of conflict prevention? Aside from the attribution issue, will donors be on-board for non-existent conflicts (prevented?) that where previously forecast? Will they really accept forecasting as a planning tool? As we have seen from Somalia on the case of drought and food security, attentions sharpen at the early stages of the crises not when it was forecast or even the onset.


HT to the always recommendable Chris Blattman (One of the authors)