This annex provides an overview of the methodological notes for the data and evidence used in this report. Further information on the methodology for the States of Fragility report series is available on the States of Fragility data and visualisation platform: http://www3.compareyourcountry.org/states-of-fragility/about/0/. This webpage also contains a link to the underlying data and statistical code (produced in R and R Studio) for the production of the fragility framework. Andrew Etchell and David Hammond (Institute for Economics and Peace) peer-reviewed and validated the statistical code to ensure the accuracy and reliability of the findings. Further information on the details of this peer review is available upon request.
The term “fragile contexts” refers to the 60 contexts identified in the 2022 edition of the OECD’s multidimensional fragility framework. Unless otherwise indicated, the grouping of “other developing contexts” refers to the countries and territories on the DAC List of ODA Recipients for reporting on aid in 2021, available at https://www.oecd.org/dac/financing-sustainable-development/development-finance-standards/DAC-List-ODA-Recipients-for-reporting-2021-flows.pdf.
The methodology for the OECD’s multidimensional fragility framework
The OECD characterises fragility as the combination of exposure to risk and insufficient coping capacities of the state, system and/or communities to manage, absorb or mitigate those risks. The OECD’s multidimensional fragility framework, introduced in the 2016 edition of States of Fragility, measures fragility on a spectrum of intensity across six dimensions: economic, environmental, human, political, security and societal. It relies on a mixed methods approach that examines contexts within each dimension and then aggregates this information to obtain an overall picture of fragility.
The methodology is based on a two-stage principal components analysis (PCA), with a hierarchical clustering procedure to group contexts according to similar characteristics in each dimension. The foundation is 57 indicators derived from independent third-party data sources, all of which are recorded and explained in greater detail on the States of Fragility platform. Each of the six dimensions contains nine to ten indicators that are aggregated into principal components in the first stage PCA; the first two principal components in each dimension are used for the second stage PCA. The first principal component that results from this second stage PCA represents the overall fragility score for each context. Based on this score, a context is classified as either fragile, if its score is lower than -1.20, or extremely fragile, if the score is lower than -2.85. This analysis assesses fragility across 176 contexts for which sufficient data were available, as denoted by data being available for a context for at least 70% of indicators.
All regional and subregional estimates of fragility were calculated using a population-weighted average of all contexts within the respective region or subregion. Population statistics were sourced from UN DESA (2019[1]), using both the “Estimates” and “Medium Variant” (e.g. for 2021). Regional and income group classifications were based on the World Bank (2022[2]) Country and Lending Groups. An extensive discussion of this methodology is available in Annex A of the working paper accompanying this publication by Desai and Yabe (2022[3]) and on the States of Fragility platform, including the step-by-step process for the PCA and hierarchical clustering procedure as well as the methodological notes and caveats regarding the data collected for the analysis. Additional information is available upon request.
Financial statistics
Unless otherwise stated, all aid statistics cited in this report are deflated to USD constant (2020) and represented in USD million disbursements. They are sourced from the OECD aid statistics database (OECD, 2022[4]), specifically the DAC2a and Creditor Reporting System.
The sources of other financial statistics are cited in the text, using the most recent values – usually 2020. Due to data limitations, not all data are available for all contexts. Where values have been imputed, they use the latest available value or a simple average of the last three years, as indicated. In time series, projected values are identified with a “p”, and estimates are identified with an “e”. Unless otherwise indicated, all financial statistics are deflated using the 2020 “Total DAC” deflator. Values after 2020 have not been deflated.