The COIN Tool is available at https://composite-indicators.jrc.ec.europa.eu/?q=coin-tool.

COIN is an abbreviation of “Composite Indicator”. COIN is also the abbreviated name of the European Commission’s Competence Centre for Composite Indicators and Scoreboards, at the Joint Research Centre, which created the COIN Tool.

See Section 1.3 for an example of some usage cases. At an absolute minimum, you need to specify your index in the Database and Framework tabs. We always recommend to check your data in the Statistics, Winsorisation and TreatedData tabs, and to check relationships in the IndCorrel tab. You may then see your scores in the Heatmap, Scores and Rankings tabs.

• Data coverage: A rule of thumb is at least 65% data for every unit and indicator. Anything less than that, and you should consider removing indicators or units, unless there are compelling reasons not to. Imputation is effectively informed guesswork, so more data coverage is always better.

• How many indicators: There is a balance between including many indicators to try to fully cover the concept, and having a streamlined framework with good data coverage and easy interpretation. Typically, a framework might include 30-40 indicators and still be easy to interpret. Consider that each indicator represents a chunk of information, and there is only so much information that can be packed into a single number (your index), so as the number of indicators increases, the amount of information retained in the final score from each indicator will decrease. This is why it is better not to build a very large framework, unless you do not intend to aggregate all indicators to a single index. On the other hand you should also be careful to include enough indicators to give a reasonable picture of the overall concept.

• Number of aggregation levels: This should be defined by breaking down the concept you are trying to measure into dimensions and sub-dimensions. Many frameworks simply consist of two aggregation levels: indicators -> pillars -> index. Complex frameworks may make use of all four aggregation levels of the COIN Tool. It is quite rare in practice that an index exceeds four aggregation levels.

• Which aggregation and normalisation method to use: The default methods here are the arithmetic average and the min-max normalisation method respectively. These are the easiest to understand and communicate. We would recommend to depart from these only if there is a good reason to do so.

• Skewness and kurtosis thresholds: The recommended thresholds are 2 and 3.5 respectively, which represent a roughly normal distribution. See Box 4 for more information.

• How many points to Winsorise: This should be taken in context of how many units you have. If you have 100 units and you Winsorise five, this is only 5% of the total and should have a small impact on the overall distribution. If you only have ten units and you Winsorise five, this has a much larger impact. Caution should be exercised in particular when Winsorising more than 10% of the total number of units.

Visit our website at composite-indicators.jrc.ec.europa.eu/ where you can a number of resources on methodology, including the JRC/OECD Handbook on Constructing Composite Indicators, many reports and audits, and academic papers.

We also organise the following events:

• An annual training course and community of practice on composite indicators (“JRC Week on Composite Indicators and Scoreboards”) - see the Events page on our website

• COIN Open days (see COIN Open Days on our website), where organisations are able to come to the JRC in Ispra and receive feedback and assistance their composite indicator from the COIN team

• Ad-hoc training on request, subject to availability

Please contact jrc-coin@ec.europa.eu for further information.

Please cite the User Guide as: Becker, W., Benavente, D., Dominguez Torreiro, M., Moura, C., Neves, A., Saisana, M., Vertesy, D., COIN Tool User Guide, 2019, ISBN 978-92-76-12385-9, doi:10.2760/523877