In simple terms, how is the aggregation methodology carried out to produce Governance estimates?
We use an Unobserved Component Model (UCM) to aggregate the various response in the broad 6 clusters. This model treats the “true” level of governance in each country as unobserved, and assumes that each of the available sources for a country provide noisy “signals” of the level of governance. The UCM then constructs a weighted average of the sources for each country as the best estimate of governance for that country. The weights are proportional to the reliability of each source. The resulting estimates of governance have an expected value (across countries) of zero, and a standard deviation (across countries) of one. This implies that virtually all scores lie between -2.5 and 2.5, with higher scores corresponding to better outcomes. For technical details, consult the Governance Matters III paper (pages 8-12) and/or Appendix D of the Governance Matters VII paper.
We use an Unobserved Component Model (UCM) to aggregate the various response in the broad 6 clusters. This model treats the “true” level of governance in each country as unobserved, and assumes that each of the available sources for a country provide noisy “signals” of the level of governance. The UCM then constructs a weighted average of the sources for each country as the best estimate of governance for that country. The weights are proportional to the reliability of each source. The resulting estimates of governance have an expected value (across countries) of zero, and a standard deviation (across countries) of one. This implies that virtually all scores lie between -2.5 and 2.5, with higher scores corresponding to better outcomes. For technical details, consult Appendix D of the Governance Matters VIII paper.
