Development acupuncture (Part I): Using network science to understand multidimensional poverty

Multidimensional poverty as a dynamic system

We know that the experience of poverty goes beyond not having enough money to meet your basic needs. It encompasses many dimensions, ranging from commonly measured deprivations in education, health, and living standards to less visible but equally significant dimensions such as social exclusion, institutional maltreatment, and disempowerment. Traditional measures of multidimensional poverty have come a long way in building the evidence base needed to identify which people in a society are poor and may need targeted and integrated support to overcome these challenges.  The policy responses, however, have not yet been able to account for the dynamic way in which these dimensions interact with one another. We know, for example, if there is poor sanitation in your home, your health is likely to suffer. Or if there is no electricity in your home, it will be difficult to study at night and your education may suffer. Cambridge economist Partha Dasgupta noted this “complexity of causation” behind poverty traps almost two decades ago.

Despite the complex interlinkages across these dimensions, policy responses typically remain compartmentalized within sector-specific silos, limiting their effectiveness. Education ministries may not have input on electrification decisions that could affect students' ability to study at night, and social development ministries might be excluded from discussions on connectivity or health policies. Sector-specific approach overlooks the potential synergies and interactions between different areas of development. As a result, policies may not fully address the multifaceted nature of poverty, missing opportunities for integrated and more impactful interventions.
 

The role of economic complexity

Economic complexity methods can provide valuable methodologies for understanding the intricate web of connections that constitute poverty. These methods use machine learning and network science to analyze the structures that shape an economy and infer how different elements within it interact. While they do not establish causal relationships, they serve as a diagnostic tool, highlighting potential risks and outcomes based on the observed structure of an economy. By applying these network science based measures to the study of poverty, we can begin to distinguish the complex relationships between different dimensions of deprivation.

While this type of analysis was originally deployed to understand productive structures, like industrial diversification, it is now being used to analyze  other aspects of development—such as how countries are progressing towards the Sustainable Development Goals. 
 

Mapping the Poverty Space and measuring Poverty Centrality

To effectively apply economic complexity to the issue of poverty, we introduce two concepts:

• The Poverty Space is a visual representation that maps out the structural relationships between various dimensions of poverty. It helps us see how different aspects of deprivation are interconnected.
• The Poverty Centrality measure, by contrast, assigns an importance score to each dimension within the network. This score reflects the prominence of each indicator and its potential as a strategic intervention point.

By targeting these key nodes, similar to acupuncture points in the human body, policymakers can potentially create ripple effects that enhance the overall system's health.
 

What does this look like in different countries?

Applying this approach to 67 developing countries at two points in time using data from the OPHI/UNDP Global Multidimensional Poverty Index (based on underlying survey data from DHS and MICS), we find a surprising similarity in the structure and stability of countries’ multidimensional poverty networks over time.

To illustrate the practical application of these tools, let's examine Ethiopia's Poverty Space at two different points in time: 2011 and 2019. During this period, Ethiopia experienced a reduction in its multidimensional poverty index . Our analysis, however, shows that during this time the structure of multidimensional poverty did not change much as the Poverty Space was very similar in both years. Certain dimensions, such as access to cooking fuel and housing, consistently appeared as central nodes within the network. This does not imply that they are more important than other dimensions like nutrition or child mortality. Instead, it indicates a higher level of structural relationships with other poverty dimensions within Ethiopia's specific context.
 

The Poverty Space of Ethiopia in 2011 and 2019

^{The Poverty Space of Ethiopia in 2011 and 2019. In these visualizations, the nodes represent poverty indicators, whereas the edges between two poverty indicators highlight a significant structural dependency. Moreover, in these networks, the size of the nodes is proportional to its poverty centrality, whereas the color of a node proportional to the censored headcount ratio of the indicator.}
 

Thinking holistically about policy

Identifying these structural linkages and central nodes within the poverty network has significant implications for policy formulation. By recognizing these critical areas, policymakers can prioritize interventions that are likely to have substantial effects across multiple dimensions of poverty. This approach encourages a more integrated and strategic policy framework, moving away from isolated actions towards coordinated efforts that address the root causes of poverty in a holistic manner.

As we look forward to Part II of this blog series, we will explore how this network structure can be integrated into a forward-looking policy analysis framework, guiding decision-makers towards more cohesive and impactful approaches to combating poverty.

This blog post draws on the research paper "Development acupuncture: The network structure of multidimensional poverty and its implications" by Viktor Stojkoski, Luis F. Lopez-Calva, Kimberly Bolch, and Almudena Fernandez.