Decision Analysis at the World Agroforestry Centre
Land Health D
Development’s decision-making dilemma
Decision-making in development occurs in an environment characterized by risk, uncertainty and
imperfect information. Moreover, the complexity of the systems decision-makers strive to influence
requires them to consider in their decision processes a host of factors – biophysical, socioeconomic and
political – so that comprehensive science-based decision support requires a transdisciplinary research
approach.
Unfortunately, it is virtually impossible to
completely understand complex socioecological
systems, because time and resources for learning
about the systems are always limited. Decisionmakers are invariably faced with knowledge gaps
and uncertainty, and they should take account of this
fact in the way they make decisions. In order to align
itself with the needs of decision-makers, decisionsupporting research should also be able to adequately
address issues of complexity and uncertainty. When
researchers ignore these issues, or when they assume
certain values for variables that are really highly
Agricultural landscapes are complex and decision impacts
uncertain, their advice can easily lead decisionare hard to predict
makers onto the wrong track.
Decision Analysis at the World Agroforestry Centre
The Land Health Decisions group at the World Agroforestry Centre tackles the problem of development
decision-making through approaches from the Decision Sciences, where the problem of decisionmaking in risky environments with limited information has been the subject of decades of research. Our
group partners with leading experts in this field, including Douglas Hubbard of Hubbard Decision
Research, as well as Norman Fenton and Martin Neil at Queen Mary University of London. These
partnerships have enabled us to enrich the research for development arena with robust strategies that are
widely used in many fields, such as business decision support, medical drug development, environmental
legislation and legal reasoning.
Our group works directly with decision-makers to
conduct research that is specific to the concrete decisions
they face. In close collaboration with decision-making
entities, experts and stakeholders, we develop decision
models that aim at including all factors deemed
important by those implementing or affected by a
particular decision. These factors are integrated into
stakeholder-informed causal impact pathways for
decisions, which link them to quantitative development
outcomes. Since decision impacts can never be known
ex-ante with absolute certainty, our approach accounts
for the current state of uncertainty about all variables and We work directly with decision-makers to ensure that
their objectives, views and values are considered in
parameters involved in a decision. It propagates these
decision models
uncertainties through decision models, converting them
into projections of the range of plausible decision outcomes, as opposed to making precise predictions
that hinge on tenuous assumptions. Decision outcomes can be projected for multiple stakeholders and
different social groups. They can be provided in their natural units (e.g. reduction of poverty, amount of
carbon sequestered) or monetized, which allows anticipation of the return on investment or value for
money that decisions can be expected to deliver.
Decision Analysis at the World Agroforestry Centre
A key principle of our work is ‘value of information’ (VoI), which describes the reduction in decisionmaking uncertainty that can be expected from measurements on particular variables. For most decisions,
this VoI varies greatly across variables. It typically identifies a small number of variables as decisionspecific research priorities, while assigning very little or no value to measurements on the majority of
uncertain inputs to a decision model. The VoI principle allows devising decision-specific measurement
activities, but it can also identify more general research priorities, through applying decision analysis
approaches to interventions and decisions at larger scale, e.g. the promotion of sustainable land
management strategies or implementation of global policy mechanisms.
Land Health D
Decisions that are supported by our work can be binary decisions – e.g. whether to implement a particular
intervention – but the approach is also
amenable to other types of decisions, such as
spatial targeting of interventions or
prioritizing among multiple investment
options for maximizing impact prospects. We
also use these procedures for project
monitoring and evaluation, as well as for
defining appropriate development indicators
that are informative for tracking development
progress while at the same time providing
guidance for concrete development decisions.
Our projects
Farmers constructing a terrace for erosion control. Is this a good
investment?
We have applied decision analysis approaches in a number of cases over the past few years. Working
with fellow scientists, we anticipated the impacts of several strategic decisions in water, land and
ecosystem management, including water interventions and payments for ecosystem services in Kenya,
a hydroelectric dam in Laos, human waste recycling in Ghana and a new seed supply system in West
Africa. Even with all uncertainties considered,
some interventions were clearly identified as
likely success stories, while for others the
prospects of positive outcomes were slim.
Variables identified as priorities for decisionsupporting research included many unexpected
ones, such as the effects of an intervention on
human migration, the impact of a dam on
fisheries in the Mekong River, or the future
carbon price. Most variables that are routinely
measured by research for development activities
had little or no information value for the
Will building a new pipeline alleviate poverty and generate
growth? Who will benefit, and who won’t?
decisions they aimed to support.
