Search Results

1 - 2 of 2 items

  • Author: Milica Bulajić x
Clear All Modify Search
Shedding Light on the Doing Business Index: a Machine Learning Approach


Background: The World Bank (WB) acknowledged the importance of business regulatory environment and therefore created a metric which ranks 190 countries based on their level of business regulation for domestic firms measured by the Doing Business Index (DBI).

Objectives: The question which attracted our attention is whether all the observed entities should be given the same weighting scheme.

Methods/Approach: The approach we propose as an answer is two-fold. First, we cluster the countries covered by the DBI. In the next step, we apply the statistical multivariate Composite I-distance Indicator (CIDI) methodology to determine new, data-driven weights for each of the retained clusters.

Results: The obtained results show that there is a difference between the weighting schemes proposed by the CIDI methodology.

Conclusions: One can argue that one weighting scheme does not fit all the observed countries, meaning that additional analyses on the DBI are suggested to explore its stability and its weighting scheme.

Open access
Multivariate approach to imposing additional constraints on the Benefit-of-the-Doubt model: The case of QS World University Rankings by Subject


Composite indexes have become a valuable asset for stakeholders as they provide ranks of entities and information upon which decisions are made. However, certain questions about their development procedure have been raised recently, especially regarding the weighting process. To tackle the observed issue, in this paper we propose a new multivariate approach for defining weights. Namely, the model based on the Data Envelopment Analysis (DEA), the Benefit-of-the-Doubt (BoD) model, has been used with significant success in the process of composite index development. On the other hand, the Composite I-distance Indicator (CIDI) methodology stands out as an impartial method for assigning weights to indicators. By combining these two approaches, some of the limitations of the original BoD model could be overcome. As a result, new entity-specific weights which maximize the value of the composite index can be proposed. As a case study, we analysed the Quacquarelli Symonds (QS) World University Rankings by Subject in the field of statistics and operational research. The obtained results, which are based on the data-driven weights, can provide new insights into the nature of the observed ranking. The approach presented here might provoke further research on the topic of composite index weights and on the university rankings by subject.

Open access