Special focus

• Author: Giulio Caperna - Eleni Papadimitriou
• Pages: 17-19

17

Figure 6. Sensitivity anal ysis on Structure (comparison of Ranks)

Note: Only Areas with a shift larger than 5 ranks are labelled.

Sourc e: European Commission, Jo int Research Centre, 2020.

5 Speci al focus

5.1 Impact of the missing dimensio ns on the weights

In the methodological report, the developers describe the system of weighting of the 15 dimensions with three

levels (50% for dimensions 1-5, 33% for 6-10 and 17% for 11-15). This st ructure is clear and intuitive . However,

the framework of the IDM is meant to be generic and adaptable to different countries. In particular, as discussed

in section 3.2, IDM Fiji is not tak ing into account dimensions 10 and 11, which leads to a different scheme where

the 33% and the 17% weights of the last two groups, are shared among 4 dimensions instead of 5.

Consequently, the respective distribution of dimensions’ weights is 10% , 8.3% and 4.2% for the first five, middle

four and last four. This means that the relat ive impact of t he dimensions from 6th to 15th is im proved and it

looks like the 4 remaining dimensions in each group act as substitutes for the removed one.

An alternat ive would be kee ping the original weights t o each dimension (10%/6.6%/3.4%) for the thirteen

remaining ones and then adjust them to sum to unity. The re sulting set of weights would t hen be

(11%/7.4%/3.7%)5, as shown in Table 3. This approach distributes the weight of the removed dimensions to all

the other dimensions, proportionally to their nominal weight.

Figure 7 compares the scores resulting from the IDM weights and those resulting from the suggestion of the

JRC. While Figure 8 represents the ranks of the same scores.

5 To adjust the weights to have sum one, it is sufficient to divide each weight by the sum of the weights.

The original weig hts with 15 dimensio ns are: (0.10, 0.10, 0. 10, 0.10, 0.10, 0.06 6, 0.066, 0.066, 0.0 66, 0.066, 0.033, 0.0 33, 0.033,

0.033, 0.033). When 10 and 11 are removed, at the moment they become (0.10, 0.10, 0.10, 0.10, 0.10, 0.083, 0.083, 0.083, 0.083,

0.044, 0.044 , 0.044, 0.044). Alt ernatively, we sug gest to take the ori ginal weights o nly for the thirteen dimensions, and di vide by their

sum, that is 0.9. Obtaining t he following set of weight s: (0.11, 0.11, 0.11, 0.11, 0. 11, 0.074, 0.074, 0.074, 0. 074, 0.037, 0.037, 0.037,

0.037).