Working directly with stakeholders and decision-makers, we evaluated the prospects of a water supply
pipeline project in Northern Kenya, taking into account all costs and benefits, monetary or nonmonetary, as well as all risks involved in the decision. Decision impacts were modeled for multiple
stakeholders, to detect and help resolve potentially inequitable distribution of project benefits. This
project helped all involved grasp the bigger picture of the planned decision, seeing beyond simply the
hydrological feasibility of the plans and including environmental, financial and equity issues into their
thinking. Model results indicated that this project was highly risky, with a good chance of net benefits,
but also substantial likelihood of undesirable results. Key uncertain variables were, among others, the
valuation of benefits, such as reduced infant mortality, the risk of political interference in project
implementation, and the financial feasibility of operating the water supply system. We also build
decision models for selected agricultural interventions in Kenya, including the introduction of droughttolerant crops, fruit trees, water-saving technologies, improved value chains and bee-keeping.
Decision Analysis at the World Agroforestry Centre
Another line of our work is the development of spatial targeting tools for interventions that adequately
consider uncertainties, can capture elements of intervention impact pathways and work with incomplete
information. Such tools, based on Bayesian Belief Networks, are currently being developed for climatesmart agriculture in Tanzania. The objective of this work is to help governments and other decisionmaking entities select options from among multiple possible interventions and decide on where to deploy
them for the greatest possible impact.
Land Health D
The way forward
A number of new case studies are currently being initiated,
for example on projecting the nutrition impacts of
agricultural interventions in East Africa, and decisions
around water structures in West Africa. These projects will
provide additional opportunities for us to refine and improve
our procedures. Our long-term vision is the widespread
application of robust decision analysis procedures in
development. We are convinced that this could be a gamechanger in the way development projects are planned,
implemented, monitored and evaluated. We aim to realize
our vision by packaging rigorous decision analysis
procedures that can be disseminated widely through
structured capacity building mechanisms. A cohort of welltrained development decision analysts emerging from such
mechanisms could make great contributions to sustainable
development.
Further reading:
Decision Analysis can consider multiple
development objectives, allowing holistic
evaluation of intervention options
Papers in scientific journals:
Shepherd K, Hubbard D, Fenton N, Claxton K, Luedeling E, De Leeuw J, 2015. Development goals should enable
decision-making. Nature 523, 152-154: http://www.nature.com/news/policy-development-goals-shouldenable-decision-making-1.17915
Luedeling E, Oord A, Kiteme B, Ogalleh S, Malesu M, Shepherd K, De Leeuw J, 2015. Fresh groundwater for
Wajir – Analysis of stakeholder uncertainty in a water supply project in Northern Kenya. Frontiers in
Environmental Science 3, 16: http://journal.frontiersin.org/article/10.3389/fenvs.2015.00016/abstract
Rosenstock TS, Mpanda M, Aynekulu E, Kimaro A, Neufeldt H, Shepherd K, Kristjanson P, Luedeling E, 2014.
Targeting conservation agriculture in the context of livelihoods and landscapes. Agriculture, Ecosystems
and Environment 187, 47-51: http://www.sciencedirect.com/science/article/pii/S0167880913004052
Reports:
Clapp A, DauSchmidt N, Millar M, Hubbard D, Shepherd K, 2013. A survey and analysis of data requirements for
stakeholders in African agriculture. World Agroforestry Centre, Nairobi, Kenya: http://r4d.dfid.
gov.uk/pdf/outputs/misc_susag/60993-DFID_Open_Data_Survey_Report_with_hyperlinks.pdf
Hubbard Decision Research, 2014. Global Intervention Decision Model – Using Applied Information Economics
for developing world agricultural research: https://cgspace.cgiar.org/bitstream/handle/10568/35093/
Report%20on%20the%20WLE%20Intervention%20Decision%20Modelling.pdf?sequence=1
Blogs:
Luedeling E, Göhring L, Shepherd K, 2015. A new tool for making big decisions without perfect information.
Agroforestry World Blog: http://blog.worldagroforestry.org/index.php/2015/05/26/a-new-tool-formaking-big-decisions-without-perfect-information/
Luedeling E, De Leeuw J, Shepherd K, 2014. Do you trust your gut instinct? Agriculture and Ecosystems Blog:
http://wle.cgiar.org/blogs/2014/09/17/trust-gut-instinct/
Contacts
Eike Luedeling: [email protected]
Keith Shepherd: [email protected]
© World Agroforestry Centre, July 2015