Statistics for international development and humanitarian action

What are the data-related challenges that those working in international development and humanitarian action face? How can statisticians contribute to solving these issues?

This is the focus of some work that I am exploring on behalf of the International Development Working Group (IDWG) of the Royal Statistical Society (RSS).

I am particular interested in understanding the challenges required to provide a strong evidence base to inform future international development and humanitarian action. In this post I want to briefly outline my understanding of the key challenges and issues. In later posts I will expand on various challenges in more detail.

First, the motivation for this work was that the RSS IDWG has initially focused on how to support and improve official statistics for international development. Here, I am looking outside the role of official statistics to the role of data more generally to support the needs of different organisations and people. This includes those working in:

• Intergovernmental/multi-lateral and regional groups – for example the World Bank or the United Nations and it agencies such as the World Health Organisation and the World Food Program
• Governmental organisations – for example the Department for International Development (DfID)
• (International) non-governmental organisations (I)NGOs – such as Save the Children or WaterAid
• (International) research institutes and academic organisations – for example the CGIAR global partnership of 15 centres of agricultural research or the Overseas Development Institute.

I am mainly considering the needs of those working in policy, advisory and research roles and those working as practitioners who manage and implement programs in the field.

A broad characterisation of the work of these groups where there are statistical challenges would be:

• Description. Providing estimates of the number of people, or the state of, a particular situation. How many children are currently suffering from severe actor malnutrition in Myanmar? How many people are living in slavery today? How strong are child protection systems in humanitarian settings?

• Monitoring. Measuring how something is changing over time. This could be: monitoring as an early warning system, for example the Famine Early Warning System (FEWS); working with communities to monitor corruption in the provision of public services; or monitoring the progress of a program to improve livelihoods using a combination of small-scale irrigation schemes, conservation agricultural practices, improved livestock health and providing enterprise development opportunities.

• What works? Evaluation and impact assessments – identifying what does and does not work and why.  These are becoming more common as a demand for a rigorous evidence base is required. For example: a randomized control trial to investigate whether unconditional cash transfers improve the economic and psychological well-being of poor rural households? Investigating whether adaptive co-management of artisanal fisheries leads to better sustainability of fish stocks and socio-economic conditions than other fishery management strategies. Understanding the impact of an individual (I)NGO’s response to a humanitarian disaster?

Of course many problems can shift from one category to another. For example a monitoring programme could be a series of snapshots or descriptions of a situation. Furthermore, projects can be monitored whilst being implemented and an evaluation may take place at the end which builds on the on-going monitoring.

A few of the common themes are:

• Complex systems: Often the situation being studied is complex. The complexity may be because there are many different  components to the problem are of interest – for example sustainability of fish stocks and the socio-economic conditions. Or there may be many different components that need to be accounted for to understand the impact of the component you are interested in – for example the role of other (I)NGOs in responding to a humanitarian disaster. These different components may all interact. The key challenge is how to make sense of this type of problem so that it is possible to identify (1) what you measure and control for (2) how you run the program or data collection exercise and (3) how you combine information and data to answer the question of interest.

• Scaling up: How do you use the results from one particular situation to apply it elsewhere?  In many ways, this is the essence of statistical inference – how can you infer something about a population (however that is defined) given the data or sample that has been obtained?  Sample selection and issues of how to interpolate or extrapolate from the particular situation need to be considered.

• Measurement difficulties: The population to be monitored may be difficult to observe – for example those in slavery. Or the key concepts you are trying to measure, for example corruption, resilience, sustainability, well-being cannot be directly or easily measured. Instead we need a strategy to observe different aspects or outcomes of these underlying unobservable concepts (in statistical terms we might call these latent variables) and combine them.

• Data from different sources: The available data may be from many different sources which may not be collected for quite the purpose that it is now needed. There may be official statistics, data collected by citizens and data collected by other agencies and organisations. These may all be collected on different spatial and temporal scales and be on a different scale to the one required. How should all these different sets of data be brought together?

• Beneficiaries and participatory methods: There is a recognition that the views of beneficiaries are necessary and should have a high priority. In addition, there is a view that participatory methods, not just simple surveys are the answer to this. There continue to be statistical challenges in how to collect and use this information.

The issues above seem to me to be ones that statisticians could contribute to. Are these areas that you think are important and/or are there others that are missing?