Wednesday 7
Decision-support systems
Pierre Bommel
› 11:30 - 12:30 (1h)
› JOFFRE 1-5
Assessing Resilience in the Horn of Africa – an Applied Information Economics Approach
Eike Luedeling  1@  , Katharine Downie  2@  , Keith Shepherd  1@  
1 : World Agroforestry Centre  (ICRAF)  -  Website
UN Avenue, Gigiri, 00100 Nairobi -  Kenya
2 : International Livestock Research Institute  (ILRI)  -  Website

The ability of communities in the drylands of Africa to adapt to or manage shocks and stresses while maintaining trajectories towards sustainable development, or “resilience”, features prominently in the objectives of donors and development partners. This concept is, however, poorly understood, particularly when it comes to measuring the impact of investments on resilience. Most definitions of resilience in development scenarios hinge upon the response of social, ecological and economic systems to shocks and stressors. It is, however, extremely difficult to quantify this response, as it is impossible to observe the full range of possible disturbances, hence assessments of system resilience normally fall short of providing comprehensive evaluations.

In this paper, we discuss the use of system models to overcome this constraint in the chronically food insecure region of the Horn of Africa. As this method allows a simulation of system responses to the full range of plausible disturbances, it then becomes possible to produce long-term system performance projections that convey insights into system resilience. Complex systems are, however, difficult to model, and there are always substantial uncertainties about model parameters, model structure and consequently model outputs. We therefore propose a modeling approach that is adequate for dealing with multiple uncertainties and that produces assessments of system resilience that can be used in planning for development activities aimed at enhancing resilience.

Our method draws on Applied Information Economics, a well-established approach to supporting business decisions under multiple uncertainties. Rather than aiming to specify system dynamics with high precision - the level of precision rarely justified by available information - this approach emphasizes the adequate framing of uncertainties with respect to data concerning input parameters, system processes and potentially even resilience definitions. Estimates of confidence intervals for input variables are obtained from historic and current datasets, as well as solicited from calibrated experts. Probability distributions of all input variables are used for constructing high-level system models that are then subjected to Monte Carlo analysis. Outputs from this procedure are analyzed with data mining techniques to determine which uncertainties constrain our ability to evaluate resilience and are thus priorities for further research.

This method also permits the identification of potential indicators of resilience that can be specifically targeted by interventions or used for monitoring resilience on the ground. Monitoring such key indicators along their respective quantified impact pathways over time provides a means of accumulating evidence indicating which outcomes can be attributed to a particular intervention. Following this logic we can also identify which variables and impact pathways cumulatively can be attributed to an outcome defined as resilience. Monitoring uncertain variables provides the opportunity for early corrective action during intervention implementation. For example, if actual adoption rates fall behind projected adoption rates, then the reasons for this can be further modeled and measures put in place to correct course. If the desired outcomes are achieved but actual adoption rates remained well below those projected to be required, then it is unlikely that the outcomes are attributable to the intervention.

Online user: 1 RSS Feed