The Index and its sub-components are calculated as aggregations through the tiers of its data model. As per the principles set out in our article on methodological principles we have aimed to ensure the Index has a simple and transparent methodology that makes it easy for external users to understand how data flows from the original sources to the overall Index.

Normalising source data into metrics

The lowest tier of the Index’s data model are “metrics”, which are translations of original source data. Metrics are calculated by first collating the processed and standardised source data into a single dataset, this dataset is then subset to the countries selected for inclusion in the Index. The source data is then normalised into metrics by scaling the data from 0 to 1, such that 0 represents the lowest performance of group of countries and 1 represents the highest performance:

• The majority of metrics are scaled using “min-max” normalisation where the lowest score in the source data is translated to 0 and the highest score is translated to 1.
• For some metrics the source must be inverted, that is their highest score represents worse performance than their lowest score and thus the highest score is translated to 0 while the lowest score is translated to 1.
• For some metrics both the highest and lowest score in the source data reflects “worse performance” compared to a reference figure. For example, gender parity in employment where the ideal would be a 50:50 split between men and women in the workforce. For these metrics the distance from the reference value is calculated and then this distance is scaled from 0 to 1 such that values closest to the reference point are scored 1 and those furthest from the reference point are scored 0.
• The Bertelsmann Transformation Index (BTI) and the Sustainable Governance Indicators (SGI) are related data projects with differing country coverage that is somewhat influenced by economic development levels. Having compared the data in these sources a differential scaling is applied such that data from the BTI is scaled from 0 to 0.8 and data from the SGI is scaled from 0.3 to 1.

Our article on re-scaling and transformation has more details about how we approach this.

Aggregation through the data model

After transforming the source data into metrics, subsequent tiers of the data model are calculated as the simple arithmetic mean of its constituent parts:

• Indicators are estimated as the mean of their constituent metrics. For example, the impartial behaviour indicator is calculated as the mean of the four metrics that are aligned to it in the data model and for which there is data (policy implementation is impartial; decisions free of interference; respect for due process; officials impartial in their duties).
• Themes are calculated as the mean of their constituent indicators. For example, the integrity theme is calculated as the mean of the four indicators aligned to it in the data model and for which there is data (impartial behaviour; corruption; sanctions; and, integrity data).
• Domains are calculated as the mean of their constituent themes. For example, the Strategy and Leadership domain is calculated as mean of the four themes aligned to it in the data model and for which there is data (strategic capacity; openness and communications; integrity; and, innovation).
• Finally, the Index is calculated as a mean of the four domains (Strategy & Leadership; Public Policy; National Delivery; and, People and Processes).

At each stage of calculation values are rounded to 2 decimal places to reduce the impact of any spurious precision arising from either the scaling of the original source data or calculation through the various aggregation levels.

The rescaling of data to the 0-1 scale occurs only when converting source data to metrics, the aggregations into the higher tiers of the data model are not rescaled. Therefore, the theoretical minimum of the Index (0) represents the situation that a country being the lowest performing country in all data sources it is present in and the theoretical maximum (1) represents the situation that a country being the highest performing country in all data sources it is present in.

Handling of missing data

As outlined in our article on data coverage, all but one country included in the Index has some degree of missing data. To ensure the methodology remains as simple and transparent as possible it was decided not to make any imputation of missing data. Based on the aggregation methodology this implies that a country’s performance in any missing data is equivalent to its average performance in its observed data points.

Weighting of data

As with the handling of missing data, to maintain the simplicity and transparency of the methodology it was decided not to apply any explicit weighting to the data. While there is no explicit weighting, the structure of the data model creates an implicit weighting structure. However, each country has its own pattern of missing data and therefore the actual weighting of an individual metric on a country’s scores will vary depending on what data is does and does not have. The implicit weight for individual metrics varies from 0.179% to 6.25%, read our separate article on the implied weighting structure.